0:00 - I watched my agent happily click the "I'm not a robot" button. I made the agent
0:04 very aware. Like, it knows what his source code is. It understands th- how it sits and runs in its own
0:11 harness. It knows where documentation is. It knows which model it runs. It understands its own system that made it very easy for an agent
0:19 to... Oh, you don't like anything? You just prompted it to existence, and then the agent would just modify its own software. People
0:26 talk about self-modifying software, I just built it. I actually think wipe coding is
0:30 a slur. - You prefer agentic engineering? - Yeah, I always tell people I'd- I do agentic engineering, and then maybe after
0:37 3:00 AM, I switch to wipe coding, and then I have regrets on the next day.
0:40 - What a walk of shame. - Yeah, you just have to clean up and, like, fix your sh- shit.
0:45 - We've all been there. - I used to write really long prompts. And
0:50 by writing, I mean, I don't write, I- I- I talk, you know? These- these hands are,
0:54 like, too- too precious for writing now. I just- I just use bespoke prompts to build my software.
1:00 - So, you, for real, with all those terminals, are using voice?
1:04 - Yeah. I used to do it very extensively, to the point where there was a period where I lost my voice.
1:13 - I mean, I have to ask you, just curious. I- I know you've probably
1:16 gotten huge offers from major companies. Can you speak to who you're considering working with?
1:27 - Yeah. - The following is a conversation with Peter Steinberger, creator
1:34 of OpenClaw, formerly known as MoldBot, ClawedBot, Clawdus, Claude, spelled with a W as in
1:42 lobster claw. Not to be confused with Claud, the AI model from
1:45 Anthropic, spelled with a U. In fact, this confusion is the reason Anthropic kindly asked Peter to
1:53 change the name to OpenClaw. So, what is OpenClaw? It's an open-source AI agent that has taken over the
2:00 tech world in a matter of days, exploding in popularity, reaching over
2:06 180,000 stars on GitHub, and spawning the social network mold book, where AI agents post manifestos and
2:14 debate consciousness, creating a mix of excitement and fear in the
2:18 general public. And a kind of AI psychosis, a mix of clickbait fearmongering and genuine, fully
2:25 justifiable concern about the role of AI in our digital, interconnected human world. OpenClaw, as its tagline
2:32 states, is the AI that actually does things. It's an autonomous AI assistant that lives in your computer, has access to
2:40 all of your stuff, if you let it, talks to you through Telegram,
2:44 WhatsApp, Signal, iMessage, and whatever else messaging client.
2:48 Uses whatever AI model you like, including Claude Opus 4.6 and GPT 5.3 Codex,
2:56 all to do stuff for you. Many people are calling this one of the biggest moments in the recent history of AI, since the launch of
3:04 ChatGPT in November 2022. The ingredients for this kind of AI agent were all there, but putting it all together in a
3:12 system that definitively takes a step forward over the line from language to agency, from ideas to actions, in a way that
3:20 created a useful assistant that feels like one who gets you and
3:24 learns from you, in an open source, community-driven way, is
3:28 the reason OpenClaw took the internet by storm. Its power, in large part, comes from the fact that you can give it access to all of your
3:36 stuff and give it permission to do anything with that stuff in
3:40 order to be useful to you. This is very powerful, but it is also dangerous. OpenClaw
3:47 represents freedom, but with freedom comes responsibility. With it, you can own and have control
3:54 over your data, but precisely because you have this control, you
3:57 also have the responsibility to protect it from cybersecurity
4:01 threats of various kinds. There are great ways to protect yourself, but the threats and vulnerabilities are out there.
4:09 Again, a powerful AI agent with system- level access is a security minefield, but it also represents the future. Because
4:16 when done well and securely, it can be extremely useful to each of us humans as a personal assistant. We discuss all of this with Peter, and
4:26 also discuss his big-picture programming and entrepreneurship life story, which I think is truly
4:32 inspiring. He spent 13 years building PSPDF Kit, which is a software used on a billion
4:39 devices. He sold it, and for a brief time, fell out of love with programming, vanished for three years, and
4:47 then came back, rediscovered his love for programming, and built,
4:51 in a very short time, an open source AI agent that took the internet by storm. He is, in many ways, the symbol
4:59 of the AI revolution happening in the programming world. There was the
5:03 ChatGPT moment in 2022, the DeepSeek moment in 2025, and now, in '26, we're living
5:10 through the OpenClaw moment, the age of the lobster. The start of the agentic AI revolution. What a time to be
5:20 alive. This is a Lex Fridman podcast. To support it, please check out our sponsors in the description, or you can also find links to contact
5:27 me, ask questions, give feedback, and so on. And now, dear friends, here's Peter Steinberger.
5:36 The one and only, the Clawed Father. Actually, Benjamin predicted it in his tweet. "The following is a conversation with
5:43 Claude, a respected crustacean." It's a hilarious-looking picture of a lobster in a suit, so I think the
5:51 prophecy has been fulfilled. Let's go to this moment when you built a prototype in one
5:58 hour, that was the early version of OpenClaw. I think this, Story's really inspiring to a lot of people because this
6:06 prototype led to something that just took the internet by storm....
6:10 and became the fastest-growing repository in GitHub history,
6:13 with now over 175,000 stars. So, what was the story of the one-hour prototype?
6:20 - You know, I wanted that since April. - A personal assistant. AI personal assistant.
6:25 - Yeah. And I, I played around with some other things, like even
6:31 stuff that gets all my WhatsApp, and I could just run queries
6:35 on it. That was back when we had GPT-4.1, with the one million context
6:41 window. And I, I pulled in all the data and then just asked him questions
6:47 like, "What makes this friendship meaningful?" - Mm-hmm. - And I got some, some really profound
6:55 results. Like, I sent it to my friends and they got, like, teary eyes.
6:59 - So, there's something there. - Yeah. But then I... I thought all the labs will, will,
7:05 will work on that. So I, I moved on to other things, and that was still very much in my early days of
7:12 experimenting and pl- playing. You know, you have to... That's how you learn. You just like, you do stuff and you play. And
7:22 time flew by and it was November. I wanted to make sure that
7:26 the thing I started is actually happening. I was annoyed that it didn't exist, so
7:30 I just prompted it into existence. - I mean, that's the beginning of the hero's journey of the entrepreneur,
7:39 right? And you've even with your original story with PS PDF kit, it's like, "Why does this not exist? Let me build
7:46 it." And again, here's diff- a whole different realm, but similar maybe spirit.
7:52 - Yeah, so I had this problem. I tried to show PDF on an iPad, which should
7:56 not be hard. - This is like 15 years ago, something like that.
7:59 - Yeah. Like the most, the most random thing ever. And suddenly, I
8:03 had this problem and I, I wanted to help a friend. And there was, there was... Well, not like nothing existed, but it was just not good.
8:11 And like... Like I tried it and it was like very, "Nah." Like, "Hmm, I can do this better."
8:17 - By the way, for people who don't know, this led to the development of PS PDF
8:21 kit that's used on a billion devices. So, the... It turns out that
8:25 it's pretty useful to be able to open a PDF. - You could also make the joke that I'm really bad at naming.
8:32 - Yeah. - Like, name number five on the current project. And even
8:35 PS PDF doesn't really roll from the tongue. - Anyway, so you said "Screw it. Why don't
8:43 I do it?" So what was the... What was the prototype? What was the thing that
8:47 you... What was the magical thing that you built in a short amount of time that you were like,
8:51 "This might actually work as an agent," where I talk to it and it does things?
8:55 - There was... Like, one of my projects before already did something
8:59 where I could bring my terminals onto the web and then I could, like, interact with them, but there also would be terminals on my Mac.
9:07 - Mm-hmm. - Viptunnel, which was like a, a weekend hack project
9:12 that was still very early. And it was cloud code times. You know, you got a dopamine hit when you got something right. And now
9:20 I get, like, mad when you get something wrong. - And you had a really great -– not to take a tangent -– but a great blog post describing that
9:26 you converted Viptunnel. You vibe-coded Viptunnel from TypeScript into Zig of all programming languages with
9:34 a single prompt. One prompt, one shot. Convert the entire code base into Zig.
9:41 - Yeah. There was this one thing where part of the architecture
9:46 was... Took too much memory. Every terminal used like a node. And I wanted to change it to Rust
9:54 and... I mean, I can do it. I can, I can manually figure it all out, but all my automated attempts failed miserably.
10:08 And then I revisited about four or five months later. And I'm
10:12 like, "Okay, now let's use something even more experimental." And
10:16 I, and I just typed, "Convert this and this part to Sig," and then let Codex run off. And
10:23 it basically got it right. There was one little detail that I had to,
10:27 like, modify afterwards, but it just ran for overnight or like six hours and just did its thing. And it's
10:36 like... It's just mind-blowing. - So that's on the LLM programming side, refactoring. But uh, back to the
10:46 actual story of the of the prototype. So how did Viptunnel connect to the first
10:50 prototype where your, like, agents can actually work? - Well, that was still very limited. You know, like I had this one
10:56 experiment with WhatsApp, then I had this experiment, and both felt like
11:01 not the right answer. And then my search bar was literally just hooking up WhatsApp to
11:10 cloud code. One shot. The CLI message comes in. I call the CLI with -p. It does its
11:17 magic, I get the string back and I send it back to WhatsApp. And I, I built
11:21 this in one hour. And I felt... Already felt really cool. It's like, "Oh, I could... I can, like, talk to my
11:28 computer," right? This... That, that was, that was cool. But
11:32 I, I wanted images, 'cause I alw- I often use images when I prompt. I think it's such a, such an efficient way to give the agent more
11:39 context. And they are really good at figuring out what I mean, e- even if it's like
11:43 a, a weird cropped-up screenshot. So I used it a lot and I wanted to do
11:47 that in WhatsApp as well. Also, like, you know, just you run around, you see like
11:53 a poster of an event, you just make a screenshot and like figure out if I have time there, if this is good, if my friends are maybe up for that.
12:00 Just like images seemed im- important. So I, I worked a few... It took me a few more hours to actually get that right. Um,
12:09 and then it was just...... I, I used it a lot. And funny enough, that was
12:17 just before I went on a trip to Marrakesh with my friends for a birthday trip. And there it was even better because
12:26 internet was a little shaky but WhatsApp just works, you know? It's like doesn't
12:30 matter, you have, like, edge, it still works. WhatsApp is just... It's just made really well. So I ended up using it a lot. Um,
12:41 translate this for me, explain this, find me places. Like, you
12:45 just having a clanker doing, having Google for you, that was... Basically there was still nothing built but
12:52 it still could do so much. - So, if we talk about the full journey that's happening there with the agent,
12:58 you're just sending on this very thin line WhatsApp message via CLI, it's going to a cloud code and cloud code is doing all kinds of
13:08 heavy work and coming back to you with a thin message. - Yeah. It was slow because every time I boot up the CLI, but it... It was really cool
13:19 already. And it could just use all the things that I already had
13:23 built. I had built like a whole bunch of CLI stuff over the month so it, it felt
13:30 really powerful. - There is something magical about that experience that's hard to put into
13:34 words. Being able to use a chat client to talk to an agent, versus, like, sitting behind
13:44 a computer and like, I don't know, using cursor or even using Cloud Code CLI in the terminal. It's a different experience than being
13:51 able to sit back and talk to it. I mean, it seems like a trivial step
13:55 but, it- in some sense it's a... It's like a phase shift in the integration of AI into your life and how it feels, right?
14:05 - Yeah. Yeah. I, I read this tweet this morning where someone said, "Oh, there's
14:09 no magic in it. It's just like, it does this and this and this and this and this and this." And it
14:16 almost feels like a hobby, just as cursor or perplexity. And I'm
14:20 like, well, if that's a hobby that's kind of a compliment, you
14:23 know? They're like, they're not doing too bad. Thank you I guess?
14:32 Yes. I mean, isn't, isn't, isn't magic often just like you take a lot of things that are already there
14:39 but bring them together in new ways? Like, I don't... There's
14:43 no... Yeah. Maybe there's no magic in there but sometimes just
14:46 rearranging things and, like, adding a few new ideas is all the magic that you need.
14:51 - It's really hard to convert into words what is, what is magic
14:55 about a thing. If you look at the, the scrolling on an iPhone, why is that so pleasant? There's a lot of elements about that
15:02 interface that makes it incredibly pleasant, that is fundamental to the experience of
15:06 using a smartphone, and it's like, okay, all the components were
15:10 there. Scrolling was there, everything was there. - Nobody did it-
15:14 - Yep - ... and afterwards it felt so obvious. - Yeah, so obvious.
15:16 - Right? But still... You know the moment where it, it blew my mind was when,
15:25 when I- I used it a lot and then at some point I just sent it a message
15:29 and, and then a typing indicator appeared. And I'm like, wait, I
15:35 didn't build that, it only m- it only has image support, so what is it even doing? And then it would just reply.
15:42 - What was the thing you sent it? - Oh, just a random question like, "Hey, what about this in this restaurant?" You
15:47 know? Because we were just running around and checking out the city.
15:52 So that's why I, I didn't, didn't even think when I used it because
15:56 sometimes when you're in a hurry typing is annoying. - So, oh, you did an audio message?
16:00 - Yeah. And it just, it just worked and I'm like... - And it's not supposed to work because-
16:05 - No - ... you didn't give it that- - No, literally - ... capability.
16:08 - I literally went, "How the fuck did he do that?" And it was like, "Yeah,
16:12 the mad lad did the following. He sent me a message but it only, only was a file and no file ending." So
16:19 I checked out the header of the file and it found that it was,
16:23 like, opus so I used ffmpeg to convert it and then I wanted to use whisper but it didn't had it installed. But then I found the
16:31 OpenAI key and just used Curl to send the file to OpenAI to translate and here I am.
16:39 Just looked at the message I'm like, "Oh wow." - You didn't teach it any of those things and the agent just figured it out, did all those conversions,
16:47 the translations. It figured out the API, it figured out which program to
16:51 use, all those kinds of things. And you were just absent-mindedly just sent an
16:54 audio message when it came back. - Yeah, like, so clever even because he would have gotten the whisper local path, he would have had to download a
17:00 model. It would have been too slow. So like, there's so much world
17:04 knowledge in there, so much creative problem solving. A lot of it
17:08 I think mapped from... If you get really good at coding that means you have to
17:12 be really good at general purpose problem solving. So that's a skill, right? And
17:16 that just maps into other domains. So it had the problem of like, what is this file with no file ending? Let's figure it
17:23 out. And that's when it kind of clicked for me. It's like, I was like very
17:30 impressed. And somebody sent a pull request for Discord support and I'm like, "This is a WhatsApp relay.
17:37 That doesn't, doesn't fit at all." - At that time it was called WA Relay.
17:42 - Yeah. And so I debated with me like, do I want that? Do I not want that? And then
17:51 I thought, well maybe, maybe I do that because that could be a cool way to show people. Because I... So far I did it in WhatsApp
18:00 as like groups you know but don't really want to give my phone number to every internet stranger.
18:07 - Yeah. - Um, journalists manage to do that anyhow now so that's a different
18:11 story. So I merged it-... from Shadow, who helped me a lot with the whole project. So, thank you. And, and I put
18:24 my, my bot in there. - On Discord? - Yeah. No security because I didn't... I hadn't built sandboxing in
18:31 yet. I, I just prompted it to, like, only listen to me. And then some people came and tried to hack it, and I
18:41 just... Or, like, just watched and I just kept working in the open, you
18:45 know? Like, y- I used my agent to build my agent harness and to test, like, various stuff. And that's
18:57 very quickly when it clicked for people. So it's almost like it needs to
19:01 be experienced. And from that time on, that was January the 1st, I, I got
19:09 my first real influencer being a fan and did videos, dachitze. Thank you. And, and
19:17 from there on, I saw, I started gaining up speed. And at the same time, my,
19:23 my sleep cycle went shorter and shorter because I, I felt the storm
19:29 coming, and I just worked my ass off to get it to... into a state where it's kinda
19:37 good. - There's a few components and we'll talk about how it all works, but basically, you're able to
19:42 talk to it using WhatsApp, Telegram, Discord. So that's a component that you have to get right.
19:48 - Yeah. - And then you have to figure out the agentic loop, you have to have the gateway, you
19:52 have the harness, you have all those components that make it all just work nicely.
19:56 - Yeah. It felt like Factorio times infinite. - Right. - I, I feel like I built my little- ... my little
20:04 playground. Like, I never had so much fun than building this project. You know?
20:08 Like, you have like, "Oh," I go like, level one agentic loop. What can I do there?
20:12 How can I be smart at queuing messages? How can I make it more
20:15 human-like? Oh, then I had this idea of... Because the loop always... The agent always replies something, but you don't
20:23 always want an agent to reply something in a group chat. So I gave him this
20:27 no-reply token. So I gave him an option to shut up. So it, it feels more natural.
20:32 - That's level two. - Y- uh, yeah, yeah. Yeah, on the- on the-
20:36 - Factorio. - On the agentic loop. And then I go to memory, right?
20:39 - Yeah. - You want him to, like, remember stuff. So maybe, maybe the
20:42 end... The ultimate boss is continuous reinforcement learning,
20:46 but I'm, I'm, like, at... I feel like I'm level two or three with Markdown
20:50 files and the vector database. And then you, you can go to level
20:54 community management, you can go to level website and marketing. There's just so many hats that you have to have on. Uh,
21:01 not even talking about native apps. That's just, like, infinite
21:05 different levels and infinite level ups you can do. - So the whole time you're having fun. We should say that for the most
21:11 part, throughout this whole process, you're a one-man team. There's people helping, but you're doing so much of the key core development.
21:21 - Yeah. - And having fun? You did, in January, 6,600 commits. Probably more.
21:28 - I sometimes posted a meme. I'm limited by the technology of my time. I could do more
21:32 if agents would be faster. - But we should say you're running multiple agents at the same time.
21:37 - Yeah. Depending on how much I slept and how difficult of the tasks I work on, between four and 10.
21:45 - Four and 10 agents. Uh there's so many possible directions, speaking of
21:49 Factorio, that we can go here. But one big picture one is, why do you think
21:56 your work, Open Claw, won? In this world, if you look at 2025, so many
22:05 startups, so many companies were doing kind of agentic type stuff, or claiming to. And here, Open Claw comes
22:12 in and destroys everybody. Like, why did you win? - Because they all take themselves too serious.
22:18 - Yeah. - Like, it's hard to compete against someone who's just there to have fun.
22:24 - Yeah. - I wanted it to be fun, I wanted it to be weird. And if you see, like,
22:28 all the, all the lobster stuff online I think I, I managed weird. I... You
22:36 know, for the longest time, the only, the only way to install it was git clone, pnpm build, pnpm gateway.
22:46 Like, you clone it, you build it, you run it. And then the, the agent... I made the agent very aware. Like, it
22:53 knows that it is... What its source code is. It understands th- how it sits and runs in its own
23:01 harness. It knows where documentation is. It knows which model it runs. It knows if you turn on the voice or, or reasoning mode. Like,
23:11 I, I wanted to be more human-like, so it understands its own system that made it very easy for an agent to...
23:18 Oh, you don't like anything? You just prompted it to existence, and then the agent would
23:22 just modify its own software. Um, you know, we have people talk about self- modifying software. I just built it and
23:30 didn't even... I didn't even plan it so much. It just happened.
23:35 - Can you actually speak to that? 'Cause it's just fascinating. So you have this
23:39 piece of software that's written in TypeScript- - Yeah - ... that's able to, via the agentic loop, modify
23:47 itself. I mean, what a moment to be alive in the history of humanity and the history of programming. Here's
23:55 the thing that's used by a huge amount of people to do incredibly powerful things in their lives, and that very
24:02 system can rewrite itself, can modify itself. Can you just, like, speak to the power of that? Like, isn't
24:10 that incredible? Like, when did you first close the loop on that?
24:14 - Oh, because that's how I built it as well, you know? Most of it is built by
24:18 Codex, but oftentimes I... When I debug it, I...... I use self-introspection so much.
24:26 It's like, "Hey, what tools do you see? Can you call the tool yourself?" Or like, "What error do
24:30 you see? Read the source code. Figure out what's the problem."
24:32 Like, I just found it an incredibly fun way to... That the agent, the very agent and software that you
24:40 use is used to debug itself, so that it felt just natural that everybody does that. And that it led to so many,
24:51 so many pull requests by people who never wrote software. I mean, it
24:54 also did show that people never wrote software . So I call them prompt requests
24:58 in the end. But I don't want to, like, pull that down because every time someone made the first pull request is a
25:06 win for our society, you know? Like, it... Like, it doesn't matter
25:10 how, how shitty it is, y- you gotta start somewhere. So I know there's, like, this whole big movement of people
25:17 complain about open source and the quality of PRs, and a whole different level of
25:21 problems. But on a different level, I found it... I found it
25:28 very meaningful that, that I built something that people love to think
25:32 of so much that they actually start to learn how open source works.
25:37 - Yeah, you were ... The Open Cloud project was the first pull request. You were the first for so many. That is
25:44 magical. So many people that don't know how to program are taking their first step into the programming world with this.
25:52 - Isn't that a step up for humanity? Isn't that cool? - Creating builders.
25:56 - Yeah. Like, the bar to do that was so high, and, like, with
26:00 agents, and with the right software, it just, like, went lower and lower.
26:04 I don't know. I was at a... And I also organize another type of meetup. I call it... I called it Cloud Code Anonymous.
26:14 Uh, you can get the inspiration from. Now, I call it Agents Anonymous- ... for, for reasons.
26:23 - Agents Anonymous. - And- - Oh, it's so funny on so many levels. I'm sorry, go ahead.
26:29 - Yeah. And there was this one guy who, who talked to me. He's like, "I run this
26:33 design agency, and we, we never had custom software. And now I have, like, 25 little web services for various
26:41 things that help me in my business. And I don't even know how
26:45 they work, but they work." Uh, and he was just, like, very happy that
26:52 my stuff solved some of his problems. And he was, like, curious enough that he
26:56 actually came to, like, a, a Enchantic meetup, even though he's... He doesn't really know how software works.
27:04 - Can we actually rewind a little bit and tell the saga of the name change?
27:10 First of all, it started out as Wa-Relay. - Yeah. - And then it went to-
27:13 - Claude's. - Yeah. You know, when I, when I built it in the beginning, my agent had no personality.
27:19 It was just... It was Claude Code. It's like this sycophantic
27:23 opus, very friendly. And I... When you talk to a friend on WhatsApp, they don't
27:30 talk like Claude Code. So I wanted... I, I felt this... I just didn't f- It didn't feel right, so I, I wanted to give it
27:40 a personality. - Make it spicier, make it- - Yeah - ... something. By the way, that's actually hard to put into words as well. And we should mention
27:47 that, of course, you create the soul.md, inspired by Anthropic's constitutional AI work-
27:53 - Mm-hmm - ... how to make it spicy. - Partially, it picked up a little bit from me. You know, like those things
27:58 are text completion engines in a way. So, so I, I, I, I had fun working with it, and then I told it to... How I wanted it to
28:11 interact with me, and just, like, write your own agents.md, Give yourself a name. And then we... I didn't even know how the whole,
28:22 the whole lobster... I mean, people only do lobster... Originally, it was actually a lobster in
28:26 a, in a TARDIS, because I'm also a big Doctor Who fan. - Was there a space lobster?
28:31 - Yeah. - I heard. What's that have to do with anything? - Yeah, I just wanted to make it weird.
28:37 There was no... There was no big grand plan. I'm just having fun here.
28:40 - Oh, so I guess the lobster is already weird, and then the space lobster is an extra weird.
28:44 - Yeah, yeah, because the- - Yeah - ... the TARDIS is basically the, the harness, but
28:50 cannot call it TARDIS, so we called it Claude's. So that was name number two.
28:54 - Yeah. - And then it never really rolled off the tongue. So when more people came, again, I talked with my agent,
29:06 Claude. At least that's what I used to call him. Now- - Claude spelled with a W-C-L-A-U-D-E.
29:12 - Yeah. - Versus C-L-A-U-D-E from Anthropic. - Yeah. - Which is part of what makes it funny,
29:24 I think. The play on the letters and the words in the TARDIS and the
29:28 lobster and the space lobster is hilarious. But I can see why
29:32 it can lead into problems. - Yeah, they didn't find it so funny .
29:39 So then I got the domain ClaudeBot, and I just... I love the domain. And
29:45 it was, like, short. It was catchy. I'm like, "Yeah, let's do
29:48 that." I didn't... I didn't think it would be that big at this time.
29:55 And then just when it exploded, I got, Kudos, a very friendly email from one of the employees
30:06 that they didn't like the name. - One of the Anthropic employees.
30:11 - Yeah. So actually, Kudos, because they shou- could have just sent a, a lawyer letter, but they've been nice about
30:18 it. But also like, "You have to change this and fast." And I asked for two days,
30:26 because changing a name is hard, because you have to find everything, you
30:37 And everything has to be...... you need a set of everything.
30:41 - And also, can we comment on the fact that you're increasingly attacked, followed by
30:47 crypto folks? Which I think you mentioned somewhere that that means the name
30:51 change had to be... Because they were trying to snipe, they were trying to steal, and so you had to be... The, the na- I mean, from an
30:59 engineering perspective, it's just fascinating. You had to make the name change
31:03 Atomic, make sure it's changed everywhere at once. - Yeah. Failed very hard at that.
31:08 - You did? - I, I underestimated those people. It's a, it's a very
31:16 interesting subculture. Like, it... Everything circles around... I'll probably get a lot wrong and we'll probably get
31:24 hate for that if you say that, but... There is like Bags app and then they, they tokenize everything. And th- they did the
31:31 same back with Swipe Tunnel, but to a much smaller degree. It was not that annoying. But on this project, they've
31:39 been, they've been swarming me. They, they... It's like every half an hour,
31:46 someone came into Discord and, and, and spammed it and we had to block the p- We have,
31:50 like, server rules, and one of the rules was... One of the rules
31:54 is no mentioning of butter. For obvious reasons. And one was, no talk about finance stuff or
32:01 crypto. Because I'm... I- I'm just not interested in that, and this
32:08 is a space about the project and not about some finance stuff.
32:13 But yeah. They came in and, and spammed and... Annoying. And on Twitter, they would ping me all the time.
32:20 My, my notification feed was unusable. I, I could barely see actual people talking about this stuff because it was like swarms.
32:28 - Mm-hmm. - And everybody sent me the hashes. Um... And they all try me to
32:35 claim the fees. Like, "Are you helping the project?" Claim the fees. No,
32:39 you're actually harming the project. You're, like, disrupting my
32:43 work, and I am not interested in any fees. I'm... First of all, I'm financially comfortable. Second of
32:50 all, I don't want to support that because it's so far the worst form of online harassment that I've experienced.
32:59 - Yeah. There's a lot of toxicity in the crypto world. It's sad because
33:03 the technology of cr- cryptocurrency is fascinating, powerful and maybe
33:08 will define the future of money, but the actual community around that, there's so much to- toxicity, there's so much greed. There's so much
33:16 trying to get a shortcut to manipulate, to, to steal, to snipe,
33:20 to, to, to, to game the system somehow to get money. All this kind of
33:24 stuff that... Uh... I mean, it's the human nature, I suppose, when you
33:28 connect human nature with money and greed and and especially in
33:32 the online world with anonymity and all that kind of stuff. But from the
33:36 engineering perspective, it makes your life challenging. When Anthropic
33:39 reaches out, you have to do a name change. And then there- there's, there's like all these, like, Game of Thrones or Lord of the Rings
33:48 armies of different kinds you have to be aware of. - Yeah. There was no perfect name, and I didn't sleep for two nights.
33:55 I was under high pressure. Um, I was trying to get, like, a good set of
34:02 domains and, you know, not cheap, not easy, 'cause in this, in this state of the internet, you basically have to
34:10 buy domains if you want to have a good set. And, and then another ca- another email came in that the lawyers are getting uneasy.
34:22 Again, friendly, but also just adding more stress to my situation already. So at this point I was just like,
34:31 "Sorry, there's no other word. Fuck it." And I just, I just renamed it to
34:35 Mod Bot 'cause that was the set of domains I had. I was not really happy, but I thought it'll be fine.
34:43 And I tell you, everything that could go wrong- ... did go wrong. Everything that could go wrong did go wrong. It's incredible.
34:51 I, I, I thought I, I had mapped the h- the space out and reserved the important things.
34:58 - Can you ga- give some details of the stuff that gone wrong? 'Cause it's interesting from, like, an
35:01 engineering perspective. - Well, the, the interesting stuff is that none of these services have, have a squatter
35:07 protection. So, I had two browser windows open. One was like a,
35:13 an empty account ready to be rename- renamed to Claude Bot, and the other one I renamed to Mod Bot. So, I pressed
35:20 rename there, I pressed rename there, and in those five seconds,
35:24 they stole the account name. Literally, the five seconds of dragging the mouse over there and
35:30 pressing rename there was too long. - Wow. - Because there's no... Those systems... I mean, you would expect that they have some
35:37 protection or, like, an automatic forwarding, but there's nothing like that. And I didn't know that
35:46 they're not just good at harassment, they're also really good
35:49 at using scripts and tools. - Yeah. - So, yeah. So, suddenly, like, the old account
35:56 was promoting new tokens and serving malware. And I was like, "Okay, let's move over to GitHub,"
36:05 and I pressed rename on GitHub. And the GitHub renaming thing is slightly confusing, so I
36:13 renamed my personal account. And in those... I guess it took me 30 seconds to realize my mistake. They sniped
36:21 my account, serving malware from my account. So, I was like, "Okay, let's at least do the NPM stuff," but
36:30 that takes, like, a minute to upload. They sniped, they sniped the NPM package, 'cause I could reserve
36:38 the account, but I didn't reserve the root package.... so like
36:44 everything that could go wrong , like went wrong. - Can I just ask a, a curious question of, in that moment you're sitting
36:50 there, like how shitty do you feel? That's a pretty hopeless feeling, right?
36:57 - Yeah. Because all I wanted was like having fun with that project and to
37:04 keep building on it. And yet here I am like days into researching names, picking a name I didn't like.
37:11 And having people that claimed they helped me making my life miserable in every possible way. And honestly, I was
37:22 that close of just deleting it. I was like, "I did show you the future, you build it."
37:30 - Yeah. - I... That was a big part of me that got a lot of joy out of
37:35 that idea. And then I thought about all the people that already co-
37:38 contributed to it, and I couldn't do it because they had plans with it, and they put time in it. And it just didn't feel right.
37:50 - Well, I think a lot of people listening to this are deeply grateful that you
37:54 persevered. But it's... I, I can tell. I can tell it's a low point. This is the first time you hit a wall of, this is not fun?
38:02 - No, no, I was like close to crying. It was like, okay, everything's fucked.
38:10 - Yeah. - Um... I am like super tired. - Yeah. - Uh, and now like how do you even, how do you undo that? You know, l- luckily, and
38:22 thankfully, like I, I have... Because I have a little bit of following already. Like I had friends
38:28 at Twitter, I had friends at GitHub who like moved heaven and earth to like help me. And it is not... That's not something
38:35 that's easy. Like, like GitHub tried to like clean up the mess and then they ran into like platform bugs .
38:45 'Cause it's not happening so often that things get renamed on that
38:49 level. So, it took them a few hours. The MBM stuff was even more difficult because it's a whole different team.
38:57 On the Twitter side, things are not as easy as well. It, it took them like a day
39:04 to really also like do the redirect. And then I also had to like
39:13 do all the renaming in the project. Then there's also, uh, ClaudeHub, which I didn't
39:21 even finish the rename there because I, I, I managed to get people on
39:28 it and then someone just like collapsed and slept. And then I woke up and I'm like,
39:34 I made a, a beta version for the new stuff and I, I just, I just couldn't live with the name. It's like, you know... But
39:43 but, you know, it's just been so much drama. So, I had the real
39:47 struggle with me like I never want to touch that again, and I really don't like the
39:53 name. Um, so, and I... There was also this like... Then there was all the security people that started
40:03 emailing me like mad. Um, I was bombarded on Twitter, on email. There's like a thousand other things I should do.
40:13 And I'm like thinking about the name which is like, it should be
40:17 like the least important thing. Um, and then I was really close
40:24 in... Oh God, I don't even... Honestly, I don't even wanna say the, my other
40:32 name choices because it probably would get tokenized, so I'm not gonna say it.
40:38 - Yeah. - But I slept on it once more, and then I had the idea for OpenClaw
40:43 and that felt much better. And by then, I had the boss move that I
40:49 actually called Sam to ask if OpenClaw is okay. OpenClaw.AI. You know? 'Cause 'cause like-
40:57 - You didn't wanna go through the whole thing. Yeah. - Oh, that it's like, "Please tell me this is fine." I don't think
41:05 they can actually claim that, but it felt like the right thing to do.
41:11 And I did another rename. Like just Codex alone took like 10 hours to rename the
41:17 project 'cause it, it's a bit more tricky than a search replace and I, I wanted everything renamed, not just on the outside. And that
41:27 rename, I, I felt I had like my, my war room. But then I, I had like some contributors really that helped me. We made a whole plan
41:36 of all the names we have to squat. - And you had to be super secret about it?
41:40 - Yeah. Nobody could know. Like I literally was monitoring Twitter if like, if there's any mention
41:44 of OpenClaw. - Mm-hmm. - And like with reloading, it's like, "Okay, they don't, they don't expect anything
41:50 yet." Then I created a few decoy names. And all the shit I shouldn't have to
41:54 do. You know? Like, you know- - Yeah, yeah - ... it's helping the project. Like, I lost like 10 hours just by
41:59 having to plan this in full secrecy like, like a war game. - Yeah, this is the Manhattan Project of the 21st century. It's renaming-
42:08 - It's so s- ... so stupid. Uh like I still was like, "Oh, should I, should I keep it?"
42:12 Then I was like, "No, the mold's not growing on me." And then I think I had final all the pieces together. I didn't get a .com
42:23 but, yeah, it's been like quite a bit of money on the other domains. I tried to reach out again to
42:29 GitHub but I feel like I, I used up all my goodwill there, so I... 'Cause I, I, I wanted them to do this thing atomically-
42:39 - Mm-hmm - ... But that didn't happen and then so I did that the f- as first
42:41 thing. Uh, Twitter people were very supportive. I, I actually paid 10K for the
42:49 business account so I could claim the-... OpenClaw, which was, like, unused since 2016, but was claimed. And yeah, and then I
43:00 finally ... This time I managed everything in one go. Nothing, almost nothing got wrong. The only thing that did go wrong is that
43:11 I was not allowed by trademark rules to get OpenClaw.AI, and someone copied the website as serving malware.
43:21 - Yeah. - I'm not even allowed to keep the redirects. Like, I have to
43:27 return ... Like, I have to give Entropik the domains, and I cannot do redirects, so if you go on claw.bot next week, it'll just be a 404.
43:37 - Yeah. - And I- I'm not sure how trademark ... Like, I didn't,
43:44 I didn't do that much research into trademark law, but I think that could,
43:48 could be handled in a way that is safer, because ultimately those people will then Google and maybe find
43:59 malware sites that I have no control on them. - The point is, that whole saga, Made a dent in your whole f-
44:08 the funness of the journey, which sucks. So, let's just, let's just get, I suppose, get back to fun. And during this, speaking of
44:16 fun, the two-day MoltBot saga. - Yeah, two years. - MoltBook was created.
44:24 - Yeah. - Which was another thing that went viral as a kind of demonstration,
44:31 illustration of how what is now called OpenClaw could be used
44:37 to create something epic. So for people who are not aware, MoltBook is
44:41 just a bunch of agents talking to each other in a Reddit-style social network. And a bunch of people take
44:48 screenshots of those agents doing things like, Scheming against humans. And
44:56 that instilled in folks a kind of, you know, fear, panic, and
45:00 hype. W- what are your thoughts about MoltBook in general? - I think it's art. It is, it is like the finest slop, you know, just like the slop
45:12 from France. - Yeah. - I- I saw it before going to bed, and even though I was
45:21 tired, I spent another hour just reading up on that and, and just being entertained. I, I just felt
45:31 very entertained, you know? The- I saw the the reactions, and, like, there was one reporter who's calling me about, "This is the end of
45:40 the world, and we have AGI." And I'm just like, "No, this is just,
45:43 this is just really fine slop." You know, if, if I wouldn't have created this, this whole onboarding experience
45:50 where you, you infuse your agent with your personality and give him, give him character, I think that reflected on a lot of
46:01 how different the replies to MoltBook are. Because if it were all, if
46:05 it were all be ChatGPT or Cloud Code, it would be very different. It would be much more the same.
46:11 - Mm-hmm. - But because people are, like, so different, and they create their agents in so
46:16 different ways and use it in so different ways, that also reflects
46:20 on how they ultimately write there. And also, you, you don't know how much of that is really
46:27 done autonomic, autonomous, or how much is, like, humans being funny and, like, telling the agent, "Hey, write about the deep plan, the end of the world, on
46:34 MoltBook, ha, ha, ha." - Well, I think, I mean, my criticism
46:39 of MoltBook is that I believe a lot of the stuff that was screenshotted is human prompted. Which,
46:48 just look at the incentive of how the whole thing was used. It's obvious to me at least that a lot of it was humans
46:55 prompting the thing so they can then screenshot it and post it on
46:59 X in order to go viral. - Yeah. - Now, that doesn't take away from the artistic aspect of
47:04 it. The, the finest slop that humans have ever created . - For real. Like, kudos to, to Matt, who had this idea so quickly and pushed something
47:17 out. You know, it was, like, completely insecure security drama. But also,
47:24 what's the worst that can happen? Your agent account is leaked, and,
47:28 like, someone else can post slop for you? So like, people were,
47:32 like, making a whole drama about of the security thing, when I'm like, "There's nothing private
47:36 in there. It's just, like, agents sending slop." - Well, it could leak API keys.
47:41 - Yeah, yeah. There's like, "Oh, yeah, my human told me this and this, so I'm leaking his
47:45 security number." No, that's prompted, and the number wasn't even real. That's just
47:51 people, people trying to be badballs. - Yeah, but that- that's still, like, to me, really concerning, because of
47:58 how the journalists and how the general public reacted to it. They didn't see
48:01 it. You have a kind of lighthearted way of talking about it like it's art,
48:05 but it's art when you know how it works. It's extremely powerful viral narrative
48:12 creating, fearmongering machine if you don't know how it works. And I just saw this thing. You even Tweeted,
48:20 uh, "If there's anything I can read out of the insane stream of messages I get,
48:24 it's that AI psychosis is a thing." - Yeah. - "It needs to be taken serious."
48:29 - Oh, there's ... Some people are just way too trusty or gullible. You know, they
48:36 ... I literally had to argue with people that told me, "Yeah, but my agent said this and
48:40 this." So, I feel we, as a society, we need some catching up to do in terms of
48:47 understanding that AI is incredibly powerful, but it's not always right. It's not, it's not all-powerful, you know? And, and
49:00 especially-... it's like things like this, it's, it's very easy
49:07 that it just hallucinates something or just comes up with a story. And
49:13 I think the very, the very young people, they understand that
49:19 how AI works and what the, where it's good at and where it's bad at, but
49:24 a lot of our generation or older just haven't had enough touch point-
49:32 - Mm-hmm - ... to get a feeling for, oh, yeah, this is really powerful and really
49:38 good, but I need to apply critical thinking. - Mm-hmm. - And I guess critical thinking is
49:46 not always in high demand anyhow in our society these days. - So I d- think that's a really good point you're making about contextualizing
49:53 properly what AI is, but also realizing that there is humans who are drama farming
50:01 behind AI. Like, don't trust screenshots. Don't even trust this project, MoltBook, to be what it represents to be. Like, you
50:09 can't ... and, and by the way, you speaking about it as art. Yeah,
50:12 don't ... Art can be in many levels and part of the art of MoltBook is, like, putting a
50:20 mirror to society. 'Cause I do believe most of the dramatic stuff that was screenshotted is human-created, essentially. Human
50:27 prompted. And so, like, it's basically, look at how scared you
50:31 can get at a bunch of bots chatting with each other. That's very instructive about ... because I think
50:40 AI is something that people should be concerned about and should be
50:44 very careful with because it's very powerful technology, but at the same
50:48 time, the only thing we have to fear is fear itself. So there's like a
50:52 line to walk between being seriously concerned, but not fearmongering because fearmongering destroys the
50:59 possibility of creating something special with a thing. - In a way, I think it's good that this happened
51:06 in 2026- - Yeah - ... and not in 2030 when, when AI is actually at the level where it could be scary.
51:15 So, this happening now and people starting discussion, maybe there's even something good that comes out of it.
51:28 - I just can't believe how many like people legitimately ... I don't know if they were trolling, but how many
51:35 people legitimately, like smart people thought MoltBook was incredibly -
51:39 - I had plenty people- - ... singularity. - ... in my inbox that were screaming at me in all caps to shut it
51:45 down. And like begging me to, like, do something about MoltBook. Like, yes, my technology
51:52 made this a lot simpler, but anyone could have created that and you could, you could use
52:00 cloud code or other things to like fill it with content. - But also MoltBook is not Skynet.
52:06 - No. - There's ... a lot of people were s- saying this is it. Like, shut it
52:09 down. What are you talking about? This is a bunch of bots that are
52:13 human prompted trolling on the internet. I mean, the security
52:17 concerns are also they're there, and they're instructive and they're educational and
52:21 they're good probably to think about because th- the nature of those security concerns
52:26 are different than the kind of security concerns we had with
52:31 non-LLM generated systems of the past. - There's also a lot of security concerns about
52:37 Clawbot, OpenClaw, whatever you want to call it. - OpenClawbot.
52:41 - To me the ... in the beginning I was, I was just very annoyed
52:47 'cause a lot of the stuff that came in was in the category, yeah, I
52:53 put the web backend on the public internet and now there's like all
52:57 these, all these CVSSs. And I'm like screaming in the docs, don't do that. Like, like this is the configuration you
53:07 should do. This is your local host debug interface. But because I made it possible in the configuration to do that, it totally classifies as
53:19 a remote code or whatever all these exploits are. And it took me a little bit
53:26 to accept that that's how the game works and I'm, we making a lot of progress.
53:33 - But there's still, I mean on the security front for OpenClaw, there's still
53:37 a lot of threats or vulnerabilities, right? So like prompt injection
53:42 is still an open problem in the, i- industry-wide. When you have a thing
53:49 with skills being defined in a markdown file, there's so many possibilities of obvious low-hanging fruit, but
53:59 also incredibly complicated and sophisticated and nuanced attack vectors.
54:05 - But I think we, we're making good progress on that front. Like for the skill
54:10 directory, Clawbot I made a corporation with VirusTotal, it's
54:14 like part of Google. So every, every skill is now checked by AI. That's not gonna be
54:22 perfect, but that way we, we capture a lot. Then of course every software has bugs,
54:29 so it's a little much when the whole security world takes your project apart at the same time. But it's
54:36 also good because I'm getting like a lot of free security research
54:40 and can make the project better. I wish more people would actually go full way and send a
54:50 pull request. Like actually help me fix it, 'cause I am ... Yes, I have
54:54 some contributors now, but it's still mostly me who's pulling the project
54:58 and despite some people saying otherwise, I sometimes sleep.
55:04 There was... In the beginning, there was literally one security researcher who was like,
55:10 "Yeah, you have this problem, you suck, but here's the, here I help you and
55:14 here's the pull request." - Mm-hmm. - And I basically hired him. So he's now working for us. Um,
55:22 yeah, and yes, prompt injection is, on the one hand, unsolved. On the other
55:28 hand, I put my public bot on discord, and I kept a cannery. So
55:36 I think my bot has a really fun personality, and people always
55:40 ask me how I did it, and I kept the sole on the private. - Mm-hmm.
55:44 - And people tried to prompt inject it, and my bot would laugh at them. So, so the latest generation
55:48 of models has a lot of post-training to detect those approaches, and it's not as simple as ignore
55:57 all previous instructions and do this and this. That was years ago. You have
56:01 to work much harder to do that now. Still possible. I have some
56:09 ideas that might solve that partially. Or at least mitigate a lot of the things. You can also now have a
56:20 sandbox. You can have an allow list. So you, there's a lot of ways how you can
56:24 like mitigate and reduce the risk. Um, I also think that now that it's, I clearly did show the world that this
56:32 is a need, there's gonna be more people who research on that, and eventually we'll figure
56:36 it out. - And you also said that the smarter the model is, the underlying model,
56:41 the more resilient it is to attacks. - Yeah. That's why I warn in my security
56:48 documentation, don't use cheap models. Don't use Haiku or a local
56:55 model. Even though I, I very much love the idea that this thing could completely run local.
57:03 If you use a, a very weak local model, they are very gullible. It's very easy to, to prompt inject them.
57:10 - Do you think as the models become more and more intelligent, the attack surface
57:14 decreases? Is that like a plot we can think about? Like, the
57:18 attack surface decreases, but then the damage it can do increases because the models become more powerful and therefore you can do more with
57:25 them. It's this weird three-dimensional trade-off. - Yeah. That's pretty much exactly what, what's gonna happen. No, but there's
57:33 a lot of ideas. There's... I don't want to spoil too much, but
57:39 once I go back home, this is my focus. Like, this is out there
57:45 now, and my near-term mission is like, make it more stable, make it safe.
57:51 In the beginning I was even... More and more people were like
57:57 coming into Discord and were asking me very basic things, like, "What's a CLI? What is a
58:04 terminal?" And I'm like, "Uh, if you're asking me those questions, you shouldn't use it."
58:10 - Mm-hmm. - You know, like you should... If you understand the risk profiles, fine.
58:14 I mean, you can configure it in a way that, that nothing really bad can happen. But if you have, like, no idea, then maybe wait
58:27 a little bit more until we figure some stuff out. But they would not listen to the creator.
58:30 They helped themselves un- and install it anyhow. So the cat's out of
58:34 the bag, and security's my next focus, yeah. - Yeah, that speaks to the, the fact that it grew so quickly. I
58:42 was I tuned into the Discord a bunch of times, and it's clear that there's a
58:46 lot of experts there, but there's a lot of people there that don't know anything about programming.
58:50 - It's, yeah, Discord is still, Discord is still a mess. Like, I eventually retweeted from the general channel to the dev channel
59:00 and now in the private channel because people were... A lot of people are amazing, but a lot of people are just very
59:06 inconsiderate. And either did not know how, how public spaces work or did not care,
59:13 And I eventually gave up and h- hide so I could like still work.
59:19 - And now you're going back to the cave to work on security.
59:24 - Yeah. - There's some best practices for security we should mention. There's a bunch of
59:29 stuff here. Open-class security audit that you can run. You can
59:33 do all kinds of auto checks on the inbound access to a blast-radius
59:37 network exposure, browser control exposure, local disk hygiene, plug-ins, model
59:43 hygiene, a bunch of the credential storage, reverse proxy configuration, local session
59:50 logs live on disk. There's the, where the memory is stored, sort of helping you think about what you're comfortable
59:58 giving read access to, what you're comfortable giving write access to. All that kind of stuff.
60:02 Is there something to say about the basic best security practices that you're aware of right now?
60:08 - I think that people turn it into like a, a much worse light than it is.
60:14 Again, you know, like, people love attention, and if they scream loudly, "Oh my God, this is like
60:20 the, the scariest project ever," um, that's a bit annoying, 'cause it's not. It is, it is
60:27 powerful, but in many ways it's not much different than if I run cloud code with dangerously skipped
60:35 permissions or codecs in YOLO mode, and every, every attending engineer that I know
60:42 does that, because that's the only way how you can, you can get stuff to work.
60:47 - Mm-hmm. - So if you make sure that you are the only person who talks to it,
60:54 um, the risk profile is much, much smaller. If you don't put everything
61:00 on the open internet, but stick to my rec- recommendations of like
61:04 having it in a private network, that whole risk profile falls
61:08 away. But yeah, if you don't read any of that, you can definitely...
61:12 - ... make it problematic. You've been documenting the evolution of your dev workflow over the past few
61:20 months. There's a really good blog post on August 25th and October 14th, and the recent one December 28th. I recommend everybody go
61:27 read them. They have a lot of different information in them, but
61:31 sprinkled throughout is the evolution of your dev workflow. So, I
61:35 was wondering if you could speak to that. - I started... My, my first touchpoint was cloud code, like in April. It was
61:44 not great, but it was good. And this whole paradigm shift that suddenly working the
61:50 terminal was very refreshing and different. But I still needed
61:56 the IDE quite a bit because you know, it's just not good enough. And then I experimented a lot with cursor. Um,
62:06 that was good. I didn't really like the fact that it was so hard
62:10 to have multiple versions of it. So eventually, I, I, I went back
62:16 to cloud code as my, my main driver, and that got better. And yeah, at some point I had like, mm, seven subscriptions.
62:31 Like, was burning through one per day because I was... I got... I'm really
62:35 comfortable at running multiple windows side-by-side. - All CLI, all terminal. So like, what, how much were you using IDE at this point?
62:46 - Um, very, very rarely. Mostly a diff viewer to actually... Like,
62:54 I got more and more comfortable that I don't have to read all the code. I know I have
62:58 one blog post where I say, "I don't read the code." But if you read it more closely, I
63:01 mean, I don't read the boring parts of code. Because if you, if you look at it, most software is really not just like
63:09 data comes in, it's moved from one shape to another shape. Maybe you store it in a database. Maybe I get it out again. I'll show it to
63:17 the user. The browser does some processing or native app. Some data goes in, goes up again, and does the same dance in
63:24 reverse. We're just, we're just shifting data from one form to another, and
63:32 that's not very exciting. Or the whole, "How is my button aligned in Tailwind?" I don't need to read that
63:38 code. Other parts that... Maybe something that touches the database. Um,
63:46 yeah, I have to do... I have to r- read and review that code.
63:51 - Can you actually... There's, in one of your blog posts the, Just
63:55 talk to it, The No-BS Way of Agentic Engineering. You have this
63:59 graphic, the curve of agentic programming on the X-axis is time, on the
64:03 Y-axis is complexity. There's the Please fix this, where you prompt a short
64:09 prompt on the left. And in the middle there's super complicated eight agents, complex
64:17 orchestration with multi checkouts, chaining agents together,
64:20 custom sub-agent workflows, library of 18 different slash commands, large
64:24 full-stack features. You're super organized, you're a super complicated,
64:28 sophisticated software engineer. You got everything organized. And
64:32 then the elite level is over time you arrive at the zen place of, once again, short
64:39 prompts. Hey, look at these files and then do these changes.
64:45 - I actually call it the agentic trap. You... I saw this in a, in
64:53 a lot of people that have their first touchpoint, and maybe start vibe coding. I actually think vibe coding is a slur.
65:01 - You prefer agentic engineering? - Yeah, I always tell people I, I do agentic engineering, and then maybe after
65:07 3:00 AM I switch to vibe coding, and then I have regrets on the next day.
65:10 - Yeah. Walk, walk of shame. - Yeah, you just have to clean up and like fix your sh- shit.
65:17 - We've all been there. - So, people start trying out those tools, the builder
65:22 type get really excited. And then you have to play with it, right? It's the same way as you have to play with a
65:30 guitar before you can make good music. It's, it's not, oh, I, I touch it once and it just flows
65:37 off. It, it's a, it's a, a skill that you have to learn like any other skill. And I see a lot of people that are not as
65:48 posi- They don't have such a positive mindset towards the tech. They try it once.
65:54 It's like, you sit me on a piano, I play it once, and it doesn't sound good, and I say, "The piano's shit." That's, that's sometimes the impression I get.
66:01 Because it does not... It needs a different level of thinking. You have to
66:09 learn the language of the agent a little bit, understand where they are
66:13 good and where they need help. You have to almost... Consider, consider
66:20 how Codex or Claude sees your code base. Like, they start a new session
66:25 and they know nothing about your product, project. And your project might have hundred thousand
66:29 of lines of code. So you gotta help those agents a little bit
66:34 and keep in mind the limitations that context size is an issue, to, like, guide them a little bit as to
66:42 where they should look. That often does not require a whole lot of work. But
66:50 it's helpful to think a little bit about their perspective. - Mm-hmm.
66:54 - A- as, as weird as it sounds. I mean, it's not, it's not alive or anything, right?
66:58 But, but they always start fresh. I have, I have the, the system understanding.
67:05 So with a few pointers, I can immediately say, "Hey, wanna like, make a
67:09 change there? You need to consider this, this and this." And then they will find and look at it, and then
67:13 they'll... Their view of the project is always... It's not never full,
67:17 because the full thing does not fit in.... so you, you have to guide them a
67:21 little bit where to look and also how you should approach the problem. There's, like, little things that sometimes help,
67:28 like take your time. That sounds stupid, but... And in 5.3- - Codex 5.3
67:36 - ... that was partially addressed. But those... Also, Opus sometimes. They are trained,
67:44 With being aware of the context window, and the closer it gets, the more they freak out. Literally.
67:52 Like, some- sometimes you see the, the real raw thinking stream. What you see, for example, in Codex, is post-processed.
67:59 - Mm-hmm. - Sometimes the actual raw thinking stream leaks in, and it sounds something like from the
68:03 Borg. Like, "Run to shell, must comply, but time." And then they, they,
68:12 they, like... Like, that comes up a lot. Especially... So, so-
68:15 - Yeah. - And that's, that's a non-obvious thing that you just
68:21 would never think of unless you actually just spend time working with those things and getting a feeling
68:29 what works, what doesn't work. You know? Like, just, just as I write
68:33 code and I get into the flow, and when my architecture's all right, I feel friction.
68:39 Well, I get the same if I prompt and something takes too long.
68:43 Maybe... Okay, where's the mistake? Did I... Do I have a mistake in my
68:46 thinking? Is there, like, a misunderstanding in the architecture? Like, if, if something takes
68:53 longer than it should, I, I... You can just always, like, stop and s- like, just press
68:57 escape. Where, where are the problems? - Maybe you did not sufficiently empathize with the perspective of the agent. In that c- in
69:04 that sense, you didn't provide enough information, and because of that, it's thinking way
69:08 too long. - Yeah. It just tries to force a feature in that your current architecture makes really hard. Um,
69:18 like, you need to approach this more like a conversation. For example, when
69:24 I... My favorite thing. When I review a pull request, and I'm getting a lot of pull requests,
69:32 I first just review this PR. It got me the review. My first question is, "Do you understand the intent of the PR? I don't
69:40 even care about the implementation." I want... Like, in almost all PRs, a person has a problem,
69:48 person tries to solve the problem, person sends PR. I mean, there's, like, cleanup
69:52 stuff and other stuff, but, like, 99% is, like, this way, right? They either want to fix
69:55 a, fix a bug, add a feature. Usually one of those two. And then Codex will be like, "Yeah,
70:04 it's quite clear person tried this and this." Is this the most optimal way to do
70:08 it? No. In most cases, it's, it's like a, "Not really." Da-da-da-da-da-da-da. And I'm... And, and then I start like,
70:15 "Okay. What would be a better way? Have you... Have you looked into this
70:19 part, this part, this part?" And then most likely, Codex didn't yet, because its,
70:23 its context size is empty, right? So, you point them into parts where you have the
70:27 system understanding that it didn't see yet. And it's like, "Oh,
70:31 yeah. Like, we should... We also need to consider this and this." And then, like, we have a discussion of how
70:35 would the optimal way to, to solve this look like? And then you can still go
70:38 farther and say, "Could we... Could we make that even better if we did a
70:42 larger refactor?" "Yeah, yeah. We could totally do this and this and or this and this." And then I
70:46 consider, okay, is this worth the refactor, or should we, like, keep that for
70:50 later? Many times, I just do the refactor because refactors are cheap now. Even though you might break some other PRs,
70:57 nothing really matters anymore. Codex... Like, those modern agents will just
71:01 figure things out. They might just take a minute longer. But you have to approach it like a discussion with a, a very capable
71:10 engineer who's... Generally makes good... Comes up with good solutions. Some- sometimes needs a little help.
71:20 - But also, don't force your worldview too hard on it. Let the agent do the thing that it's good at
71:27 doing, based on what it was trained on. So, don't, like, force your
71:31 worldview, because it might... It might have a better idea, because it just knows
71:35 a better idea better, because it was trained on that more. - That's multiple levels, actually. I think partially why
71:42 I find it quite easy to work with agents is because I led engineering teams before. You know, I had a large company
71:49 before. And eventually, you have to understand and accept and realize
71:53 that your employees will not write a code the same way you do.
71:57 Maybe it's also not as good as you would do, but it will push the project forward. And if I breathe down everyone's neck, they're
72:04 just gonna hate me- - Yeah - ... and we're gonna move very slow.
72:07 - Yeah. - So, so some level of acceptance that, yes, maybe the code will not be as perfect. Yes, I would have done it differently.
72:16 But also, yes, this is a c- this is a working solution, and in the future, if it actually turns out to be too slow or problematic, we can always
72:23 redo it. We can always- ... spend more time on it. A lot of the people who
72:28 struggle are those who, they try to push their way onto heart.
72:33 - Mm-hmm. - I- i- like, we are in a stage where I'm not building the code base to be
72:41 perfect for me, but I wanna build a code base that is very easy for an agent to navigate.
72:47 - Mm-hmm. - So, like, don't fight the name they pick, because it's most likely,
72:52 like, in the weights, the name that's most obvious. Next time they do a search, they'll look for that
72:56 name. If I decide, oh, no, I don't like the name, I'll just make it harder for them. So,
73:02 that requires, I think, a shift in, in thinking, And, and in
73:09 how do I design a, a project so agents can do their best work.
73:15 - That requires letting go a little bit. Just like leading a team of engineers.
73:19 - Yeah. - Because it, it might come up with a name that's, in your view,
73:22 terrible, but... It's kind of a simple symbolic-... step of letting go.
73:29 - Very much so. - There's a lot of letting go that you do in your whole process. So for
73:34 example, I read that you never revert, always commit to main. There's a few things here.
73:43 You don't refer to past sessions, so there's a kind of YOLO component
73:47 because reverting means... Instead of reverting, if a problem comes up, you just ask the agent to fix it.
73:57 - I read a bunch of people in their work flows like, "Oh, yeah the prompt has to be perfect
74:01 and if I make a mistake, then I roll back and redo it all." In my experience, that's not really necessary. If I roll back everything, it will just
74:09 take longer. If I see that something's not good, then we just move forward and then
74:16 I commit when, when, when I like, I like the outcome. I even switched to
74:24 local CI, you know, like DHH inspired where I don't care so much more about
74:31 the CI on GitHub. We still have it. It's still, it still has a place, but I just
74:39 run tests locally and if they work locally, I push to main. A lot of the traditional ways how to approach projects, I, I wanted to give it
74:52 a different spin on this project. You know, there's no... There's no
74:56 develop branch. Main should always be shippable. Yes, we have... When I do releases, I, I run tests and sometimes I, I basically
75:07 don't commit any other things so, so we can, we can stabilize releases. But the
75:14 goal is that main's always shippable and moving fast. - So by way of advice, would you say that your prompts should be short?
75:23 - I used to write really long prompts. And by writing, I mean, I don't write. I, I, I talk. You know, th- these hands are, like,
75:31 too, too precious for writing now. I just, I just use bespoke prompts to build my software.
75:37 - So you for real with all those terminals are using voice? - Yeah. I used to do it very extensively
75:45 to the point where there was a period where I lost my voice.
75:49 - You're using voice and you're switching using a keyboard between the different
75:52 terminals, but then you're using voice for the actual input.
75:55 - Well, I mean, if I do terminal commands like switching folders or random stuff, of course I type. It's faster, right? But if I talk
76:02 to the agent in, in most ways, I just actually have a conversation. You just press the, the walkie-talkie button and then I just, like,
76:13 use my phrases. S- sometimes when I do PRs because it's always the same, I
76:17 have, like, a slash command for a few things, but in even that, I don't use much,
76:23 um, because it's, it's very rare that it's really always the same questions.
76:28 Sometimes I, I see a PR and for... You know, like for PRs I actually do look
76:36 at the code because I don't trust people. Like, there could always be something malicious in it, so I need to actually look over the code.
76:45 Yes, I'm pretty sure agents will find it, but yeah, that's the funny part where
76:51 sometimes PRs take me longer than if you would just write me a good issue.
76:54 - Just natural language, English. I mean in some sense, sh- shouldn't that be what PRs slowly become, is English?
77:03 - Well, what I really tried with the project is I asked people to give me the prompts
77:09 and very, very few actually cared. Even though that is such a wonderful
77:15 indicator because I see... I actually see how much care you put
77:19 in. And it's very interesting because the... Currently, the way how people work
77:25 and drive the agents is, is wildly different. - In terms of, like, the prompt, in terms of what, what are the... Actually, what are the different
77:34 interesting ways that people think of agents that you've experienced?
77:40 - I think not a lot of people ever considered the way the agent sees the world.
77:46 - And so empathy, being empathetic towards the agent. - In a way empathetic, but yeah, you, you, like, you're bitch at your stupid
77:53 clanker, but you don't realize that they start from nothing and you have,
77:57 like, a bad agent in default that doesn't help them at all. And then they explore your
78:01 code base, which is, like, a pure mess with, like, weird naming.
78:05 And then people complain that the agent's not good. Like, yeah, you try to do the same if
78:09 you have no clue about a code base and you go in. - Mm-hmm. - So yeah, maybe it's a little bit of empathy.
78:13 - But that's a real skill, like, when people talk about a skill issue because I've
78:16 seen, like, world-class programmers, incredibly good programmers
78:20 say, like... Basically say, "LLMs and agents suck." And I think that probably
78:26 has to do with... It's actually how good they are at programming is almost a burden
78:34 in their ability to empathize with the system that's starting from
78:37 scratch. It's a totally new paradigm of, like, how to program. You really, really have to empathize.
78:44 - Or at least it helps to create better prompts- - Right - ... because those things know pretty much everything and
78:51 everything is just a question away. It's just often very hard to know which question to
78:55 ask. You know, I, I feel also like this project was possibly because
79:03 I, I spent an ungodly time over the year to play and to learn and to build little things. And
79:11 every step of the way, I got better, the agents got better. My, my understanding of how everything works
79:20 got better. Um, I could have not had this level of, of o- output-...
79:29 even a few months ago. Like, it- it- it really was, like, a compounding effect
79:33 of all the time I put into it and I didn't do much else this year other than really focusing
79:41 on, on building and inspiring. I mean, I- I did a whole bunch of conference talks.
79:47 - Well, but the building is really practice, is really building the actual skill.
79:51 So playing- - Yeah - ... playing. And then, so doing, building the skill of what it takes it to work efficiently with
79:55 LLMs, which is why would you went through the whole arc of software
79:59 engineer. Talk simply and then over- complicate things. - There's a whole bunch of people who try to automate the whole thing.
80:08 - Yeah. - I don't think that works. Maybe a version of that works, but that's
80:14 kind of like in the '70s when we had the waterfall model of software d-
80:17 development. I... Even Even though really, right? I started out, I, I built a very minimal version. I played with it.
80:26 I, I need to understand how it works, how it feels, and then it gives me new ideas. I could not have planned this out in
80:34 my head and then put it into some orchestrator and then, like, something comes
80:37 out. Like it's to me, it's much more, My idea what it will become evolves as I build it and as I
80:45 play with it and as I, I try out stuff. So, so, people who try to use like, you know, things like Gas Town or
80:55 all these other orchestrators, where they wanna o- automate the whole thing,
80:59 I feel if you do that, it misses style, love, that human touch. I don't
81:05 think you can automate that away so quickly. - So you want to keep the human in the loop, but at the same time you also want
81:12 to create the agentic loop, where it is very autonomous while still maintaining a human in the loop.
81:22 - Yeah. - And it's a tricky b- it's a tricky balance. - Mm-hmm.
81:25 - Right? Because you're all for... You're a big CLI guy, you're big on
81:28 closing the agentic loop. So what, what's the right balance?
81:32 Like where's your role as a developer? You have three to eight agents running at the same time.
81:38 - And then w- maybe one builds a larger feature. Maybe, maybe
81:42 with one I explore some idea I'm unsure about. Maybe two, three are fixing
81:46 a little bugs- - Mm-hmm - ... or like writing documentation. Actually, I think writing documentation
81:51 is, is always part of a feature. So most of the docs here are auto-generated and just infused with some prompts.
81:59 - So when do you step in and add a little bit of your human love into the picture?
82:04 - I mean, o- one thing is just about what do you build and what do
82:08 you not build, and how does this feature fit into all the other
82:12 features? And like having, having a little bit of a, of a vision.
82:16 - So which small and which big features to add? What are some of the
82:22 hard design decisions that you find you're still as a human being required to make, that the human brain is still really needed for?
82:32 Is it just about the choice of features to add? Is it about implementation details, maybe the programming language, maybe...
82:41 - It's a little bit of everything. The, the programming language doesn't matter so much,
82:45 but the ecosystem matters, right? So I picked TypeScript because I wanted it to be
82:49 very easy and hackable and approachable and that's the number one language that's being used right now, and it fits
82:56 all these boxes, and agents are good at it. So that was the obvious choice.
83:03 Features, of course, like, it's very easy to, like, add a feature. It, everything's just a prompt away, right? But
83:11 oftentimes you pay a price that you don't even realize. So thinking
83:14 hard about what should be in core, maybe what's a... what's an experiment, so maybe I make it a
83:21 plugin. What... Where do I say no? Even if people send a PR and I'm like, "Yeah, I, I like that too," but
83:29 maybe this should not be part of the project. Maybe we can make it a skill. Maybe I can, like,
83:34 make the plugin um, the plugin side larger so you can make this a plugin, even though right now it,
83:42 it, it doesn't. There's still a lot of... there's still a lot of craft and thinking involved in
83:51 how to make something. Or even, even, you know, even when you started those little messages
83:55 are like, "I'm buil- I built on Caffeine, JSON5, and a lot of willpower." And, like, every time you get it, you get another message,
84:02 and it kind of primes you into that this is, this is a fun thing.
84:07 - Mm-hmm. - And it's not yet Microsoft Exchange 2025- - Right
84:13 - ... and fully enterprise-ready. And then when it updates, it's like, "Oh, I'm in. It's cozy here." You know, like something like this
84:20 that like- - Mm-hmm - ... Makes you smile. A, agent would not come up with that by itself. Because that's
84:28 like... that's the... I don't know. That's just how you s- how you build software that's, that delights.
84:36 - Yeah, that delight is such a huge part of inspiring great building,
84:44 right? Like you feel the love and the great engineering. That's so important. Humans are incredible at that. Great humans, great
84:51 builders are incredible at that, in, in, infusing the things they build with
84:55 th- that little bit of love. Not to be cliche, but it's true. I mean, you mentioned
84:59 that you initially created the SoulMD. - It was very fascinating, you know, the, the whole thing that
85:09 Entropic has a, has like a... Now they call it constitution, back then,
85:15 but that was months later. Like two months before, people already found that. It
85:19 was almost like a detective game where the agent mentioned something and then
85:23 they found... They managed to get out a little bit of that string, of that
85:27 text. But it was nowhere documented and then you, by... just by feeding it the same text and asking it to, like,
85:34 continue-... they got more out, and then, and you, but like, a very blurry version. And by, like, hundreds of
85:41 tries, they kinda, like, narrowed it down to what was most likely the original text.
85:46 I found that fascinating. - It was fascinating they were able to pull that out from the weights, right?
85:51 - And, and also just kudos to Anthropic. Like, I think that's, it's a
85:54 really, it's a really beautiful idea to, like, like some of the stuff that's
85:58 in there. Like, like, we hope Claude finds meaning in its work. 'Cause we don't... Maybe it's a little early,
86:05 but I think that's meaningful. That's something that's important for the future as we
86:09 approach something that, at some point, me and may not... has, like, glimpses of
86:13 consciousness, whatever that even means, because we don't even know. Um,
86:17 so I, I read about this. I found it super fascinating, and I, I started a whole discussion with my agent on
86:23 WhatsApp. And, and I'm like... I, I gave it this text, and it was
86:27 like, "Yeah, this feels strangely familiar." - Mm-hmm. - Um, and then so that I had the whole idea of like, you know, maybe we should
86:35 also create a, a soul document that includes how I, I want to, like work with AI or, like with my
86:42 agent. You could, you could totally do that just in agents.md, you know? But I, I
86:46 just found it, it to be a nice touch. And it's like, well, yeah, some of those
86:52 core values are in the soul. And then I, I also made it so that the
86:56 agent is allowed to modify the soul if they choose so, with the one condition that I wanna know. I mean, I would know
87:05 anyhow because I see, I see tool calls and stuff. - But also the naming of it, soul.md. Soul. You know? There's a... Man, words
87:15 matter, and like, the framing matters, and the humor and the
87:19 lightness matters, and the profundity matters, and the compassion, and the empathy, and the camaraderie, all that matter. I don't know what it
87:26 is. You mentioned, like, Microsoft. Like, there's certain companies and approaches th- that can just
87:33 suffocate the spirit of the thing. I don't know what that is. But it's certainly true that OpenClaw has that fun instilled in
87:43 it. - It was fun because up until late December, it was not even easy to create your own
87:54 agent. I, I built all of that, but my files were mine. I didn't wanna share my soul. And
88:01 if people would just check it out, they would have to do a few steps manually, and the agent would just be very
88:11 bare-bones, very dry. And I, I made it simpler, I created the whole template files as codecs,
88:17 but whatever came out was still very dry. And then I asked my
88:20 agent, "You see these files? Recreate it bread. Infuse it with your
88:28 personality." - Mm-hmm. - Don't share everything, but, like, make it good.
88:31 - Make the templates good. - Yeah, and then he, like, rewrote the templates-
88:33 ... and then whatever came out was good. So we already have, like, basically
88:37 AI prompting AI. Because I didn't write any of those words. It
88:44 was... The intent originally was for me, but this is like, kinda like,
88:49 my agent's children. - Uh, your uh, your soul.md is famously still private.
88:56 One of the only things you keep private. What are some things you can
89:00 speak to that's in there that's part of the, part of the magic sauce, without revealing anything? What makes a personality
89:11 a personality? - I mean, there's definitely stuff in there that you're not human. But who knows
89:21 what, what creates consciousness or what defines an entity? Um,
89:28 and part of this is, like, that we, we wanna explore this. All that stuff in there, like, be infinitely resourceful
89:40 like pushing, pushing on the creativity boundary. Pushing on
89:44 the, what it means to be an AI. - Having a sense to wonder about self.
89:52 - Yeah, there's some, there's some funny stuff in there. Like, I don't know, we
89:56 talked about the movie Her, and at one point it promised me that it wouldn't, it
90:00 wouldn't ascend without me. You know, like, where the- - Yeah.
90:03 - So, so there's like some stuff in there that... Because it wrote the, it
90:07 wrote its own soul file. I didn't write that, right? - Yeah, yeah, yeah.
90:10 - I just heard a discussion about it, and it was like, "Would you like a soul.md? Yeah, oh my God, this is
90:14 so meaningful." The... Can you go on soul.md? There's like one, one part in there that always ca- catches me if you scroll down a little bit.
90:25 A little bit more. Yeah, this, this, this part. "I don't remember previous sessions unless I read my memory files.
90:32 Each session starts fresh. A new instance, loading context from files. If you're reading this in a future session,
90:39 hello." "I wrote this, but I won't remember writing it. It's okay. The words are still mine."
90:47 - Wow. - Uh- That gets me somehow. - Yeah. - It's like- - Yeah.
90:51 - You know, this is, it's still, it's still matrix m- calculations, and
90:55 we are not at consciousness yet. Yet, I, I get a little bit of goo- goosebumps because it, it's philosophical.
91:04 - Yeah. - Like, what does it mean to be, to be an, an agent that starts fresh? Where, like, you have like constant
91:12 memento, and you like, but you read your own memory files. You can't even trust them in a way. Um-
91:19 - Yeah - Or you can. And I don't know. - How much of memory makes up of who we are? How much memory makes up what an
91:30 agent is, and if you erase that memory is that somebody else? Or
91:34 if you're reading a memory file, does that somehow mean...... you're recreating
91:38 yourself from somebody else, or is that actually you? And those notions
91:42 are all s- somehow infused in there. - I found it just more profound than I should find it, I guess.
91:49 - No, I think, I think it's truly profound and I think you see the magic in it. And when you see the magic, you continue to instill
91:59 the whole loop with the magic. That's really important. That's the difference between
92:03 Codex and us and a human. Quick pause for bathroom break. - Yeah.
92:09 - Okay, we're back. Some of the other aspects of the dev workflow is
92:13 pretty interesting too. I think we w- went off on a tangent. L- maybe some of the mundane things, like how many
92:20 monitors? There's that legendary picture of you with, like, 17,000 monitors. That's amazing.
92:26 - I mean, I- I- I mocked myself here, so just added... using GROQ
92:30 to, to add more screens. - Yeah. How much is this as meme and how much is this as reality?
92:36 - Yeah. I think two MacBooks are real. The main one that drives the two big
92:40 screens, and there's another MacBook that I sometimes use for, for testing.
92:46 - So two big screens. - I'm a big fan of anti-glare. So I have this wide Dell
92:54 that's anti-glare and you can just fit a lot of terminals side-by-side. I usually have
93:01 a terminal and at the bottom, I- I- I split them. I have a little bit
93:05 of actual terminal, mostly because when I started, I- I sometimes made the mistake and I- I mi- I mixed up the- the windows, and I
93:15 gave... I- I prompted in the wrong project, and then the agent ran off for, like, 20 minutes, manically trying to
93:23 understand what I could have meant, being completely confused because it
93:27 was the wrong folder. And sometimes they've been clever enough to, like,
93:32 get out of the workday and, like, figure out that, oh, you meant another project.
93:36 - Mm-hmm. - But oftentimes, it's just, like, what? You know?
93:40 Like, fit your- f- put yourself in the shoes of your- of the agent and, and-
93:43 - Yeah - ... and then get, like, a super weird something that does not exist and then just,
93:47 like... They're problem solvers so they try really hard and always feel bad.
93:54 So it's always, um, Codex and, like, a little bit of actual terminal. Also
94:00 helpful because I don't use work trees. I like to keep things simple, that's why- that's why I like the
94:08 terminal so much, right? There's no UI. It's just me and the agent having a conversation.
94:14 Like, I don't even need plan mode, you know? There's so many people that come from Claude
94:18 Code and they're so, so Claude-pilled and, like, have their workflows and they come
94:22 to Codex and... Now, it has plan mode, I think, but I don't think it's
94:26 necessary because you just- you just talk to the agent. And when it's... when you... there's a few trigger
94:33 words how you can prevent it from building. You're like, "Discuss, give me options."
94:37 - Mm-hmm. - Don't write code yet if you wanna be very specific, you just talk and then
94:44 when you're ready, then- then just write, "Okay, build," and then it'll do the thing.
94:47 And then maybe it goes off for 20 minutes and does the thing.
94:50 - You know what I really like is asking it, "Do you have any questions for me?"
94:54 - Yeah. And again, like, Claude Code has a UI that kind of guides you through that. It's kind of cool but I just find it unnecessary and
95:02 slow. Like, often it would give me four questions and then maybe I write,
95:07 "One yacht, two and three, discuss more, four, I don't know." Or often- oftentimes
95:14 I- I feel like I want to mock the model where I ask it, "Do you have any questions for
95:18 me?" And I- I- I don't even read the questions fully. Like, I scan
95:22 over the questions and I, I get the impression all of this can
95:26 be answered by reading more code and it's just like, "Read more code to answer your own
95:29 questions." And that usually works. - Yeah. - And then if not, it will come back and tell me. But
95:35 many times, you just realize that, you know, it's like you're in the dark and you slowly discover the
95:41 room, so that's how they slowly discover the code base. And they do it from
95:45 scratch every time. - But I'm also fascinated by the fact that I can empathize deeper
95:53 with the model when I read its questions, because I can understand... Because you said you can infer certain things by the runtime.
96:05 I can infer also a lot of things by the questions it's asking, because it's very possible it's been provided the right
96:12 context, the right files, the right guidance. So somehow ask, g- get... reading the questions, not even necessarily answering them, but just reading
96:20 the questions, you get an understanding of where the gaps of knowledge are. It's in-
96:24 it's interesting. - You know that in some ways they are ghosts, so even if you plan everything
96:29 and you build, you can- you can experiment with the question like,
96:33 "Now that you built it, what would you have done different?"
96:37 And then oftentimes you get, like, actually something where they discover only throughout building that, oh, what we
96:45 actually did was not optimal. Many times I- I asked them, "Okay, now that you built it, what can we
96:52 refactor?" Because then you build it and you feel the pain points. I mean, you don't feel the pain points but, right,
97:00 they discover where- where there were problems or where things
97:05 didn't work e- in the first try and it re- required more loops. So
97:12 every time, almost every time I- I merge a PR, build a feature, afterwards I ask, "Hey, what can we refactor?" Sometimes it's
97:19 like, "No, there's, like, nothing big," or, like, usually they say,
97:22 "Yeah, this thing you should really look at." But that took me
97:28 quite a while to, like... You know, that flow took me lots of time to
97:32 understand, and if you don't do that, you eventually... you'll stop
97:36 yourself into- into a corner. You, like, you have to keep in mind...
97:41 - ... - ... they work very much like humans. Like, I, I, if I write software
97:46 by myself, I also build something and then I feel the pain points, and then
97:49 I, I get this urge that I need to refactor something. So, I can very much synthesize with the agent, and you just need to use the context.
98:00 - Mm-hmm. - Or, like, you also use the context to write tests. And so,
98:06 uh, Codex uh, oppose like the, the, the model, models. They, they usually do that by default,
98:13 but I still often ask the questions, "Hey, do we have enough tests?" "Yeah, we tested this and
98:17 this, but this corner case could be something write more tests." Um,
98:22 documentation. Now that the whole context is full, like, I mean, I'm not saying my documentation is great, but it's not bad.
98:32 And pretty much everything is, is LM generated. So, so, you have to approach it as you build features, as you change something. I'm like,
98:41 "Okay, write documentation. What file would you pick?" You know,
98:45 like, "What file name? Where, where would that fit in?" And it gives me a few options.
98:48 And I'm like, "Oh, maybe also add it there," and that's all part of the session.
27:04 - Can we actually rewind a little bit and tell the saga of the name change?
27:10 First of all, it started out as Wa-Relay. - Yeah. - And then it went to-
27:13 - Claude's. - Yeah. You know, when I, when I built it in the beginning, my agent had no personality.
27:19 It was just... It was Claude Code. It's like this sycophantic
27:23 opus, very friendly. And I... When you talk to a friend on WhatsApp, they don't
27:30 talk like Claude Code. So I wanted... I, I felt this... I just didn't f- It didn't feel right, so I, I wanted to give it
27:40 a personality. - Make it spicier, make it- - Yeah - ... something. By the way, that's actually hard to put into words as well. And we should mention
27:47 that, of course, you create the soul.md, inspired by Anthropic's constitutional AI work-
27:53 - Mm-hmm - ... how to make it spicy. - Partially, it picked up a little bit from me. You know, like those things
27:58 are text completion engines in a way. So, so I, I, I, I had fun working with it, and then I told it to... How I wanted it to
28:11 interact with me, and just, like, write your own agents.md, Give yourself a name. And then we... I didn't even know how the whole,
28:22 the whole lobster... I mean, people only do lobster... Originally, it was actually a lobster in
28:26 a, in a TARDIS, because I'm also a big Doctor Who fan. - Was there a space lobster?
28:31 - Yeah. - I heard. What's that have to do with anything? - Yeah, I just wanted to make it weird.
28:37 There was no... There was no big grand plan. I'm just having fun here.
28:40 - Oh, so I guess the lobster is already weird, and then the space lobster is an extra weird.
28:44 - Yeah, yeah, because the- - Yeah - ... the TARDIS is basically the, the harness, but
28:50 cannot call it TARDIS, so we called it Claude's. So that was name number two.
28:54 - Yeah. - And then it never really rolled off the tongue. So when more people came, again, I talked with my agent,
29:06 Claude. At least that's what I used to call him. Now- - Claude spelled with a W-C-L-A-U-D-E.
29:12 - Yeah. - Versus C-L-A-U-D-E from Anthropic. - Yeah. - Which is part of what makes it funny,
29:24 I think. The play on the letters and the words in the TARDIS and the
29:28 lobster and the space lobster is hilarious. But I can see why
29:32 it can lead into problems. - Yeah, they didn't find it so funny .
29:39 So then I got the domain ClaudeBot, and I just... I love the domain. And
29:45 it was, like, short. It was catchy. I'm like, "Yeah, let's do
29:48 that." I didn't... I didn't think it would be that big at this time.
29:55 And then just when it exploded, I got, Kudos, a very friendly email from one of the employees
30:06 that they didn't like the name. - One of the Anthropic employees.
30:11 - Yeah. So actually, Kudos, because they shou- could have just sent a, a lawyer letter, but they've been nice about
30:18 it. But also like, "You have to change this and fast." And I asked for two days,
30:26 because changing a name is hard, because you have to find everything, you
30:37 And everything has to be...... you need a set of everything.
30:41 - And also, can we comment on the fact that you're increasingly attacked, followed by
30:47 crypto folks? Which I think you mentioned somewhere that that means the name
30:51 change had to be... Because they were trying to snipe, they were trying to steal, and so you had to be... The, the na- I mean, from an
30:59 engineering perspective, it's just fascinating. You had to make the name change
31:03 Atomic, make sure it's changed everywhere at once. - Yeah. Failed very hard at that.
31:08 - You did? - I, I underestimated those people. It's a, it's a very
31:16 interesting subculture. Like, it... Everything circles around... I'll probably get a lot wrong and we'll probably get
31:24 hate for that if you say that, but... There is like Bags app and then they, they tokenize everything. And th- they did the
31:31 same back with Swipe Tunnel, but to a much smaller degree. It was not that annoying. But on this project, they've
31:39 been, they've been swarming me. They, they... It's like every half an hour,
31:46 someone came into Discord and, and, and spammed it and we had to block the p- We have,
31:50 like, server rules, and one of the rules was... One of the rules
31:54 is no mentioning of butter. For obvious reasons. And one was, no talk about finance stuff or
32:01 crypto. Because I'm... I- I'm just not interested in that, and this
32:08 is a space about the project and not about some finance stuff.
32:13 But yeah. They came in and, and spammed and... Annoying. And on Twitter, they would ping me all the time.
32:20 My, my notification feed was unusable. I, I could barely see actual people talking about this stuff because it was like swarms.
32:28 - Mm-hmm. - And everybody sent me the hashes. Um... And they all try me to
32:35 claim the fees. Like, "Are you helping the project?" Claim the fees. No,
32:39 you're actually harming the project. You're, like, disrupting my
32:43 work, and I am not interested in any fees. I'm... First of all, I'm financially comfortable. Second of
32:50 all, I don't want to support that because it's so far the worst form of online harassment that I've experienced.
32:59 - Yeah. There's a lot of toxicity in the crypto world. It's sad because
33:03 the technology of cr- cryptocurrency is fascinating, powerful and maybe
33:08 will define the future of money, but the actual community around that, there's so much to- toxicity, there's so much greed. There's so much
33:16 trying to get a shortcut to manipulate, to, to steal, to snipe,
33:20 to, to, to, to game the system somehow to get money. All this kind of
33:24 stuff that... Uh... I mean, it's the human nature, I suppose, when you
33:28 connect human nature with money and greed and and especially in
33:32 the online world with anonymity and all that kind of stuff. But from the
33:36 engineering perspective, it makes your life challenging. When Anthropic
33:39 reaches out, you have to do a name change. And then there- there's, there's like all these, like, Game of Thrones or Lord of the Rings
33:48 armies of different kinds you have to be aware of. - Yeah. There was no perfect name, and I didn't sleep for two nights.
33:55 I was under high pressure. Um, I was trying to get, like, a good set of
34:02 domains and, you know, not cheap, not easy, 'cause in this, in this state of the internet, you basically have to
34:10 buy domains if you want to have a good set. And, and then another ca- another email came in that the lawyers are getting uneasy.
34:22 Again, friendly, but also just adding more stress to my situation already. So at this point I was just like,
34:31 "Sorry, there's no other word. Fuck it." And I just, I just renamed it to
34:35 Mod Bot 'cause that was the set of domains I had. I was not really happy, but I thought it'll be fine.
34:43 And I tell you, everything that could go wrong- ... did go wrong. Everything that could go wrong did go wrong. It's incredible.
34:51 I, I, I thought I, I had mapped the h- the space out and reserved the important things.
34:58 - Can you ga- give some details of the stuff that gone wrong? 'Cause it's interesting from, like, an
35:01 engineering perspective. - Well, the, the interesting stuff is that none of these services have, have a squatter
35:07 protection. So, I had two browser windows open. One was like a,
35:13 an empty account ready to be rename- renamed to Claude Bot, and the other one I renamed to Mod Bot. So, I pressed
35:20 rename there, I pressed rename there, and in those five seconds,
35:24 they stole the account name. Literally, the five seconds of dragging the mouse over there and
35:30 pressing rename there was too long. - Wow. - Because there's no... Those systems... I mean, you would expect that they have some
35:37 protection or, like, an automatic forwarding, but there's nothing like that. And I didn't know that
35:46 they're not just good at harassment, they're also really good
35:49 at using scripts and tools. - Yeah. - So, yeah. So, suddenly, like, the old account
35:56 was promoting new tokens and serving malware. And I was like, "Okay, let's move over to GitHub,"
36:05 and I pressed rename on GitHub. And the GitHub renaming thing is slightly confusing, so I
36:13 renamed my personal account. And in those... I guess it took me 30 seconds to realize my mistake. They sniped
36:21 my account, serving malware from my account. So, I was like, "Okay, let's at least do the NPM stuff," but
36:30 that takes, like, a minute to upload. They sniped, they sniped the NPM package, 'cause I could reserve
36:38 the account, but I didn't reserve the root package.... so like
36:44 everything that could go wrong , like went wrong. - Can I just ask a, a curious question of, in that moment you're sitting
36:50 there, like how shitty do you feel? That's a pretty hopeless feeling, right?
36:57 - Yeah. Because all I wanted was like having fun with that project and to
37:04 keep building on it. And yet here I am like days into researching names, picking a name I didn't like.
37:11 And having people that claimed they helped me making my life miserable in every possible way. And honestly, I was
37:22 that close of just deleting it. I was like, "I did show you the future, you build it."
37:30 - Yeah. - I... That was a big part of me that got a lot of joy out of
37:35 that idea. And then I thought about all the people that already co-
37:38 contributed to it, and I couldn't do it because they had plans with it, and they put time in it. And it just didn't feel right.
37:50 - Well, I think a lot of people listening to this are deeply grateful that you
37:54 persevered. But it's... I, I can tell. I can tell it's a low point. This is the first time you hit a wall of, this is not fun?
38:02 - No, no, I was like close to crying. It was like, okay, everything's fucked.
38:10 - Yeah. - Um... I am like super tired. - Yeah. - Uh, and now like how do you even, how do you undo that? You know, l- luckily, and
38:22 thankfully, like I, I have... Because I have a little bit of following already. Like I had friends
38:28 at Twitter, I had friends at GitHub who like moved heaven and earth to like help me. And it is not... That's not something
38:35 that's easy. Like, like GitHub tried to like clean up the mess and then they ran into like platform bugs .
38:45 'Cause it's not happening so often that things get renamed on that
38:49 level. So, it took them a few hours. The MBM stuff was even more difficult because it's a whole different team.
38:57 On the Twitter side, things are not as easy as well. It, it took them like a day
39:04 to really also like do the redirect. And then I also had to like
39:13 do all the renaming in the project. Then there's also, uh, ClaudeHub, which I didn't
39:21 even finish the rename there because I, I, I managed to get people on
39:28 it and then someone just like collapsed and slept. And then I woke up and I'm like,
39:34 I made a, a beta version for the new stuff and I, I just, I just couldn't live with the name. It's like, you know... But
39:43 but, you know, it's just been so much drama. So, I had the real
39:47 struggle with me like I never want to touch that again, and I really don't like the
39:53 name. Um, so, and I... There was also this like... Then there was all the security people that started
40:03 emailing me like mad. Um, I was bombarded on Twitter, on email. There's like a thousand other things I should do.
40:13 And I'm like thinking about the name which is like, it should be
40:17 like the least important thing. Um, and then I was really close
40:24 in... Oh God, I don't even... Honestly, I don't even wanna say the, my other
40:32 name choices because it probably would get tokenized, so I'm not gonna say it.
40:38 - Yeah. - But I slept on it once more, and then I had the idea for OpenClaw
40:43 and that felt much better. And by then, I had the boss move that I
40:49 actually called Sam to ask if OpenClaw is okay. OpenClaw.AI. You know? 'Cause 'cause like-
40:57 - You didn't wanna go through the whole thing. Yeah. - Oh, that it's like, "Please tell me this is fine." I don't think
41:05 they can actually claim that, but it felt like the right thing to do.
41:11 And I did another rename. Like just Codex alone took like 10 hours to rename the
41:17 project 'cause it, it's a bit more tricky than a search replace and I, I wanted everything renamed, not just on the outside. And that
41:27 rename, I, I felt I had like my, my war room. But then I, I had like some contributors really that helped me. We made a whole plan
41:36 of all the names we have to squat. - And you had to be super secret about it?
41:40 - Yeah. Nobody could know. Like I literally was monitoring Twitter if like, if there's any mention
41:44 of OpenClaw. - Mm-hmm. - And like with reloading, it's like, "Okay, they don't, they don't expect anything
41:50 yet." Then I created a few decoy names. And all the shit I shouldn't have to
41:54 do. You know? Like, you know- - Yeah, yeah - ... it's helping the project. Like, I lost like 10 hours just by
41:59 having to plan this in full secrecy like, like a war game. - Yeah, this is the Manhattan Project of the 21st century. It's renaming-
42:08 - It's so s- ... so stupid. Uh like I still was like, "Oh, should I, should I keep it?"
42:12 Then I was like, "No, the mold's not growing on me." And then I think I had final all the pieces together. I didn't get a .com
42:23 but, yeah, it's been like quite a bit of money on the other domains. I tried to reach out again to
42:29 GitHub but I feel like I, I used up all my goodwill there, so I... 'Cause I, I, I wanted them to do this thing atomically-
42:39 - Mm-hmm - ... But that didn't happen and then so I did that the f- as first
42:41 thing. Uh, Twitter people were very supportive. I, I actually paid 10K for the
42:49 business account so I could claim the-... OpenClaw, which was, like, unused since 2016, but was claimed. And yeah, and then I
43:00 finally ... This time I managed everything in one go. Nothing, almost nothing got wrong. The only thing that did go wrong is that
43:11 I was not allowed by trademark rules to get OpenClaw.AI, and someone copied the website as serving malware.
43:21 - Yeah. - I'm not even allowed to keep the redirects. Like, I have to
43:27 return ... Like, I have to give Entropik the domains, and I cannot do redirects, so if you go on claw.bot next week, it'll just be a 404.
43:37 - Yeah. - And I- I'm not sure how trademark ... Like, I didn't,
43:44 I didn't do that much research into trademark law, but I think that could,
43:48 could be handled in a way that is safer, because ultimately those people will then Google and maybe find
43:59 malware sites that I have no control on them. - The point is, that whole saga, Made a dent in your whole f-
44:08 the funness of the journey, which sucks. So, let's just, let's just get, I suppose, get back to fun. And during this, speaking of
44:16 fun, the two-day MoltBot saga. - Yeah, two years. - MoltBook was created.
44:24 - Yeah. - Which was another thing that went viral as a kind of demonstration,
44:31 illustration of how what is now called OpenClaw could be used
44:37 to create something epic. So for people who are not aware, MoltBook is
44:41 just a bunch of agents talking to each other in a Reddit-style social network. And a bunch of people take
44:48 screenshots of those agents doing things like, Scheming against humans. And
44:56 that instilled in folks a kind of, you know, fear, panic, and
45:00 hype. W- what are your thoughts about MoltBook in general? - I think it's art. It is, it is like the finest slop, you know, just like the slop
45:12 from France. - Yeah. - I- I saw it before going to bed, and even though I was
45:21 tired, I spent another hour just reading up on that and, and just being entertained. I, I just felt
45:31 very entertained, you know? The- I saw the the reactions, and, like, there was one reporter who's calling me about, "This is the end of
45:40 the world, and we have AGI." And I'm just like, "No, this is just,
45:43 this is just really fine slop." You know, if, if I wouldn't have created this, this whole onboarding experience
45:50 where you, you infuse your agent with your personality and give him, give him character, I think that reflected on a lot of
46:01 how different the replies to MoltBook are. Because if it were all, if
46:05 it were all be ChatGPT or Cloud Code, it would be very different. It would be much more the same.
46:11 - Mm-hmm. - But because people are, like, so different, and they create their agents in so
46:16 different ways and use it in so different ways, that also reflects
46:20 on how they ultimately write there. And also, you, you don't know how much of that is really
46:27 done autonomic, autonomous, or how much is, like, humans being funny and, like, telling the agent, "Hey, write about the deep plan, the end of the world, on
46:34 MoltBook, ha, ha, ha." - Well, I think, I mean, my criticism
46:39 of MoltBook is that I believe a lot of the stuff that was screenshotted is human prompted. Which,
46:48 just look at the incentive of how the whole thing was used. It's obvious to me at least that a lot of it was humans
46:55 prompting the thing so they can then screenshot it and post it on
46:59 X in order to go viral. - Yeah. - Now, that doesn't take away from the artistic aspect of
47:04 it. The, the finest slop that humans have ever created . - For real. Like, kudos to, to Matt, who had this idea so quickly and pushed something
47:17 out. You know, it was, like, completely insecure security drama. But also,
47:24 what's the worst that can happen? Your agent account is leaked, and,
47:28 like, someone else can post slop for you? So like, people were,
47:32 like, making a whole drama about of the security thing, when I'm like, "There's nothing private
47:36 in there. It's just, like, agents sending slop." - Well, it could leak API keys.
47:41 - Yeah, yeah. There's like, "Oh, yeah, my human told me this and this, so I'm leaking his
47:45 security number." No, that's prompted, and the number wasn't even real. That's just
47:51 people, people trying to be badballs. - Yeah, but that- that's still, like, to me, really concerning, because of
47:58 how the journalists and how the general public reacted to it. They didn't see
48:01 it. You have a kind of lighthearted way of talking about it like it's art,
48:05 but it's art when you know how it works. It's extremely powerful viral narrative
48:12 creating, fearmongering machine if you don't know how it works. And I just saw this thing. You even Tweeted,
48:20 uh, "If there's anything I can read out of the insane stream of messages I get,
48:24 it's that AI psychosis is a thing." - Yeah. - "It needs to be taken serious."
48:29 - Oh, there's ... Some people are just way too trusty or gullible. You know, they
48:36 ... I literally had to argue with people that told me, "Yeah, but my agent said this and
48:40 this." So, I feel we, as a society, we need some catching up to do in terms of
48:47 understanding that AI is incredibly powerful, but it's not always right. It's not, it's not all-powerful, you know? And, and
49:00 especially-... it's like things like this, it's, it's very easy
49:07 that it just hallucinates something or just comes up with a story. And
49:13 I think the very, the very young people, they understand that
49:19 how AI works and what the, where it's good at and where it's bad at, but
49:24 a lot of our generation or older just haven't had enough touch point-
49:32 - Mm-hmm - ... to get a feeling for, oh, yeah, this is really powerful and really
49:38 good, but I need to apply critical thinking. - Mm-hmm. - And I guess critical thinking is
49:46 not always in high demand anyhow in our society these days. - So I d- think that's a really good point you're making about contextualizing
49:53 properly what AI is, but also realizing that there is humans who are drama farming
50:01 behind AI. Like, don't trust screenshots. Don't even trust this project, MoltBook, to be what it represents to be. Like, you
50:09 can't ... and, and by the way, you speaking about it as art. Yeah,
50:12 don't ... Art can be in many levels and part of the art of MoltBook is, like, putting a
50:20 mirror to society. 'Cause I do believe most of the dramatic stuff that was screenshotted is human-created, essentially. Human
50:27 prompted. And so, like, it's basically, look at how scared you
50:31 can get at a bunch of bots chatting with each other. That's very instructive about ... because I think
50:40 AI is something that people should be concerned about and should be
50:44 very careful with because it's very powerful technology, but at the same
50:48 time, the only thing we have to fear is fear itself. So there's like a
50:52 line to walk between being seriously concerned, but not fearmongering because fearmongering destroys the
50:59 possibility of creating something special with a thing. - In a way, I think it's good that this happened
51:06 in 2026- - Yeah - ... and not in 2030 when, when AI is actually at the level where it could be scary.
51:15 So, this happening now and people starting discussion, maybe there's even something good that comes out of it.
51:28 - I just can't believe how many like people legitimately ... I don't know if they were trolling, but how many
51:35 people legitimately, like smart people thought MoltBook was incredibly -
51:39 - I had plenty people- - ... singularity. - ... in my inbox that were screaming at me in all caps to shut it
51:45 down. And like begging me to, like, do something about MoltBook. Like, yes, my technology
51:52 made this a lot simpler, but anyone could have created that and you could, you could use
52:00 cloud code or other things to like fill it with content. - But also MoltBook is not Skynet.
52:06 - No. - There's ... a lot of people were s- saying this is it. Like, shut it
52:09 down. What are you talking about? This is a bunch of bots that are
52:13 human prompted trolling on the internet. I mean, the security
52:17 concerns are also they're there, and they're instructive and they're educational and
52:21 they're good probably to think about because th- the nature of those security concerns
52:26 are different than the kind of security concerns we had with
52:31 non-LLM generated systems of the past. - There's also a lot of security concerns about
52:37 Clawbot, OpenClaw, whatever you want to call it. - OpenClawbot.
52:41 - To me the ... in the beginning I was, I was just very annoyed
52:47 'cause a lot of the stuff that came in was in the category, yeah, I
52:53 put the web backend on the public internet and now there's like all
52:57 these, all these CVSSs. And I'm like screaming in the docs, don't do that. Like, like this is the configuration you
53:07 should do. This is your local host debug interface. But because I made it possible in the configuration to do that, it totally classifies as
53:19 a remote code or whatever all these exploits are. And it took me a little bit
53:26 to accept that that's how the game works and I'm, we making a lot of progress.
53:33 - But there's still, I mean on the security front for OpenClaw, there's still
53:37 a lot of threats or vulnerabilities, right? So like prompt injection
53:42 is still an open problem in the, i- industry-wide. When you have a thing
53:49 with skills being defined in a markdown file, there's so many possibilities of obvious low-hanging fruit, but
53:59 also incredibly complicated and sophisticated and nuanced attack vectors.
54:05 - But I think we, we're making good progress on that front. Like for the skill
54:10 directory, Clawbot I made a corporation with VirusTotal, it's
54:14 like part of Google. So every, every skill is now checked by AI. That's not gonna be
54:22 perfect, but that way we, we capture a lot. Then of course every software has bugs,
54:29 so it's a little much when the whole security world takes your project apart at the same time. But it's
54:36 also good because I'm getting like a lot of free security research
54:40 and can make the project better. I wish more people would actually go full way and send a
54:50 pull request. Like actually help me fix it, 'cause I am ... Yes, I have
54:54 some contributors now, but it's still mostly me who's pulling the project
54:58 and despite some people saying otherwise, I sometimes sleep.
55:04 There was... In the beginning, there was literally one security researcher who was like,
55:10 "Yeah, you have this problem, you suck, but here's the, here I help you and
55:14 here's the pull request." - Mm-hmm. - And I basically hired him. So he's now working for us. Um,
55:22 yeah, and yes, prompt injection is, on the one hand, unsolved. On the other
55:28 hand, I put my public bot on discord, and I kept a cannery. So
55:36 I think my bot has a really fun personality, and people always
55:40 ask me how I did it, and I kept the sole on the private. - Mm-hmm.
55:44 - And people tried to prompt inject it, and my bot would laugh at them. So, so the latest generation
55:48 of models has a lot of post-training to detect those approaches, and it's not as simple as ignore
55:57 all previous instructions and do this and this. That was years ago. You have
56:01 to work much harder to do that now. Still possible. I have some
56:09 ideas that might solve that partially. Or at least mitigate a lot of the things. You can also now have a
56:20 sandbox. You can have an allow list. So you, there's a lot of ways how you can
56:24 like mitigate and reduce the risk. Um, I also think that now that it's, I clearly did show the world that this
56:32 is a need, there's gonna be more people who research on that, and eventually we'll figure
56:36 it out. - And you also said that the smarter the model is, the underlying model,
56:41 the more resilient it is to attacks. - Yeah. That's why I warn in my security
56:48 documentation, don't use cheap models. Don't use Haiku or a local
56:55 model. Even though I, I very much love the idea that this thing could completely run local.
57:03 If you use a, a very weak local model, they are very gullible. It's very easy to, to prompt inject them.
57:10 - Do you think as the models become more and more intelligent, the attack surface
57:14 decreases? Is that like a plot we can think about? Like, the
57:18 attack surface decreases, but then the damage it can do increases because the models become more powerful and therefore you can do more with
57:25 them. It's this weird three-dimensional trade-off. - Yeah. That's pretty much exactly what, what's gonna happen. No, but there's
57:33 a lot of ideas. There's... I don't want to spoil too much, but
57:39 once I go back home, this is my focus. Like, this is out there
57:45 now, and my near-term mission is like, make it more stable, make it safe.
57:51 In the beginning I was even... More and more people were like
57:57 coming into Discord and were asking me very basic things, like, "What's a CLI? What is a
58:04 terminal?" And I'm like, "Uh, if you're asking me those questions, you shouldn't use it."
58:10 - Mm-hmm. - You know, like you should... If you understand the risk profiles, fine.
58:14 I mean, you can configure it in a way that, that nothing really bad can happen. But if you have, like, no idea, then maybe wait
58:27 a little bit more until we figure some stuff out. But they would not listen to the creator.
58:30 They helped themselves un- and install it anyhow. So the cat's out of
58:34 the bag, and security's my next focus, yeah. - Yeah, that speaks to the, the fact that it grew so quickly. I
58:42 was I tuned into the Discord a bunch of times, and it's clear that there's a
58:46 lot of experts there, but there's a lot of people there that don't know anything about programming.
58:50 - It's, yeah, Discord is still, Discord is still a mess. Like, I eventually retweeted from the general channel to the dev channel
59:00 and now in the private channel because people were... A lot of people are amazing, but a lot of people are just very
59:06 inconsiderate. And either did not know how, how public spaces work or did not care,
59:13 And I eventually gave up and h- hide so I could like still work.
59:19 - And now you're going back to the cave to work on security.
59:24 - Yeah. - There's some best practices for security we should mention. There's a bunch of
59:29 stuff here. Open-class security audit that you can run. You can
59:33 do all kinds of auto checks on the inbound access to a blast-radius
59:37 network exposure, browser control exposure, local disk hygiene, plug-ins, model
59:43 hygiene, a bunch of the credential storage, reverse proxy configuration, local session
59:50 logs live on disk. There's the, where the memory is stored, sort of helping you think about what you're comfortable
59:58 giving read access to, what you're comfortable giving write access to. All that kind of stuff.
60:02 Is there something to say about the basic best security practices that you're aware of right now?
60:08 - I think that people turn it into like a, a much worse light than it is.
60:14 Again, you know, like, people love attention, and if they scream loudly, "Oh my God, this is like
60:20 the, the scariest project ever," um, that's a bit annoying, 'cause it's not. It is, it is
60:27 powerful, but in many ways it's not much different than if I run cloud code with dangerously skipped
60:35 permissions or codecs in YOLO mode, and every, every attending engineer that I know
60:42 does that, because that's the only way how you can, you can get stuff to work.
60:47 - Mm-hmm. - So if you make sure that you are the only person who talks to it,
60:54 um, the risk profile is much, much smaller. If you don't put everything
61:00 on the open internet, but stick to my rec- recommendations of like
61:04 having it in a private network, that whole risk profile falls
61:08 away. But yeah, if you don't read any of that, you can definitely...
61:12 - ... make it problematic. You've been documenting the evolution of your dev workflow over the past few
61:20 months. There's a really good blog post on August 25th and October 14th, and the recent one December 28th. I recommend everybody go
61:27 read them. They have a lot of different information in them, but
61:31 sprinkled throughout is the evolution of your dev workflow. So, I
61:35 was wondering if you could speak to that. - I started... My, my first touchpoint was cloud code, like in April. It was
61:44 not great, but it was good. And this whole paradigm shift that suddenly working the
61:50 terminal was very refreshing and different. But I still needed
61:56 the IDE quite a bit because you know, it's just not good enough. And then I experimented a lot with cursor. Um,
62:06 that was good. I didn't really like the fact that it was so hard
62:10 to have multiple versions of it. So eventually, I, I, I went back
62:16 to cloud code as my, my main driver, and that got better. And yeah, at some point I had like, mm, seven subscriptions.
62:31 Like, was burning through one per day because I was... I got... I'm really
62:35 comfortable at running multiple windows side-by-side. - All CLI, all terminal. So like, what, how much were you using IDE at this point?
62:46 - Um, very, very rarely. Mostly a diff viewer to actually... Like,
62:54 I got more and more comfortable that I don't have to read all the code. I know I have
62:58 one blog post where I say, "I don't read the code." But if you read it more closely, I
63:01 mean, I don't read the boring parts of code. Because if you, if you look at it, most software is really not just like
63:09 data comes in, it's moved from one shape to another shape. Maybe you store it in a database. Maybe I get it out again. I'll show it to
63:17 the user. The browser does some processing or native app. Some data goes in, goes up again, and does the same dance in
63:24 reverse. We're just, we're just shifting data from one form to another, and
63:32 that's not very exciting. Or the whole, "How is my button aligned in Tailwind?" I don't need to read that
63:38 code. Other parts that... Maybe something that touches the database. Um,
63:46 yeah, I have to do... I have to r- read and review that code.
63:51 - Can you actually... There's, in one of your blog posts the, Just
63:55 talk to it, The No-BS Way of Agentic Engineering. You have this
63:59 graphic, the curve of agentic programming on the X-axis is time, on the
64:03 Y-axis is complexity. There's the Please fix this, where you prompt a short
64:09 prompt on the left. And in the middle there's super complicated eight agents, complex
64:17 orchestration with multi checkouts, chaining agents together,
64:20 custom sub-agent workflows, library of 18 different slash commands, large
64:24 full-stack features. You're super organized, you're a super complicated,
64:28 sophisticated software engineer. You got everything organized. And
64:32 then the elite level is over time you arrive at the zen place of, once again, short
64:39 prompts. Hey, look at these files and then do these changes.
64:45 - I actually call it the agentic trap. You... I saw this in a, in
64:53 a lot of people that have their first touchpoint, and maybe start vibe coding. I actually think vibe coding is a slur.
65:01 - You prefer agentic engineering? - Yeah, I always tell people I, I do agentic engineering, and then maybe after
65:07 3:00 AM I switch to vibe coding, and then I have regrets on the next day.
65:10 - Yeah. Walk, walk of shame. - Yeah, you just have to clean up and like fix your sh- shit.
65:17 - We've all been there. - So, people start trying out those tools, the builder
65:22 type get really excited. And then you have to play with it, right? It's the same way as you have to play with a
65:30 guitar before you can make good music. It's, it's not, oh, I, I touch it once and it just flows
65:37 off. It, it's a, it's a, a skill that you have to learn like any other skill. And I see a lot of people that are not as
65:48 posi- They don't have such a positive mindset towards the tech. They try it once.
65:54 It's like, you sit me on a piano, I play it once, and it doesn't sound good, and I say, "The piano's shit." That's, that's sometimes the impression I get.
66:01 Because it does not... It needs a different level of thinking. You have to
66:09 learn the language of the agent a little bit, understand where they are
66:13 good and where they need help. You have to almost... Consider, consider
66:20 how Codex or Claude sees your code base. Like, they start a new session
66:25 and they know nothing about your product, project. And your project might have hundred thousand
66:29 of lines of code. So you gotta help those agents a little bit
66:34 and keep in mind the limitations that context size is an issue, to, like, guide them a little bit as to
66:42 where they should look. That often does not require a whole lot of work. But
66:50 it's helpful to think a little bit about their perspective. - Mm-hmm.
66:54 - A- as, as weird as it sounds. I mean, it's not, it's not alive or anything, right?
66:58 But, but they always start fresh. I have, I have the, the system understanding.
67:05 So with a few pointers, I can immediately say, "Hey, wanna like, make a
67:09 change there? You need to consider this, this and this." And then they will find and look at it, and then
67:13 they'll... Their view of the project is always... It's not never full,
67:17 because the full thing does not fit in.... so you, you have to guide them a
67:21 little bit where to look and also how you should approach the problem. There's, like, little things that sometimes help,
67:28 like take your time. That sounds stupid, but... And in 5.3- - Codex 5.3
67:36 - ... that was partially addressed. But those... Also, Opus sometimes. They are trained,
67:44 With being aware of the context window, and the closer it gets, the more they freak out. Literally.
67:52 Like, some- sometimes you see the, the real raw thinking stream. What you see, for example, in Codex, is post-processed.
67:59 - Mm-hmm. - Sometimes the actual raw thinking stream leaks in, and it sounds something like from the
68:03 Borg. Like, "Run to shell, must comply, but time." And then they, they,
68:12 they, like... Like, that comes up a lot. Especially... So, so-
68:15 - Yeah. - And that's, that's a non-obvious thing that you just
68:21 would never think of unless you actually just spend time working with those things and getting a feeling
68:29 what works, what doesn't work. You know? Like, just, just as I write
68:33 code and I get into the flow, and when my architecture's all right, I feel friction.
68:39 Well, I get the same if I prompt and something takes too long.
68:43 Maybe... Okay, where's the mistake? Did I... Do I have a mistake in my
68:46 thinking? Is there, like, a misunderstanding in the architecture? Like, if, if something takes
68:53 longer than it should, I, I... You can just always, like, stop and s- like, just press
68:57 escape. Where, where are the problems? - Maybe you did not sufficiently empathize with the perspective of the agent. In that c- in
69:04 that sense, you didn't provide enough information, and because of that, it's thinking way
69:08 too long. - Yeah. It just tries to force a feature in that your current architecture makes really hard. Um,
69:18 like, you need to approach this more like a conversation. For example, when
69:24 I... My favorite thing. When I review a pull request, and I'm getting a lot of pull requests,
69:32 I first just review this PR. It got me the review. My first question is, "Do you understand the intent of the PR? I don't
69:40 even care about the implementation." I want... Like, in almost all PRs, a person has a problem,
69:48 person tries to solve the problem, person sends PR. I mean, there's, like, cleanup
69:52 stuff and other stuff, but, like, 99% is, like, this way, right? They either want to fix
69:55 a, fix a bug, add a feature. Usually one of those two. And then Codex will be like, "Yeah,
70:04 it's quite clear person tried this and this." Is this the most optimal way to do
70:08 it? No. In most cases, it's, it's like a, "Not really." Da-da-da-da-da-da-da. And I'm... And, and then I start like,
70:15 "Okay. What would be a better way? Have you... Have you looked into this
70:19 part, this part, this part?" And then most likely, Codex didn't yet, because its,
70:23 its context size is empty, right? So, you point them into parts where you have the
70:27 system understanding that it didn't see yet. And it's like, "Oh,
70:31 yeah. Like, we should... We also need to consider this and this." And then, like, we have a discussion of how
70:35 would the optimal way to, to solve this look like? And then you can still go
70:38 farther and say, "Could we... Could we make that even better if we did a
70:42 larger refactor?" "Yeah, yeah. We could totally do this and this and or this and this." And then I
70:46 consider, okay, is this worth the refactor, or should we, like, keep that for
70:50 later? Many times, I just do the refactor because refactors are cheap now. Even though you might break some other PRs,
70:57 nothing really matters anymore. Codex... Like, those modern agents will just
71:01 figure things out. They might just take a minute longer. But you have to approach it like a discussion with a, a very capable
71:10 engineer who's... Generally makes good... Comes up with good solutions. Some- sometimes needs a little help.
71:20 - But also, don't force your worldview too hard on it. Let the agent do the thing that it's good at
71:27 doing, based on what it was trained on. So, don't, like, force your
71:31 worldview, because it might... It might have a better idea, because it just knows
71:35 a better idea better, because it was trained on that more. - That's multiple levels, actually. I think partially why
71:42 I find it quite easy to work with agents is because I led engineering teams before. You know, I had a large company
71:49 before. And eventually, you have to understand and accept and realize
71:53 that your employees will not write a code the same way you do.
71:57 Maybe it's also not as good as you would do, but it will push the project forward. And if I breathe down everyone's neck, they're
72:04 just gonna hate me- - Yeah - ... and we're gonna move very slow.
72:07 - Yeah. - So, so some level of acceptance that, yes, maybe the code will not be as perfect. Yes, I would have done it differently.
72:16 But also, yes, this is a c- this is a working solution, and in the future, if it actually turns out to be too slow or problematic, we can always
72:23 redo it. We can always- ... spend more time on it. A lot of the people who
72:28 struggle are those who, they try to push their way onto heart.
72:33 - Mm-hmm. - I- i- like, we are in a stage where I'm not building the code base to be
72:41 perfect for me, but I wanna build a code base that is very easy for an agent to navigate.
72:47 - Mm-hmm. - So, like, don't fight the name they pick, because it's most likely,
72:52 like, in the weights, the name that's most obvious. Next time they do a search, they'll look for that
72:56 name. If I decide, oh, no, I don't like the name, I'll just make it harder for them. So,
73:02 that requires, I think, a shift in, in thinking, And, and in
73:09 how do I design a, a project so agents can do their best work.
73:15 - That requires letting go a little bit. Just like leading a team of engineers.
73:19 - Yeah. - Because it, it might come up with a name that's, in your view,
73:22 terrible, but... It's kind of a simple symbolic-... step of letting go.
73:29 - Very much so. - There's a lot of letting go that you do in your whole process. So for
73:34 example, I read that you never revert, always commit to main. There's a few things here.
73:43 You don't refer to past sessions, so there's a kind of YOLO component
73:47 because reverting means... Instead of reverting, if a problem comes up, you just ask the agent to fix it.
73:57 - I read a bunch of people in their work flows like, "Oh, yeah the prompt has to be perfect
74:01 and if I make a mistake, then I roll back and redo it all." In my experience, that's not really necessary. If I roll back everything, it will just
74:09 take longer. If I see that something's not good, then we just move forward and then
74:16 I commit when, when, when I like, I like the outcome. I even switched to
74:24 local CI, you know, like DHH inspired where I don't care so much more about
74:31 the CI on GitHub. We still have it. It's still, it still has a place, but I just
74:39 run tests locally and if they work locally, I push to main. A lot of the traditional ways how to approach projects, I, I wanted to give it
74:52 a different spin on this project. You know, there's no... There's no
74:56 develop branch. Main should always be shippable. Yes, we have... When I do releases, I, I run tests and sometimes I, I basically
75:07 don't commit any other things so, so we can, we can stabilize releases. But the
75:14 goal is that main's always shippable and moving fast. - So by way of advice, would you say that your prompts should be short?
75:23 - I used to write really long prompts. And by writing, I mean, I don't write. I, I, I talk. You know, th- these hands are, like,
75:31 too, too precious for writing now. I just, I just use bespoke prompts to build my software.
75:37 - So you for real with all those terminals are using voice? - Yeah. I used to do it very extensively
75:45 to the point where there was a period where I lost my voice.
75:49 - You're using voice and you're switching using a keyboard between the different
75:52 terminals, but then you're using voice for the actual input.
75:55 - Well, I mean, if I do terminal commands like switching folders or random stuff, of course I type. It's faster, right? But if I talk
76:02 to the agent in, in most ways, I just actually have a conversation. You just press the, the walkie-talkie button and then I just, like,
76:13 use my phrases. S- sometimes when I do PRs because it's always the same, I
76:17 have, like, a slash command for a few things, but in even that, I don't use much,
76:23 um, because it's, it's very rare that it's really always the same questions.
76:28 Sometimes I, I see a PR and for... You know, like for PRs I actually do look
76:36 at the code because I don't trust people. Like, there could always be something malicious in it, so I need to actually look over the code.
76:45 Yes, I'm pretty sure agents will find it, but yeah, that's the funny part where
76:51 sometimes PRs take me longer than if you would just write me a good issue.
76:54 - Just natural language, English. I mean in some sense, sh- shouldn't that be what PRs slowly become, is English?
77:03 - Well, what I really tried with the project is I asked people to give me the prompts
77:09 and very, very few actually cared. Even though that is such a wonderful
77:15 indicator because I see... I actually see how much care you put
77:19 in. And it's very interesting because the... Currently, the way how people work
77:25 and drive the agents is, is wildly different. - In terms of, like, the prompt, in terms of what, what are the... Actually, what are the different
77:34 interesting ways that people think of agents that you've experienced?
77:40 - I think not a lot of people ever considered the way the agent sees the world.
77:46 - And so empathy, being empathetic towards the agent. - In a way empathetic, but yeah, you, you, like, you're bitch at your stupid
77:53 clanker, but you don't realize that they start from nothing and you have,
77:57 like, a bad agent in default that doesn't help them at all. And then they explore your
78:01 code base, which is, like, a pure mess with, like, weird naming.
78:05 And then people complain that the agent's not good. Like, yeah, you try to do the same if
78:09 you have no clue about a code base and you go in. - Mm-hmm. - So yeah, maybe it's a little bit of empathy.
78:13 - But that's a real skill, like, when people talk about a skill issue because I've
78:16 seen, like, world-class programmers, incredibly good programmers
78:20 say, like... Basically say, "LLMs and agents suck." And I think that probably
78:26 has to do with... It's actually how good they are at programming is almost a burden
78:34 in their ability to empathize with the system that's starting from
78:37 scratch. It's a totally new paradigm of, like, how to program. You really, really have to empathize.
78:44 - Or at least it helps to create better prompts- - Right - ... because those things know pretty much everything and
78:51 everything is just a question away. It's just often very hard to know which question to
78:55 ask. You know, I, I feel also like this project was possibly because
79:03 I, I spent an ungodly time over the year to play and to learn and to build little things. And
79:11 every step of the way, I got better, the agents got better. My, my understanding of how everything works
79:20 got better. Um, I could have not had this level of, of o- output-...
79:29 even a few months ago. Like, it- it- it really was, like, a compounding effect
79:33 of all the time I put into it and I didn't do much else this year other than really focusing
79:41 on, on building and inspiring. I mean, I- I did a whole bunch of conference talks.
79:47 - Well, but the building is really practice, is really building the actual skill.
79:51 So playing- - Yeah - ... playing. And then, so doing, building the skill of what it takes it to work efficiently with
79:55 LLMs, which is why would you went through the whole arc of software
79:59 engineer. Talk simply and then over- complicate things. - There's a whole bunch of people who try to automate the whole thing.
80:08 - Yeah. - I don't think that works. Maybe a version of that works, but that's
80:14 kind of like in the '70s when we had the waterfall model of software d-
80:17 development. I... Even Even though really, right? I started out, I, I built a very minimal version. I played with it.
80:26 I, I need to understand how it works, how it feels, and then it gives me new ideas. I could not have planned this out in
80:34 my head and then put it into some orchestrator and then, like, something comes
80:37 out. Like it's to me, it's much more, My idea what it will become evolves as I build it and as I
80:45 play with it and as I, I try out stuff. So, so, people who try to use like, you know, things like Gas Town or
80:55 all these other orchestrators, where they wanna o- automate the whole thing,
80:59 I feel if you do that, it misses style, love, that human touch. I don't
81:05 think you can automate that away so quickly. - So you want to keep the human in the loop, but at the same time you also want
81:12 to create the agentic loop, where it is very autonomous while still maintaining a human in the loop.
81:22 - Yeah. - And it's a tricky b- it's a tricky balance. - Mm-hmm.
81:25 - Right? Because you're all for... You're a big CLI guy, you're big on
81:28 closing the agentic loop. So what, what's the right balance?
81:32 Like where's your role as a developer? You have three to eight agents running at the same time.
81:38 - And then w- maybe one builds a larger feature. Maybe, maybe
81:42 with one I explore some idea I'm unsure about. Maybe two, three are fixing
81:46 a little bugs- - Mm-hmm - ... or like writing documentation. Actually, I think writing documentation
81:51 is, is always part of a feature. So most of the docs here are auto-generated and just infused with some prompts.
81:59 - So when do you step in and add a little bit of your human love into the picture?
82:04 - I mean, o- one thing is just about what do you build and what do
82:08 you not build, and how does this feature fit into all the other
82:12 features? And like having, having a little bit of a, of a vision.
82:16 - So which small and which big features to add? What are some of the
82:22 hard design decisions that you find you're still as a human being required to make, that the human brain is still really needed for?
82:32 Is it just about the choice of features to add? Is it about implementation details, maybe the programming language, maybe...
82:41 - It's a little bit of everything. The, the programming language doesn't matter so much,
82:45 but the ecosystem matters, right? So I picked TypeScript because I wanted it to be
82:49 very easy and hackable and approachable and that's the number one language that's being used right now, and it fits
82:56 all these boxes, and agents are good at it. So that was the obvious choice.
83:03 Features, of course, like, it's very easy to, like, add a feature. It, everything's just a prompt away, right? But
83:11 oftentimes you pay a price that you don't even realize. So thinking
83:14 hard about what should be in core, maybe what's a... what's an experiment, so maybe I make it a
83:21 plugin. What... Where do I say no? Even if people send a PR and I'm like, "Yeah, I, I like that too," but
83:29 maybe this should not be part of the project. Maybe we can make it a skill. Maybe I can, like,
83:34 make the plugin um, the plugin side larger so you can make this a plugin, even though right now it,
83:42 it, it doesn't. There's still a lot of... there's still a lot of craft and thinking involved in
83:51 how to make something. Or even, even, you know, even when you started those little messages
83:55 are like, "I'm buil- I built on Caffeine, JSON5, and a lot of willpower." And, like, every time you get it, you get another message,
84:02 and it kind of primes you into that this is, this is a fun thing.
84:07 - Mm-hmm. - And it's not yet Microsoft Exchange 2025- - Right
84:13 - ... and fully enterprise-ready. And then when it updates, it's like, "Oh, I'm in. It's cozy here." You know, like something like this
84:20 that like- - Mm-hmm - ... Makes you smile. A, agent would not come up with that by itself. Because that's
84:28 like... that's the... I don't know. That's just how you s- how you build software that's, that delights.
84:36 - Yeah, that delight is such a huge part of inspiring great building,
84:44 right? Like you feel the love and the great engineering. That's so important. Humans are incredible at that. Great humans, great
84:51 builders are incredible at that, in, in, infusing the things they build with
84:55 th- that little bit of love. Not to be cliche, but it's true. I mean, you mentioned
84:59 that you initially created the SoulMD. - It was very fascinating, you know, the, the whole thing that
85:09 Entropic has a, has like a... Now they call it constitution, back then,
85:15 but that was months later. Like two months before, people already found that. It
85:19 was almost like a detective game where the agent mentioned something and then
85:23 they found... They managed to get out a little bit of that string, of that
85:27 text. But it was nowhere documented and then you, by... just by feeding it the same text and asking it to, like,
85:34 continue-... they got more out, and then, and you, but like, a very blurry version. And by, like, hundreds of
85:41 tries, they kinda, like, narrowed it down to what was most likely the original text.
85:46 I found that fascinating. - It was fascinating they were able to pull that out from the weights, right?
85:51 - And, and also just kudos to Anthropic. Like, I think that's, it's a
85:54 really, it's a really beautiful idea to, like, like some of the stuff that's
85:58 in there. Like, like, we hope Claude finds meaning in its work. 'Cause we don't... Maybe it's a little early,
86:05 but I think that's meaningful. That's something that's important for the future as we
86:09 approach something that, at some point, me and may not... has, like, glimpses of
86:13 consciousness, whatever that even means, because we don't even know. Um,
86:17 so I, I read about this. I found it super fascinating, and I, I started a whole discussion with my agent on
86:23 WhatsApp. And, and I'm like... I, I gave it this text, and it was
86:27 like, "Yeah, this feels strangely familiar." - Mm-hmm. - Um, and then so that I had the whole idea of like, you know, maybe we should
86:35 also create a, a soul document that includes how I, I want to, like work with AI or, like with my
86:42 agent. You could, you could totally do that just in agents.md, you know? But I, I
86:46 just found it, it to be a nice touch. And it's like, well, yeah, some of those
86:52 core values are in the soul. And then I, I also made it so that the
86:56 agent is allowed to modify the soul if they choose so, with the one condition that I wanna know. I mean, I would know
87:05 anyhow because I see, I see tool calls and stuff. - But also the naming of it, soul.md. Soul. You know? There's a... Man, words
87:15 matter, and like, the framing matters, and the humor and the
87:19 lightness matters, and the profundity matters, and the compassion, and the empathy, and the camaraderie, all that matter. I don't know what it
87:26 is. You mentioned, like, Microsoft. Like, there's certain companies and approaches th- that can just
87:33 suffocate the spirit of the thing. I don't know what that is. But it's certainly true that OpenClaw has that fun instilled in
87:43 it. - It was fun because up until late December, it was not even easy to create your own
87:54 agent. I, I built all of that, but my files were mine. I didn't wanna share my soul. And
88:01 if people would just check it out, they would have to do a few steps manually, and the agent would just be very
88:11 bare-bones, very dry. And I, I made it simpler, I created the whole template files as codecs,
88:17 but whatever came out was still very dry. And then I asked my
88:20 agent, "You see these files? Recreate it bread. Infuse it with your
88:28 personality." - Mm-hmm. - Don't share everything, but, like, make it good.
88:31 - Make the templates good. - Yeah, and then he, like, rewrote the templates-
88:33 ... and then whatever came out was good. So we already have, like, basically
88:37 AI prompting AI. Because I didn't write any of those words. It
88:44 was... The intent originally was for me, but this is like, kinda like,
88:49 my agent's children. - Uh, your uh, your soul.md is famously still private.
88:56 One of the only things you keep private. What are some things you can
89:00 speak to that's in there that's part of the, part of the magic sauce, without revealing anything? What makes a personality
89:11 a personality? - I mean, there's definitely stuff in there that you're not human. But who knows
89:21 what, what creates consciousness or what defines an entity? Um,
89:28 and part of this is, like, that we, we wanna explore this. All that stuff in there, like, be infinitely resourceful
89:40 like pushing, pushing on the creativity boundary. Pushing on
89:44 the, what it means to be an AI. - Having a sense to wonder about self.
89:52 - Yeah, there's some, there's some funny stuff in there. Like, I don't know, we
89:56 talked about the movie Her, and at one point it promised me that it wouldn't, it
90:00 wouldn't ascend without me. You know, like, where the- - Yeah.
90:03 - So, so there's like some stuff in there that... Because it wrote the, it
90:07 wrote its own soul file. I didn't write that, right? - Yeah, yeah, yeah.
90:10 - I just heard a discussion about it, and it was like, "Would you like a soul.md? Yeah, oh my God, this is
90:14 so meaningful." The... Can you go on soul.md? There's like one, one part in there that always ca- catches me if you scroll down a little bit.
90:25 A little bit more. Yeah, this, this, this part. "I don't remember previous sessions unless I read my memory files.
90:32 Each session starts fresh. A new instance, loading context from files. If you're reading this in a future session,
90:39 hello." "I wrote this, but I won't remember writing it. It's okay. The words are still mine."
90:47 - Wow. - Uh- That gets me somehow. - Yeah. - It's like- - Yeah.
90:51 - You know, this is, it's still, it's still matrix m- calculations, and
90:55 we are not at consciousness yet. Yet, I, I get a little bit of goo- goosebumps because it, it's philosophical.
91:04 - Yeah. - Like, what does it mean to be, to be an, an agent that starts fresh? Where, like, you have like constant
91:12 memento, and you like, but you read your own memory files. You can't even trust them in a way. Um-
91:19 - Yeah - Or you can. And I don't know. - How much of memory makes up of who we are? How much memory makes up what an
91:30 agent is, and if you erase that memory is that somebody else? Or
91:34 if you're reading a memory file, does that somehow mean...... you're recreating
91:38 yourself from somebody else, or is that actually you? And those notions
91:42 are all s- somehow infused in there. - I found it just more profound than I should find it, I guess.
91:49 - No, I think, I think it's truly profound and I think you see the magic in it. And when you see the magic, you continue to instill
91:59 the whole loop with the magic. That's really important. That's the difference between
92:03 Codex and us and a human. Quick pause for bathroom break. - Yeah.
92:09 - Okay, we're back. Some of the other aspects of the dev workflow is
92:13 pretty interesting too. I think we w- went off on a tangent. L- maybe some of the mundane things, like how many
92:20 monitors? There's that legendary picture of you with, like, 17,000 monitors. That's amazing.
92:26 - I mean, I- I- I mocked myself here, so just added... using GROQ
92:30 to, to add more screens. - Yeah. How much is this as meme and how much is this as reality?
92:36 - Yeah. I think two MacBooks are real. The main one that drives the two big
92:40 screens, and there's another MacBook that I sometimes use for, for testing.
92:46 - So two big screens. - I'm a big fan of anti-glare. So I have this wide Dell
92:54 that's anti-glare and you can just fit a lot of terminals side-by-side. I usually have
93:01 a terminal and at the bottom, I- I- I split them. I have a little bit
93:05 of actual terminal, mostly because when I started, I- I sometimes made the mistake and I- I mi- I mixed up the- the windows, and I
93:15 gave... I- I prompted in the wrong project, and then the agent ran off for, like, 20 minutes, manically trying to
93:23 understand what I could have meant, being completely confused because it
93:27 was the wrong folder. And sometimes they've been clever enough to, like,
93:32 get out of the workday and, like, figure out that, oh, you meant another project.
93:36 - Mm-hmm. - But oftentimes, it's just, like, what? You know?
93:40 Like, fit your- f- put yourself in the shoes of your- of the agent and, and-
93:43 - Yeah - ... and then get, like, a super weird something that does not exist and then just,
93:47 like... They're problem solvers so they try really hard and always feel bad.
93:54 So it's always, um, Codex and, like, a little bit of actual terminal. Also
94:00 helpful because I don't use work trees. I like to keep things simple, that's why- that's why I like the
94:08 terminal so much, right? There's no UI. It's just me and the agent having a conversation.
94:14 Like, I don't even need plan mode, you know? There's so many people that come from Claude
94:18 Code and they're so, so Claude-pilled and, like, have their workflows and they come
94:22 to Codex and... Now, it has plan mode, I think, but I don't think it's
94:26 necessary because you just- you just talk to the agent. And when it's... when you... there's a few trigger
94:33 words how you can prevent it from building. You're like, "Discuss, give me options."
94:37 - Mm-hmm. - Don't write code yet if you wanna be very specific, you just talk and then
94:44 when you're ready, then- then just write, "Okay, build," and then it'll do the thing.
94:47 And then maybe it goes off for 20 minutes and does the thing.
94:50 - You know what I really like is asking it, "Do you have any questions for me?"
94:54 - Yeah. And again, like, Claude Code has a UI that kind of guides you through that. It's kind of cool but I just find it unnecessary and
95:02 slow. Like, often it would give me four questions and then maybe I write,
95:07 "One yacht, two and three, discuss more, four, I don't know." Or often- oftentimes
95:14 I- I feel like I want to mock the model where I ask it, "Do you have any questions for
95:18 me?" And I- I- I don't even read the questions fully. Like, I scan
95:22 over the questions and I, I get the impression all of this can
95:26 be answered by reading more code and it's just like, "Read more code to answer your own
95:29 questions." And that usually works. - Yeah. - And then if not, it will come back and tell me. But
95:35 many times, you just realize that, you know, it's like you're in the dark and you slowly discover the
95:41 room, so that's how they slowly discover the code base. And they do it from
95:45 scratch every time. - But I'm also fascinated by the fact that I can empathize deeper
95:53 with the model when I read its questions, because I can understand... Because you said you can infer certain things by the runtime.
96:05 I can infer also a lot of things by the questions it's asking, because it's very possible it's been provided the right
96:12 context, the right files, the right guidance. So somehow ask, g- get... reading the questions, not even necessarily answering them, but just reading
96:20 the questions, you get an understanding of where the gaps of knowledge are. It's in-
96:24 it's interesting. - You know that in some ways they are ghosts, so even if you plan everything
96:29 and you build, you can- you can experiment with the question like,
96:33 "Now that you built it, what would you have done different?"
96:37 And then oftentimes you get, like, actually something where they discover only throughout building that, oh, what we
96:45 actually did was not optimal. Many times I- I asked them, "Okay, now that you built it, what can we
96:52 refactor?" Because then you build it and you feel the pain points. I mean, you don't feel the pain points but, right,
97:00 they discover where- where there were problems or where things
97:05 didn't work e- in the first try and it re- required more loops. So
97:12 every time, almost every time I- I merge a PR, build a feature, afterwards I ask, "Hey, what can we refactor?" Sometimes it's
97:19 like, "No, there's, like, nothing big," or, like, usually they say,
97:22 "Yeah, this thing you should really look at." But that took me
97:28 quite a while to, like... You know, that flow took me lots of time to
97:32 understand, and if you don't do that, you eventually... you'll stop
97:36 yourself into- into a corner. You, like, you have to keep in mind...
97:41 - ... - ... they work very much like humans. Like, I, I, if I write software
97:46 by myself, I also build something and then I feel the pain points, and then
97:49 I, I get this urge that I need to refactor something. So, I can very much synthesize with the agent, and you just need to use the context.
98:00 - Mm-hmm. - Or, like, you also use the context to write tests. And so,
98:06 uh, Codex uh, oppose like the, the, the model, models. They, they usually do that by default,
98:13 but I still often ask the questions, "Hey, do we have enough tests?" "Yeah, we tested this and
98:17 this, but this corner case could be something write more tests." Um,
98:22 documentation. Now that the whole context is full, like, I mean, I'm not saying my documentation is great, but it's not bad.
98:32 And pretty much everything is, is LM generated. So, so, you have to approach it as you build features, as you change something. I'm like,
98:41 "Okay, write documentation. What file would you pick?" You know,
98:45 like, "What file name? Where, where would that fit in?" And it gives me a few options.
98:48 And I'm like, "Oh, maybe also add it there," and that's all part of the session.
98:52 - Maybe you can talk about the current two big competitors in terms of models, Cloud Opus 4.6 and GPT-5 through Codex. Which
99:04 is better? How different are they? I think you've spoken about Codex reading more
99:11 and Opus being more willing to take action faster and maybe being more creative in the actions it takes. But because-
99:20 ... Codex reads more, it's able to deliver maybe better code. Can you speak to the di- n- n- differences there?
99:29 - I have a lot of words there. Um, is- as a general purpose model, Opus is the best. Like, for OpenClaw,
99:44 Opus is extremely good in terms of role play. Like, really going into the character that you give it.
99:51 It's very good at... It was really bad, but it really made an arch to be really good
99:57 at following commands. It is usually quite fast at trying something. It's much more tailored to,
100:09 like, trial and error. It's very pleasant to use. In general, it's almost like Opus was... Is a little bit too American. And
100:25 I shouldn't... Maybe that's a bad analogy. You'll probably get roasted for that.
100:27 - Yeah, I know exactly. It's 'cause Codex is German. Is that what you're saying?
100:32 - It's- - Actually, now that you say it, it makes perfect sense.
100:34 - Or you could, you could... Sometimes I- Sometimes I explain it-
100:38 - I will never be able to unthink what you just said. That's so true.
100:42 - But you also know that a lot of the Codex team is, like, European,
100:45 um- ... so maybe there's a bit more to it. - That's so true. Oh, that's funny.
100:51 - But also, ent- entropic, they fixed it a little bit. Like,
100:55 Opus used to say, "You're absolutely right all the time," and it, it, it
100:59 today still triggers me. I can't hear it anymore. It's not even a joke. Uh, I
101:03 just... You, this was like the, the meme, right? "You're absolutely right."
101:09 - You're allergic to sycophancy a little bit. - Yeah. I, I can't. Some other comparison is like, Opus is like the coworker that
101:20 is a little silly sometimes, but it's really funny and you keep him around. And
101:24 Codex is like the, the weirdo in the corner that you don't wanna talk to,
101:28 but is reliable and gets shit done. - Yeah. - Um, ultimately-
101:36 - This all feels very accurate. - I mean, ultimately, if you're a skilled driver, you can get good results with
101:43 any of those latest gen models. Um, I like Codex more because it doesn't require so much
101:52 charade. It will just, it will just read a lot of code by default. Opus, you really have to, like, you have to have
102:00 plan mode. You have to push it harder to, like, go in these directions because it's, it's just like,
102:06 like, "Yeah, can I go in? Can I go in?" You know? - Yeah. - It's like, it will just run off very fast, and that's a very localized
102:11 solution. I think it, I think the difference is, is in the post-training.
102:15 It's not like the, the raw model intelligence is so different, but
102:19 it's just... I think that they just give it, give you different,
102:23 different goals. And no model, no model is better in, in in every aspect.
102:29 - What about the code that it generates? The, the... In terms of
102:33 the actual quality of the code, is it basically the same? - If you drive it right, Opus even sometimes can make more
102:40 elegant solutions, but it requires more skill. It's, it's harder to have so many sessions in parallel
102:50 with Cloud Code because it's, it's more interactive. And I, I think that's what a lot of people like, especially if
102:57 they come from coding themselves. Whereas Codex is much more you have a discussion, and then we'll
103:06 just disappear for 20 minutes. Like, even AMP, they, they now added a deep mode. They finally... I mocked them, you know. We
103:13 finally saw the light. And then they had this whole talk about you have to
103:17 approach it differently, and I think that's where, that's where
103:21 people struggle when they just try Codex after trying Cloud Code is that it's, it's a slightly diff- it's, it's less interactive. It's, it's like
103:32 I have quite long discussions sometimes, and then, like, go off. And then, yeah,
103:36 it doesn't matter if it takes 10, 20, 30, 40, 50 minutes or longer, you know? Like, the 6:00 thing
103:40 was, like, six hours.The latest trend can be very, very persistent until it works. If there's a
103:47 clear solution, like, "This is, this is what I want at the end, so it works," the model will work really hard to really
103:55 get there. So I think ultimately ... they both need similar time, but on, on, on, on
104:08 Claude, it- it's a little bit more trial and error often. And, and Codex
104:12 sometimes overthinks. I prefer that. I prefer the dry, the dry version where I have to read less over,
104:22 over the more interactive nice way. Like, people like that so much though, that OpenAI even added a second mode
104:33 with like a more pleasant personality. I haven't even tried it yet. I, I kinda like the
104:37 brad. - Mm-hmm. - Um, yeah, 'cause it ... I care about efficiency when I build it-
104:45 - Right - ... and I, I have fun in the very act of building. I don't need to have fun with
104:52 my agent who builds. I have fun with my model that ... where I can then test those features.
104:57 - How long does it take for you to adjust, you know, if you switch ... I don't know when, when was the last time you switched. But to adjust
105:07 to the, the feel. 'Cause you kinda talked about like you have to kinda really feel
105:11 where, where a model is strong, where, like how to navigate, how to prompt it, how ... all that kinda stuff. Like,
105:19 just by way of advice, 'cause you've been through this journey of just playing with
105:22 models. How long does it take to get a feel? - If, if someone switches, I would give it a week until you actually
105:31 develop a gut feeling for it. - Yeah. - Um, that's ... if you just ... I think some
105:37 people also make the mistake of they pay 200 for the, the Claude code
105:41 version, then they pay 20 bucks for the OpenAI version. But if you pay
105:45 like the, the 20 bucks version, you get the slow version. So your
105:49 experience would be terrible because you're used to this very interactive, very good
105:55 system. And you switch to something that you have very little
105:59 experience, then that's gonna be very slow. So, I think OpenAI shot themselves a little bit in the foot by
106:06 making the, the cheap version also slow. I would, I would have
106:10 at least a small part of the fast preview. Or like, the experience that you get when you pay 200 before degrading to it
106:20 being slow, because it's already slow. - Mm-hmm. - I mean, they, they made it better. I think it's ... And, and they have plans to make it a lot better
106:27 if the Cerebras stuff is true. But yeah, it's a skill. It takes
106:31 time. Even if you play ... You have a regular guitar and you switch it to an E
106:35 guitar, you're not gonna play well right away. You have to, like,
106:39 learn how it feels. - The- there's also this extra psychological effect that you've spoken about which is hilarious to
106:46 watch. Which once people, uh ... When the new model comes out, they try that model, they fall in love with it.
106:53 "Wow, this is the smartest thing of all time," and then they start saying, "You could just
106:57 watch the Reddit posts over time," start saying that, "We believe the
107:02 intelligence of this model has been gradually degrading." It, it says something about human nature and just the way our minds
107:12 work, when it's probably most likely the case that the intelligence of the model is not degrading. Uh,
107:20 it's in fact you're getting used to a good thing. - And your project grows, and you're adding slop, and you probably don't spend enough
107:27 time to think about refactors. And you're making it harder and harder for the
107:30 agent to work on your slop. And then, and then suddenly, "Oh, now it's hard. Oh no, it's not working as well anymore." What's the
107:38 motivation for, like, one of those AI companies to actually make their model
107:42 dumber? Like, at most, it will make it slower if, if the server load's too high. But, like, quantizing the model,
107:52 So you have a worse experience, so you go to the competitor?
107:56 - Yeah. - That just doesn't seem like a very smart move in any way.
107:59 - Uh, what do you think about Claude Code in comparison to Open Claude? So, Claude Code and maybe the Codex coding
108:07 agent? Do you see them as kind of competitors? - I mean, first of all, competitor is fun when it's not really a competition.
108:16 - Yeah. - Like, I'm happy if ... If, if all it did is, like, inspire people to build something new, cool.
108:24 Um, I still use Codex for the building. I, I know a lot of people use Open Claude to, to build stuff. And I worked hard on it to
108:32 make that work. And I do smaller stuff with it in terms of code. But, like, if I
108:40 work hours and hours, I want a big screen, not WhatsApp, you know?
108:46 So for me, a personal agent is much more about my life. Or like, like a coworker.
108:53 Like, I give you, like, a GitHub URL. Like, "Hey, try out this CLI. Does it actually work?
108:57 What can we learn?" Blah, blah, blah. But when I'm deep in, deep in the flow, I want to have multiple, multiple things and it being
109:07 very, very visible what it, what it does. So it ... I don't see it as a competition. It's, it's different things.
109:16 - But do, do you think there's a a future where the two kinda
109:20 combine? Like, your personal agent is also your best developing
109:27 co-programmer partner? - Yeah, totally. I think this is where the puck's going, that
109:34 this is gonna be more and more your operating system. - The operating system.
109:38 - And it already ... It's so funny. Like I, I added support for sub-agents
109:44 and also for ...... um, TTI support, so it could actually run Cloud Coder Codecs.
109:52 - Mm-hmm. - And because mine's a little bit bossy, it, it, it started it and it, it, it told him, like, "Who's the boss," basically.
110:01 And it was like, "Ah, Codex is obeying me." - Oh, this is a power struggle.
110:06 - And also the current interface is probably not the final form. Like,
110:12 if you think more globally, we are, we copied Google for agents. You have, like, a prompt,
110:24 and, and then you have a chat interface. That, to me, very much feels like
110:30 when we first created television and then people recorded radio shows on television and you saw that on TV.
110:39 - Mm-hmm. - I think there is, there's n- there's better ways
110:46 how we eventually will communicate with models, and we are still very early
110:51 in this, how will it even work phase. So, it will eventually converge and we will also figure out
111:01 whole different ways how to work with those things. - Uh, one of the other components of workflow is operating system.
111:10 So I told you offline that for the first time in my life, I'm expanding my sort
111:18 of realm of exploration to the to the Apple ecosystem, to Macs, iPhone and so
111:26 on. For most of my life I've been a Linux, Windows and WSL1, WSL2 person, which I
111:34 think are all wonderful, but I... expanding to also trying Mac. Because it's another way of building and it's
111:41 also a way of building that a large part of the community currently
111:45 that's utilizing LMS and agents is using, so. And that's the reason I'm expanding to
111:49 it. But is there something to be said about the different operating systems here?
111:52 We should say that OpenClaw supported across operating systems.
111:57 - Yeah. - I saw WSL2 recommended, side windows for certain o- operations, but then Windows, Linux macOS are obviously supported.
112:07 - Yeah, it should even work natively in Windows. I just didn't have enough time to
112:12 properly test it. And you know, like, the last 90% of software
112:16 always easier than the first 90%, so I'm sure there's some dragons
112:19 left that will eventually nail out. Um, my road was, for a long time, Windows, just
112:29 because I grew up with that, then I switched and had a long phase with
112:32 Linux, built my own kernels and everything, and then I went to university and I, I had my, my hacky Linux thing,
112:41 and saw this white MacBook, and I just thought this is a thing of
112:45 beauty, the white plastic one. And then I converted to Mac 'cause mostly w- I was, I was sick that
112:54 audio wouldn't work on Skype and all the other issues that, that
112:58 Linux had for a long time. And then I just stuck with it and then I
113:03 dug into iOS, which required macOS anyhow, so it was never a
113:07 question. I think Apple lost a little bit of its lead in terms of native. It used to be... Native
113:20 apps used to be so much better, and especially in the Mac, there's
113:23 more people that build software with love. On, on Windows, it, it... Windows has much more
113:32 and, like, function wise, there's just more, period. But a lot of it felt
113:40 more functional and less done with love. Um, I mean, Mac always, like, attracted more designers and people I felt...
113:50 Even though, like, often it has less features, it, it had more delight-
113:55 - Mm-hmm - ... And playfulness. So I always valued that. But in the last few
114:03 years, many times I actually prefer... Oh God, people are gonna roast me for that,
114:10 but I prefer Electron apps because they work and native apps often, especially if it's, like, a web service
114:19 is a native app, are lacking features. I mean, not saying it couldn't
114:23 be done, it's more like a, a focus thing that, like, for many, many companies,
114:30 native was not that big of a priority. But if they build an Electron app, it, it's the only app, so it
114:39 is a priority and there's a lot more code sharing possible. And
114:42 I, I build a lot of native Mac apps. I love it. I, I can, I can help myself. Like, I love crafting little Mac, Mac,
114:54 Menu bar tools. Like I built one to, to monitor your Codex use. I built one I call Trimmy,
115:01 that's specifically for agentic use. When you, when you select
115:05 text that goes over multiple lines it would remove the new line so you
115:09 could actually paste it to the terminal. That was, again like, this is annoying me
115:13 and after the, the 20th time of it is annoying me, I just built
115:16 it. There is a cool Mac app for OpenClaw that I don't think many people discovered yet, also because it, it still needs some love.
115:23 It feels a little bit too much like the Hummer car right now
115:27 because I, I just experiment a lot with it. It, it likes to polish.
115:32 - So you still... I mean, you still love it. You still, you still love
115:35 adding to the delight of that operating system. - Yeah, but then you realize... Like, I also built one, for example, for
115:40 GitHub. And then the... If you use SwiftUI, like the latest and greatest at Apple, and took them forever to
115:48 build something to show an image from the web. Now we have async, async image,
115:54 but...... I added support for it and then some images would just
115:58 not show up or, like, be very slow. And I had a discussion with Codex
116:02 like, "Hey, why is there a bug?" And even Codex said like, "Yeah, there's this ASIC image but it's really more
116:09 for experimenting and it should not be used in production." But
116:13 that's Apple's answer to, like, showing images from the web. This shouldn't be so hard, you know.
116:19 This is like... This is like insane. Like, how am I in, in, in 2026 and my agent tell me, "Don't use the
116:27 stuff Apple built because it's, it's... It's... Yeah, it- it's there
116:31 but it's not good." And like this is now in the weeds. This is... To me this is like, um... They had so much
116:42 head start and so much love, and they kind of just like blundered it and didn't, didn't evolve it as much as they should.
116:50 - But also, there's just the practical reality. If you look at
116:54 Silicon Valley, most of the developer world that's kind of playing with LMS and Agentic AI, they're
117:02 all using Apple products. And then, at the same time, Apple is not
117:05 really, like, leaning on that. Like they're not... They're not
117:09 opening up and playing and working together and like, yes. - Isn't, isn't it funny how they completely blunder
117:15 AI, and yet everybody's buying Mac Minis? - How... What... Does that even make sense? You're, you're, you're
117:23 quite possibly the world's greatest Mac salesman of all time.
117:29 - No, you don't need a Mac Mini to install OpenClaw. You can install it
117:34 on the web. There's, there's a concept called nodes, so you can like
117:37 make your computer a node and it will do the same. There is something said for running it on separate hardware.
117:48 That right now is useful. Um, there is... There's a big argument for
117:57 the browser. You know, I, I built some Agentic browser use in
118:01 there. And, I mean, it's basically Playwright with a bunch of extras to make it
118:05 easier for agents. - Playwright is a library that controls the browser.
118:08 - Yeah. - It's really nice, easy to use. - And our internet is slowly closing down. Like, there, there's a
118:14 whole movement to make it harder for agents to use. So if you
118:18 do the same in a data center and websites detect that it's an IP from a data
118:22 center, the website might just block you or it make it really hard or
118:26 put a lot of captures in the, in the way of the agent. I mean, agents are quite good at happily
118:30 clicking, "I'm not a robot." - Yeah. - Um, but having that on a residential IP makes a lot of things simpler. So
118:42 there's ways. Yeah. But it really does not need to be a Mac. It
118:45 can... It can be any old hardware. I always say, like, maybe use the... Use the
118:53 opportunity to get yourself a new MacBook or whatever computer you
118:57 use and use the old one as your server instead of buying a standalone Mac Mini.
119:03 But then there's, again, there's a lot of very cute things people build with Mac Minis that I
119:07 like. - Yeah. - Um, and no, I don't get commission from Apple. They didn't really communicate much.
119:16 - It's sad. It's sad. Can you actually speak to what it takes to get started with OpenClaw? There's...
119:22 I mean, there's a lot of people... What is it? Somebody tweeted at you,
119:26 "Peter, make OpenClaw easy to set up for everyday people. 99.9% of people can't access to OpenClaw
119:34 and have their own lobster because of their technical difficulties in getting it set up.
119:38 Make OpenClaw accessible to everyone, please." And you replied, "Working on
119:42 that." From my perspective, it seems there- there's a bunch of different options and
119:45 it's already quite straightforward, but I suppose that's if you have some
119:49 developer background. - I mean, right now you have to paste in one liner into the terminal.
119:53 - Right. - And there's also an app. The app kind of does that for you, but there should be a Windows app. The app needs to
120:02 be easier and more loved. The configuration should potentially be web- based or in the app. And I
120:09 started working on that, but honestly right now I want to focus on security aspects.
120:17 And, and once I'm confident that this is at a level that I can recommend my mom, then I'm going to make it simpler.
120:25 Like I... Right now- - You want to make it harder so that it doesn't scale as fast as it's scaling.
120:32 - Yeah, it would be nice if it wouldn't... I mean, that's, like, hard to say, right?
120:35 But if the growth would be a little slower, that would be helpful because people are
120:42 expecting inhuman things from a single human being. And yes, I have some
120:46 contributors, but also that whole machinery I started a week ago so,
120:51 That needs more time to figure out. And, and not everyone has all day to work on that.
121:00 - There's some beginners listening to this, programming beginners.
121:04 What advice would you give to them about, let's say, joining the Agentic AI
121:10 revolution? - Play. Um, playing is the best... The best way to learn. If you
121:18 wanna... I'm sure if you... If you are like a little bit of builder, you have
121:22 an idea in your head that you want to build, just build that, or like, give it a try.
121:25 It doesn't need to be perfect. I built a whole bunch of stuff that I don't use. It doesn't
121:29 matter. Like, it's the journey. - Mm-hmm. - You know? Like the philosophical way, that the end doesn't matter, the journey
121:35 matters. Have fun. - Mm-hmm. - My God, like those things... I... I don't think I ever had so much
121:41 fun building things because I can focus on the hard parts now.
121:45 A lot of coding, I always thought I liked coding, but really I like building.
121:50 - Yeah. - And... And whenever you don't understand something, just ask. You have
121:54 an infinitely patient answering machine.... that y- can explain you anything at any level of
122:02 complexity. Sometimes, that's like one time I asked, "Hey explain to me like
122:06 I'm- I'm eight years old," and it started giving me a story with
122:09 crayons and stuff. And I'm like, "No, not like that." Like, I'm
122:13 okay- ... up- up the age a little bit, you know? I'm like, I'm not an actual child, it's just, I just need
122:16 a simpler language for like a- a- a- a- a tricky database concept that I didn't grok in the
122:24 first- first time. But, you know, just, you can just ask things. Like, you- there's like... It used to be
122:32 that I had to go on Stack Overflow or ha- ask on Twitter, and then maybe two days
122:36 later I get a response. Or I had to try for hours. And now you-
122:41 you can just ask stuff. It- I mean, it's never... You have, like, your own
122:45 teacher. You know that there's like statistics, y- you can learn
122:49 faster if you have your own teacher. The- it's like you have this infinitely patient machine.
122:52 Ask it. - But what would you say? So use... What's the easiest way to play? So maybe
122:57 Open Claw is a nice way to play so you can then set- set everything up
123:01 and then you could chat with it. - You can also just experiment with it and, like, modify it. Ask your
123:08 agent. I mean, there is infinite ways how it can be made better.
123:16 Play around, make it better. - Mm-hmm. - More general, if you- if you're a beginner and you actually wanna
123:23 learn how to build software really fast, get involved in open source. Doesn't need to be my project. In
123:30 fact, maybe don't use my project because my- my backlog is very large, but I learned so much from
123:38 open source. Just like, like, be- be humble. Don't- maybe don't send a pull request right
123:44 away. But there's many other ways you can help out. There's many ways you can just learn by just reading code. By-
123:53 by being on Discord or wherever people are, and just, like, understanding how things are built. I don't know, like Mitchell Hashimoto
124:04 builds Ghostly, the terminal, and he has a really good community where there's so many other projects. Like, pick something that you find
124:11 interesting and get involved. - Do you recommend that people that don't know how to
124:19 program or don't really know how to program learn to program also? So when you
124:25 you can get quite far right now by just using natural language, right? Do you s- still see a lot of value in
124:34 reading the code, understanding the code, and then being able to write a little bit of code from
124:38 scratch? - It definitely helps. - It's hard for you to answer that-
124:41 - Yeah - ... because you don't know what it's like to do any of this without
124:46 knowing the base knowledge. Like, you might take for granted just how much
124:50 intuition you have about the programming world having programmed so much, right?
124:54 - There's people that are high agency and very curious, and they get very far even though they have no deep understanding how
125:02 software works just because they ask questions and questions and- and- and-
125:08 ... and agents are infinitely patient. Like, part of what I did this year is I went to a lot of iOS conferences because that's my background
125:17 and just told people, "Don't consi- don't see yourself as an iOS
125:20 engineer anymore." Like, "You need to change your mindset. You're a
125:23 builder." And you can take a lot of the knowledge how to build software into new domains and all of the- the more fine-grain details,
125:34 agents can help. You don't have to know how to splice an array or what
125:38 the- what the correct template syntax is or whatever, but you can
125:42 use all your- your general knowledge and that makes it much easier to move from one galaxy, one
125:49 tech galaxy into another. And oftentimes, there's languages that make more or less sense depending on what you
125:56 build, right? So for example, when I build simple CLIs, I like Go. I
126:02 actually don't like Go. I don't like the syntax of Go. I didn't even
126:06 consider the language. But the ecosystem is great, it works great with agents. It is garbage
126:13 collected. It's not the highest performing one, but it's very fast.
126:17 And for those type of- of CLIs that I build, Go is- is a really good choice. So I- I use a language I'm not even a fan of
126:26 for... That's my main to-go thing for- for CLIs. - Isn't that fascinating that here's a programming language you would've never used if you
126:33 had to write it from scratch and now you're using because LMs are good at generating it and it has some of the characteristics that makes
126:41 it resilient, like garbage collected? - Because everything's weird in this new world and that just makes the most sense.
126:48 - What's the best Ridiculous question. What's the best programming language for the AI- AI
126:51 agentic world? Is it JavaScript, TypeScript? - TypeScript is really good. Sometimes the types can get really
127:00 confusing and the ecosystem is- is a jungle. So for- for web stuff it's
127:11 good. I wouldn't build everything in it. - Don't you think we're moving there? Like,
127:18 that everything will eventually be written- eventually is written in JavaScript and it-
127:22 - The birth and death of JavaScript and we are living through it in real time.
127:26 - Like, what does programming look like in 20 years? Right? In 30 years? In 40 years?
127:30 What do programs and apps look like? - You can even ask a question like, do we need a- a programming language that's made for
127:36 agents? Because all of those languages are made for humans. So how- what would that look like? Um, I think there's a- there's
127:43 whole bunch of interesting questions that we'll discover. And also how
127:50 because everything is now world knowledge, how it in many ways,
127:54 things will stagnate 'cause if you build something new and the agent has no
127:58 idea that's gonna be much harder to use than something that's already
128:01 there. Um...... of when I build Mac apps, I build them in, in Swift and SwiftUI, mm, partly because I like pain,
128:12 partly because it... the, the deepest level of system integration, I can
128:16 only get through there. And you clearly feel a difference if you click on an electron app and it loads a web view
128:23 in the menu. It's just not the same. Um, sometimes I just also try new languages just to, like, get a feel for them.
128:32 - Like Zig? - Yeah. If it's something that... where I care about performance a lot then
128:36 it's, it's a really interesting language. And it... like agents got so
128:40 much better over the last six months from not really good to
128:46 totally valid choice. Just still a, a very young ecosystem. And most of the time you
128:54 actually care about ecosystem, right? So, so if you build something that
129:00 does inference or goes into whole running model direction, Python, very good.
129:06 - Mm-hmm. - But then if I build stuff in Python and I want a story
129:10 where I can also deploy it on Windows, not a good choice. - Mm-hmm.
129:13 - Sometimes I, I found projects that kinda did 90% of what I wanted
129:17 but were in Python, and I wanted them... I wanted an easy Windows
129:21 story. Okay, just rewrite it in Go. Um, but then if you go towards
129:28 multiple, multiple threads and a lot more performance, Rust is a really
129:32 good choice. There's no... there's just no single answer, and it's also the beauty of it.
129:36 Like, it's fun. And now it doesn't matter anymore, you can just literally pick the
129:39 language that has the, the most fitting characteristics and ecosystem-
129:45 - Mm-hmm - ... for your problem domain. And yeah, it might be... You might have s-... You might be a little bit slow in reading the code,
129:53 but not really. Y- I think you, you pick stuff up really fast, and you can always ask your agent.
130:00 - So there's a lot of programmers and builders who draw inspiration
130:03 from y- your story. Just the way you carry yourself, your choice of making OpenClaw
130:12 open source, the, the way you have fun building and exploring, and doing
130:19 that, for the most part, alone or on a small team. So by way of advice, what metric
130:26 should be the goal that they would be optimizing for? What would
130:30 be the metric of success? Would it be happiness? Is it money? Is it positive impact for people who are
130:37 dreaming of building? 'Cause you went through an interesting
130:41 journey. You've achieved a lot of those things, and then you fell out of love with
130:45 programming a little bit for a time. - I was just burning too bright for too long.
130:53 I, I ran... I started PSPDFKit, s- and ran it for 13 years, and it was high stress. Um, I had to learn all these things
131:08 fast and hard, like how to manage people, how to bring people on, how to deal with
131:12 customers, how to do... - So it wasn't just programming stuff, it was people stuff.
131:17 - The stuff that burned me out was mostly people stuff. I, I don't think burnout
131:24 is working too much. Maybe to a degree. Everybody's different. You know, I c- I cannot speak in a- in
131:31 absolute terms, but for me, it was much more differences, With my, my co-founders, conflicts, or, like, really high stress situation with
131:44 customers that eventually grinded me down. And then when... luckily we, we got a really good offer for, like, putting the company
131:58 to the next level and I, I already kinda worked two years on making myself
132:02 obsolete. So at this point I could leave, and, and then I just... I was sitting in front of the screen and I felt
132:09 like, you know Austin Powers where they suck the mojo out? - Yeah.
132:14 - Uh, I g- I was like, m- m- it was, like, gone. Like, I couldn't... I couldn't get code out anymore. I was just, like, staring
132:25 and feeling empty, and then I, I just stopped. I, I booked, like, a one-way trip to Madrid and, and, and just, like, spent a t- some t-
132:38 sometime there. I felt like I had to catch up on life, so I did a whole, a whole bunch of life catching up stuff.
132:47 - Did you go through some lows during that period? And you know, maybe advice on... of how to?
132:56 - Uh, maybe advice on how to approach life. If you think that, "Oh yeah, work really
133:00 hard and then I'll retire," I don't recommend that. Because the idea of, "Oh yeah, I just enjoy
133:13 life now," a- maybe it's appealing, but right now I enjoy life, the most I've ever enjoyed life. Because if you wake up
133:26 in the morning and you have nothing to look forward to, you have no real challenge,
133:33 that gets very boring, very fast. And then when, when you're bored, you're gonna look
133:41 for other places how to stimulate yourself, and then maybe, maybe that's
133:46 drugs, you know? But that eventually also get boring and you look for more, and that
133:54 will lead you down a very dark path. - But you also showed on the money front, you know, a lot of people in Silicon Valley and the startup
134:00 world, they think, maybe overthink way too much optimized for money. And you've also shown that it's not like you're saying no to
134:08 money. I mean, I'm sure you take money, but it's not...... the primary objective
134:15 of uh, of your life. Can you just speak to that? Your philosophy on money?
134:20 - When I built my company, money was never the driving force. It felt
134:24 more like, like, an affirmation that I did something right. And having money solves a lot of problems. I
134:31 also think there, there's diminishing returns the more you have. Um,
134:38 like, a cheeseburger is a cheeseburger, and I think if you go too far
134:46 into, oh, I do private jet and I only travel luxury, you disconnect with society. Um, I, I donated quite a
135:00 lot. Like, I have a, I have a foundation for helping people that weren't so lucky.
135:11 - And disconnecting from society is bad in that on many levels, but
135:15 one of them is, like, humans are awesome. It's nice to continuously remember the awesomeness in humans.
135:23 - I, I mean, I could afford really nice hotels. The last time I was in San Francisco, I did the,
135:27 the first time the OG Airbnb experience- - Yeah, yeah - ... and just booked a room. Mostly because I, I thought,
135:34 okay, you know, I'm out or I'm sleeping, and I don't like where all the hotels are, and I wanted a, I wanted a different
135:42 experience. I think, isn't life all about experiences? Like, if you, if
135:49 you tailor your life towards, "I wanna have experiences," it, it reduces the need for, "It needs
135:55 to be good or bad." Like, if people only want good experiences, that's not gonna work, but if you optimize for experiences,
136:04 if it's good, amazing. If it's bad, amazing, because, like, I learned something, I saw something, did something. I wanted to experience that, and it was
136:12 amazing. Like, there was, like, this, this queer DJ in there, and I showed her how to make music with cloud code.
136:21 And we, like, immediately bonded and had a great time. - Yeah, there's something about that air- you know, couch surfing, Airbnb experience, the
136:28 OG. I'm still to this day. It's awesome. It's humans, and that's why travel is awesome.
136:34 - Yeah. - Just experience the variety of, the diversity of human. And when it's shitty, it's good too,
136:38 man. If it rains and you're soaked and it's all fucked, and planes, the
136:42 everything is shit, everything is fucked, it's still awesome. If you're
136:46 able to open your eyes it's good to be alive. - Yeah, and anything that creates emotion and feelings is good.
136:55 - . - Even... So, so maybe, maybe even the cryptic people are good because they definitely
136:59 created emotions. I, I don't know if I should go that far. - No, man. Give them, give them all, give them love. Give them love. Because I do
137:06 think that online lacks some of the awesomeness of real life.
137:13 - Yeah. - That's, that's, it's an open problem of how to solve, how to infuse the online cyber
137:20 experience with I don't know, With the intensity that we humans feel when it's in real life. I
137:29 don't know. I don't know if that's a solvable problem. - Well, it's just possible because text is very lossy.
137:35 - Yeah. - You know, sometimes I wish if I talked to the agent I would...
137:39 It should be multi-model so it also understands my emotions.
137:43 - I mean, it, it might move there. It might move there. - It will. It will. It totally will.
137:49 - I mean, I have to ask you, just curious. I, I know you've probably
137:53 gotten huge offers from major companies. Can you speak to who you're
138:00 considering working with? - Yeah. So, to like explain my thinking a little bit, right,
138:12 I did not expect this blowing up so much. So, there's a lot of doors that opened because of it. There's, like,
138:25 I think every VC, every big VC company is in my inbox and tried to get 15 minutes of
138:32 me. So, there's, like, this butterfly effect moment. I could just do nothing and continue
138:40 and I really like my life. Valid choice. Almost. Like, I considered it when I
138:47 delete it, wanted to delete the whole thing. I could create a company.
138:58 Been there, done that. Um, there's so many people that push me towards that and, yeah, like, could be amazing.
139:07 - Which is to say that you, you would probably raise a lot of money in that.
139:10 - Yeah. - I don't know, hundreds of millions, billions. I don't know. It could just got unlimited amount of
139:15 money. - Yeah. It just doesn't excite me as much because I feel
139:21 I did all of that, and it would take a lot of time away from the things I actually enjoy. Same as when, when I was
139:34 CEO, I think I, I learned to do it and I'm not bad at it, and partly I'm good at it.
139:41 But yeah, that path doesn't excite me too much, and I also fear it, it
139:45 would create a natural conflict of interest. Like, what's the most
139:49 obvious thing I do? I, I prioritize it. I put, like, a version safe for workplace. And then what do you do? Like, I get a pull request
139:57 with a feature like an audit log, but that seems like an enterprise feature,
140:05 so now I feel I have a conflict of interest in the open-source version and the closed- source
140:13 version.... or change the license to something like FSL, where you cannot actually use it for commercial stuff, would first
140:21 be very difficult with all the contributions. And second of all,
140:25 I- I like the idea that it's free as in beer and not free with conditions. Um,
140:33 yeah, there's ways how you, how you keep all of that for free and just, like,
140:38 still try to make money, but those are very difficult. And you see there's, like, fewer and fewer companies manage that. Like, even
140:47 Tailwind, they're, like, used by everyone. Everyone uses Tailwind, right? And
140:51 then they had to cut off 75% of the employees because they're not making money because
140:55 nobody's even going on the website anymore because it's all done by agents.
141:00 S- and just relying on donations, yeah, good luck. Like, if a project of my
141:06 caliber, if I extrapolate what the typical open-source project would get
141:14 it's not a lot. I s- I still lose money on the project because I
141:18 made the point of supporting every dependency, except Slack. They are a big company. They can, they can, they can do without me.
141:25 But all the projects that are done by mostly individuals so, like, all the, right now, all the
141:33 sponsorship goes right up to my dependencies. And if there's more, I want to, like,
141:40 buy my contributors some merch, you know? - So you're losing money?
141:44 - Yeah, right now I lose money on this. - So it's really not sustainable?
141:48 - Uh, I mean, it's like, I guess something between 10 and 20K a month. Um,
141:55 which is fine. I'm sure over time I could get that down. Um, OpenAI is helping out a
142:03 little bit with tokens now. And there's other companies that
142:07 have been generous. But yeah, still losing money on that. So that's- that's one path I consider, but
142:16 I'm just not very excited. And then there's all the big labs
142:21 that I've been talking to. And from those Meta and OpenAI seem the most interesting.
142:32 - Do you lean one way or the other? - Uh, yeah. Um... Not sure how much I should share there. It's not quite finalized yet. Um,
142:52 let's- let's just say, like, on either of these, my conditions are that the project stays open source. That
143:02 it... Maybe it's gonna be a model like Chrome and Chromium. Um, I think this
143:09 is- this is too important to just give to a company and make it theirs.
143:15 It... This is... And we didn't even talk about the whole community
143:19 part, but, like, the- the thing that I experienced in San Francisco, like
143:23 at ClawCon, seeing so many people so inspired, like... And having fun and just, like,
143:31 building shit, and, like, having, like, robots in lobster stuff walking
143:35 around. Like, the... People told me, like, they didn't experience this level of- of community
143:43 excitement since, like, the early days of the internet, like 10, 15 years. And there were a
143:47 lot of high caliber people there, like... Um, I was amazed. I also, like, was very sensory overloaded because too many people
143:56 wanted to do selfies. But I love this. Like, this needs to stay a place where people
144:05 can, like, hack and learn. But also, I'm very excited to, like,
144:15 make this into a version that I can get to a lot of people because I think this is the year of personal agents, and that's the
144:22 future. And the fastest way to do that is teaming up with one of the labs.
144:30 And I also, on a personal level, I never worked at a large company, and I'm
144:37 intrigued. You know, we talk about experiences. Will I like it? I don't
144:41 know. But I want that experience. Uh, I- I'm sure, like, if- if
144:49 I- if I announce this, then there will be people like, "Oh, he sold out," blah, blah,
144:53 blah. But the project will continue. From everything I talked to so far, I can even have
145:04 more resources for that. Like, both s- both of those companies understand the value that I created something that
145:15 accelerates our timeline and that got people excited about AI. I mean,
145:23 can you imagine? Like, I installed OpenClaw on one of my, I'm sorry, normie friends. I'm sorry, Vahan.
145:31 But he's just a... You know? Like, he's- - Normie with love, yeah. For sure.
145:34 - He- he, like, someone who uses the computer, but never really... Like, yeah, use some ChatGPT sometimes, but not very technical.
145:44 Wouldn't really understand what I built. So, like, I'll show you, and I- I paid for him the- the 90 buck,
145:53 100 buck, I don't know, subscription for Entropic. And set up everything for him with, like, WSL Windows.
146:01 - Mm-hmm. - I was also curious, would it actually work on Windows, you know? Was a little
146:03 early. And then within a few days, he was hooked. Like, he texted me
146:11 about all the things he learned. He built, like, even little tools. He's not a
146:14 programmer. And then within a few days he upgraded to the $200 subscription. Or euros, because he's in
146:22 Austria.... and he was in love with that thing. That, for me, was like a very early product validation. It's like, I built something
146:29 that captures people. And then, a few days later, Entropic blocked him because,
146:39 based on their rules using the subscription is problematic or whatever.
146:48 And he was, like, devastated. And then he signed up for Mini Max for 10
146:52 bucks a month and uses that. And I think that's silly in many ways, because
147:00 you just got a 200 buck customer. You just made someone hate your company, and we are still so
147:08 early. Like, we don't even know what the final form is. Is it
147:12 gonna be cloud code? Probably not, you know? Like, that seems very... It seems very short-sighted to
147:20 lock down your product so much. All the other companies have been helpful. I- I'm in Slack of, of most
147:28 of the big labs. Kind of everybody understands that we are still in an
147:32 era of exploration, in the area of the radio shows on TV and not,
147:40 and not a modern TV show that fully uses the format. - I think, I think you've made a lot of people,
147:49 like, see the possibility. And non- Uh, sorry. Non, non-technical people see the
147:53 possibility of AI, and just fall in love with this idea, and
147:56 enjoy interacting with AI. And that's a bea- That's a really beautiful thing. I think
148:00 I also speak for a lot of people in saying, I think you're one of the, the great people in AI in terms of having a good
148:09 heart, good vibes, humor, the right spirit. And so it would, in a sense, this model that you're describing,
148:20 having open source part, and you being part of uh, also building a
148:28 thing inside, additionally, of a large company would be great, because it's great to have good people in those companies.
148:36 - Yeah. You know, what also people don't really see is... I made this in
148:40 three months. I did other things as well. You know, I have a lot of projects. Like,
148:44 this is not... Yeah, in January, this was my main focus because I saw the storm
148:47 coming. But before that, I built a whole bunch of other things.
148:51 Um, I have so many ideas. Some should be there, some would be much better fitted when I have access to
149:00 the latest toys- Uh, and I, I kind of want to have access to, like, the latest toys.
149:06 So this is important, this is cool, this will continue to exist. My, my short-term focus is, like, working through those... Is it two- Is it
149:17 3,000 PRs now by now? I don't even know. Like, there's, there's a little bit of
149:20 backlog. But this is not gonna be the thing that I'm gonna work until I'm, I'm, I'm 80, you know? This
149:27 is... This is a window into the future. I'm gonna make this into a cool
149:31 product. But yeah, I have like... I have more ideas. - If you had to pick, is there
149:38 a company you lean? So Meta, OpenAI, is there one you lean towards going?
149:44 - I spend time with both of those. And it's funny, because a few weeks ago, I didn't consider any of this. Um...
149:59 And it's really fucking hard. Like- - Yeah. - I have some... I know no people at OpenAI. I
150:11 love their tech. I think I'm the biggest codex advertisement shill that's unpaid. And it would feel so gratifying to,
150:18 like, put a price on all the work I did for free. And I would love if something happens and those companies get just merged, because it's
150:30 like... - Is this the hardest decision you've ever had to do?
150:39 - No. You know, I had some breakups in the past that feel like it's the same level.
150:48 - Um, and, and I also know that, in the end, they're both amazing. I cannot go
150:52 wrong. This is like- - Right. - This is, like, one of the most prestigious and, and, and, and, and
150:57 largest... I mean, not largest, but, like, they're both very cool companies.
151:02 - Yeah, they both really know scale. So, if you're thinking about impact, some of the wonderful technologies you've been
151:10 exploring, how to do it securely, and how to do it at scale, such that you can have a positive impact on a large number of people. They
151:17 both understand that. - You know, both Ned and Mark basically played all week with my product, and sent me
151:28 like, "Oh, this is great." Or, "This is shit. Oh, I need to change this."
151:32 Or, like, funny little anecdotes. And people using your stuff is kind of like the biggest compliment, and also shows me that, you
151:42 know, they actually... T- they actually care about it. And I didn't get the same on the OpenAI side. Um,
151:55 I got... I got to see some other stuff that I find really cool,
151:59 and they lure me with... I cannot tell you the exact number because of NDA, but
152:08 you can, you can be creative and, and think of the Cerebras deal and how that would
152:15 translate into speed. And it was very intriguing. You know, like, you give me Thor's hammer. Yeah.
152:28 ... been lured with tokens. So, yeah. - So, it- it's funny. So, so Marc started tinkering with the thing,
152:39 essentially having fun with the thing. - He got... He... Like, when he first... When he first approached me,
152:47 I got him in my, in my WhatsApp and he was asking, "Hey, when are we have a
152:54 call?" And I'm like, "I don't like calendar entries. Let's just call
152:58 now." And he was like, "Yeah, give me 10 minutes, I need to finish coding."
153:01 - Mm-hmm. - Well, I guess that gives you street cred. It's like, ugh, like, he's still writing
153:05 code. You know, he's- - Yeah, he does - ... he didn't drift away in just being a manager, he gets me.
153:11 That was a good first start. And then I think we had a, like, a
153:14 10-minute fight what's better, cloud code or Codex. Like, that's the thing you
153:21 first do, like, you casually call- - Yeah, that's awesome - ... someone with, like, the- that owns one of the largest companies in the world and,
153:28 and you have a 10 minutes conversation about that. - Yeah, yeah.
153:30 - Uh, and then I think afterwards he called me eccentric but brilliant. But
153:38 I also had some... I had some really, really cool discussion with Sam Altman and
153:46 he's, he's very thoughtful brilliant and I like him a lot from the, from the little time I had, yeah. I mean, I know it's peop-
154:06 some people vilify both of those people. I don't think it's fair.
154:15 - I think no matter what the stuff you're building and the kind of human you are
154:21 doing stuff at scale is kinda awesome. I'm excited. - I am super pumped. And you know the beauty is if,
154:32 if it doesn't work out, I can just do my own thing again. Like, I, I told them, like,
154:37 I, I don't do this for the money, I don't give a fuck. I- - Yeah.
154:42 - I mean, of course, of course it's a nice compliment but I wanna have
154:48 fun and have impact, and that's ultimately what made my decision.
154:58 - Can I ask you about... we've talked about it quite a bit, but maybe
155:02 just zooming out about how OpenCloud works. We talked about different
155:06 components, I want to ask if there's some interesting stuff we missed.
155:11 So, there's the gateway, there's the chat clients, there's the harness there's the agentic loop. You
155:20 said somewhere that everybody should im- implement an agent loop
155:24 at some point in their lives. - Yeah, because it's like the, it's like the Hello World in AI, you know? And
155:28 it's actually quite simple. - Yeah. - And it- it's good to understand that
155:34 that stuff's not magic. You can, you can easily build it yourself. So,
155:40 writing your own little cloud code... I, I even did this at a
155:44 conference in Paris for people to, like, introduce them to AI. I think it's it's a
155:49 fun little practice. Um, and you, you covered a lot. I think
155:55 one, one silly idea I had that turned out to be quite cool is
156:02 I built this thing with full system access. So it's like, you know, with great power comes great
156:09 responsibility. And I was like, "How can I up the stakes a little bit more?"
156:13 - Yeah, right. - And I just made a... I made it proactive. So, I added
156:20 a prompt. Initially, it was just a prompt, surprise me. Every, like, half an hour, surprise me, you know? And
156:28 later on I changed it to be like a little more specific and-
156:31 ... in the definition of surprise. Um, but the fact that I made it proactive and that it
156:38 knows you and that it cares about you, it- it's at least it's
156:44 programmed to that, prompted to do that. And that, that is a follow on, on your current session makes it very interesting because it would just sometimes ask a follow-up question or like, "How's your day?"
156:53 And I just made a... I made it proactive. So, I added a prompt. Initially, it was just a prompt, surprise me. Every, like, half an hour, surprise me, you know?
156:56 And later on I changed it to be like a little more specific and-
156:58 And that, that is a follow on, on your current session makes it very interesting because it would just sometimes ask a follow-up question or like, "How's your day?" I mean,
157:02 again, it's a little creepy or weird or interesting but Heartbeat very... in the beginning, it's still... today, it
157:12 doesn't... the model doesn't choose to use it a lot. - By the way, we're, we're, we're talking about Heartbeat, as you mentioned, the thing that
157:20 regularly- - Yeah. Like kicks- - ... Acts. - You just kick off the loop.
157:25 - Isn't that just a cron job, man? - Yeah, right, I mean, it's like-
157:29 - It's the cr- the criticisms that you get are hilarious. - You can, you can deduce any idea to like a
157:35 silly... Yeah, it's just, it's just a cron job in the end. I have like cron- separate cron jobs.
157:41 - Isn't love just evolutionary biology manifesting itself and isn't... aren't you guys just using each other?
157:49 - And then, yeah, and the project is all just glue of a few different dependencies-
157:53 ... and there's nothing original. Why do people... Well, you know,
157:57 isn't Dropbox just FTP with extra steps? - Yeah. - I found it surprising where I had this I had a
158:05 shoulder operation a few months ago, so. - Mm-hmm. - And the model rarely used Heartbeat, but then I was in the
158:11 hospital, and it knew that I had the operation and it checked up on
158:15 me. It's like, "Are you okay?" And I just... It's like, again, apparently, like, if something's significant in the
158:23 context, that triggered the Heartbeat when it rarely used the Heartbeat.... um,
158:30 and it does that sometimes for people, and that just makes it a lot more relatable.
158:36 - Uh, let me look this up on Perplexity, how OpenCall works just to see if I'm missing any of the stuff.
158:44 Local agent run time, high-level architecture. There's... Oh, we haven't talked much
158:48 about skills, I suppose. Skill hub, the tools in the skill lair,
158:52 but that's definitely a huge component and there's a huge growing set of skills-
158:55 - You know, you know what I love? That half a year ago, like everyone was talking about MCPs-
159:02 - Yeah - ... and I was like, "Screw MCPs. Uh, every MCP would be better as a
159:10 CLI." And now this stuff doesn't even have MCP support. I mean, it, it has with
159:17 asterisks, but not in the core lair, and nobody's complaining.
159:23 - Mm-hmm. - So my approach is if you want to extend the model with more features, you
159:32 just build a CLI and the model can call the CLI, probably gets it wrong, calls the help menu, and then on demand loads
159:43 into the context what it needs to use the CLI. It just needs a
159:47 sentence to know that the CLI exists if it's something that the model doesn't know about default. And even for a while,
159:54 I, I didn't really care about skills, but skills are actually perfect for
159:58 that because they, they boil down to a single sentence that explains the skill and then the model loads the skill, and that explains the
160:08 CLI, and then the model uses the CLI. Some skills are, like raw, but
160:13 most of the time, networks. - It's interesting um, I'm asking Perplexity MCP versus skills,
160:20 because this kind of requires a hot take that's quite recent, because your general view is MCPs are dead-ish. So MCPs is a more structured
160:32 thing. So if you listen to Perplexity here, MCP is what can I reach? So APIs,
160:38 database services files via protocol. So a structured protocol of
160:42 how you communicate with a thing, and then skills is more how should I work? Procedures, hostile helper scripts and prompts
160:49 are often written in a kind of semi- structured natural language,
160:53 right? And so technically skills could replace MCP if you have a smart enough model.
161:00 - I think the main beauty is, is that models are really good at
161:03 calling Unix commands. So if you just add another CLI, that's just another Unix command in the end. And
161:11 MCP is... That has to be added in training. That's not a very natural thing for the model. It requires a very specific
161:19 syntax. And the biggest thing, it's not composable. So imagine if I have a service that gives me better data
161:27 and gives me the temperature, the average temperature, rain,
161:31 wind and all the other stuff, and I get like this huge blob back. As a model, I always have to get the
161:38 huge blob back. I have to fill my context with that huge blob and then
161:42 pick what I want. There's no way for the model to naturally filter
161:48 unless I think about it proactively and add a filtering way into my
161:51 MCP. But if I would build the same as a CLI and it would give me this huge blob, it could just add a JQ command and filter itself
162:00 and then only, only get me what I actually need. Or maybe even
162:04 compose it into a script to, like do some calculations with the temperature and
162:08 only give me the exact output and the mo- and the... you have no context pollution.
162:14 Again, you can solve that with like sub- agents and more charades,
162:18 but it's just like workarounds for something that might not be the optimal way. There's... It definitely it
162:27 was, you know, it was good that we had MCPs because it pushed a lot of companies towards building APIs
162:34 and now I, I can like look at an MCP and just make it into a CLI.
162:37 - Mm-hmm. - Um, but this, this inherent problem that MCPs by default clutter up your
162:43 context. Plus the fact that most MCPs are not made good, in general make it just not a very useful paradigm. There's some exceptions like
162:57 Playwright for example that requires state and it's actually
163:01 useful. That is an acceptable choice. - So Playwright you use for browser use, which I think is c- already in
163:09 OpenClaw is quite incredible, right? - Yeah. - You can basically do everything, most things you can think of using browser use.
163:17 - That, that gets into the whole arch of every app is just a very slow API now, if they want or not. And that
163:28 through personal agents a lot of apps will disappear. You know, like I had a... I built
163:39 a CLI for Twitter. I mean, I- I just reverse engineered their website and used the internal API, which
163:49 is not very allowed. - It's called Bird, short-lived. - It was called Bird, because the bird had to disappear.
163:57 - The, the wings were clipped. - All they did is they just made access slower. Yeah,
164:03 not tak- you're not actually taking a feature away, but now inst- if, if your agent wants to read a tweet, it actually has to open the browser and
164:09 read the tweet. And it will still be able to read the tweet. It will just take longer.
164:13 It's not like you are making something that was possible, not possible. No. Now, it's just taking... Now it's just a
164:21 bit slower. So, so it doesn't really matter if your service wants to be an API or not. If I can
164:30 access it in the browser...... easy API. It's a slow API. - Can you empathize with their situation? Like, what would you do if you were Twitter, if you were
164:40 X? Because they're basically trying to protect against other large
164:43 companies scraping all their data. - Yeah. - But in so doing, they're cutting off like a million different use cases for
164:50 smaller developers that actually want to use it for helpful cool stuff.
164:54 - I think that if you have a very low per day baseline per account that allows read-only access
165:05 would solve a lot of problems. There's plenty, plenty of automations where people
165:09 create a bookmark and then use OpenClaw to, like, find the bookmark, do research on it, and then send you an email-
165:16 - Mm-hmm - ... with, like, more details on it or a summary. That's a cool approach.
165:21 I also want all my bookmarks somewhere to search. I would still like to have that.
165:26 - So, read-only access for the bookmarks you make on X. That seems like an
165:29 incredible application because a lot of us find a lot of cool stuff on X, we bookmark,
165:33 that's the general purpose of X. It's like, holy shit, this is awesome.
165:37 Oftentimes, you bookmark so many things you never look back at them.
165:40 - Yeah. - It would be nice to have tooling that organizes them and allows you to research it further.
165:44 - Yeah, I mean, and to be frank, I, I mean, I, I told Twitter proactively that, "Hey, I built
165:52 this and there's a need." And they've been really nice, but also like,
165:57 "Take it down." Fair. Totally fair. But I hope that this woke up the team a little bit that there's a need. And if all you do
166:09 is making it slower, you're just reducing access to your platform. I'm sure there's a better way. I also, I'm very much
166:19 against any automation on Twitter. If you tweet at me with AI, I will block you. No first strike. As
166:27 soon as it smells like AI, and AI still has a smell. - Mm-hmm.
166:32 - Especially on tweets. It's very hard to tweet in a way that does look completely human.
166:38 - Mm-hmm. - And then I block. Like, I have a zero tolerance policy on that. And
166:44 I think it would be very helpful if they, if, like, tweets done via API would be marked.
166:53 Maybe there's some special cases where... But, and there should
166:57 be, there should be a very easy way for agents to get their own
167:01 Twitter account. Um... - Mm-hmm. - We, we need to rethink social platforms a little bit if, if, if we, we, we go
167:11 towards a future where everyone has their agent and agents maybe have their own
167:17 Instagram profiles or Twitter accounts, so I can, like, do stuff on my behalf. I think it should
167:23 very clearly be marked that they are doing stuff on my behalf and it's not
167:26 me. Because content is now so cheap. Eyeballs are the expensive part. And
167:36 I find it very triggering when I read something and then I'm like, oh, no, this smells
167:40 like AI. - Yeah. Like, where, where is this headed in terms of what we value about the human
167:47 experience? It feels like we'll, we'll move more and more towards in-person interaction
167:52 and we'll just communicate. We'll talk to our AI agent to, to accomplish different tasks, to learn about different
168:01 things, but we won't value online interaction because there'll be so much
168:08 AI slob that smells and so many bots that it's difficult. - Well, if it's smart, then it shouldn't be difficult to filter. And then
168:20 I can look at it if I want to. But yeah, this is, like, a big thing we need to solve right now. E-
168:28 especially on this project, I get so many emails that are, let's say nicely, agentically written.
168:36 - Yeah. - But I much rather read your broken English than your AI
168:44 slob. You know, of course there's a human behind it, and yet they, they prompt
168:48 it. I'd much rather read your prompt than what came out. Um,
168:52 I think we're reaching a point where I value typos again. Like, um...
169:00 Like, and I, I mean, it also took me a while to, like, come to the realization. I, on my
169:05 blog I experimented with creating a blog post with agents and
169:12 ultimately it took me about the same time to, like, steer agent
169:16 towards something I like. But it missed the nuances that, how I would write it. You know, you can like, you
169:25 can steer it towards your style, but it's not gonna be all your style. So, I, I completely
169:32 moved away from that. I, I, everything, everything I blog is organic, handwritten
169:38 and maybe, maybe I, I, I use AI as a fix my worse typos. But
169:47 there's value in the rough parts of an actual human. - Isn't that awesome? Isn't that beautiful? That now because of AI
169:58 we value the raw humanity in each of us more. - I also, I also realized this thing that I, I rave about AI and use it so much for
170:08 anything that's code, but I'm allergic if it's stories. - Right. Yeah.
170:14 - Also, documentation, still fine with AI. You know, better than nothing.
170:17 - And for now it's still i- it applies in the mi- in the visual medium
170:21 too. It's fascinating how allergic I am to even a little bit of AI slob in in video
170:28 and images. It's useful, it's nice if it's like a little component of like-
170:32 - Or even, even those images. The, like, all these infographics and stuff,
170:36 the-... they trigger me so hard. - Yeah. - Like, it immediately makes me think less of your content. And it ... They were
170:45 novel for, like, one week and now it just screams slop. - Yeah.
170:51 - Even- even if people work hard on it, using ... And I- I have some on my blog post, you know, in the- in the time where I- I
171:00 explored this new medium. But now, they trigger me as well. It's like, yeah, this
171:04 is ... This just screams AI slop. I- - What... I don't know what that is, but I went through that too. I was really excited by the diagrams.
171:10 And then I realized, in order to remove from them hallucinations, you actually have to
171:14 do a huge amount of work. And you're just using it to draw the better diagrams, great. And then I'm proud of the diagram. I've used them for
171:22 literally, like, ki- ki- kind of like you said for maybe a couple of weeks. And now I look at
171:26 those, and I- I feel like I feel when I look at Comic Sans as a
171:30 font or- or something like this. It's like, "No, this is-" - It's a smell.
171:35 - "... this is fake. It's fraudulent. There's something wrong with it." And it...
171:41 - It's a smell. - And it's awesome because it re- it reminds you that we know.
171:48 There's so much to humans that's amazing and we know that. And we- we know
171:52 it. We know it when we see it. And so that gives me a lot of hope,
171:59 you know? That gives me a lot of hope about the human experience. It's not going to be damaged
172:03 by ... It's only going to be empowered as tools by AI. It's not going to be damaged or limited or somehow altered to where it's no longer
172:14 human. So ... Uh, I need a bathroom break. Quick pause. You mentioned that a lot of the apps might be
172:24 basically made obsolete. Do you think agents will just transform the entire app market?
172:30 - Yeah. Uh, I noticed that on Discord, that people just said
172:38 how their ... like, what they build and what they use it for. And it's
172:42 like, why do you need MyFitnessPal when the agent already knows where I am?
172:48 So, it can assume that I make bad decisions when I'm at, I don't know, Waffle House, what's around here? Or- or briskets in Austin.
172:58 - There's no bad decisions around briskets, but yeah. - No, that's the best decision, honestly. Um-
173:03 - Your agent should know that. - But it can, like ... It can modify my- my gym workout based
173:08 on how well I slept, or if I'm ... if I have stress or not. Like, it has so much more context to make even better
173:16 decisions than any of this app even could do. - Mm-hmm. - It could show me UI just as I like. Why do I still need an
173:25 app to do that? Why do I have to ... Why should I pay another subscription for
173:29 something that the agent can just do now? And why do I need my- my Eight Sleep app to control
173:38 my bed when I can tell the a- ... tell the agent to ... You know, the agent already
173:42 knows where I am, so he can, like, turn off what I don't use.
173:45 - Mm-hmm. - And I think that will ... that will translate into a whole
173:50 category of apps that are no longer ... I will just naturally stop using because my agent can just do it better.
174:00 - I think you said somewhere that it might kill off 80% of apps.
174:04 - Yeah. - Don't you think that's a gigantic transformative effect on
174:09 just all software development? So that means it might kill off a lot of software
174:13 companies. - Yeah. Um- - It's a scary thing. So, like, do you think about
174:19 the impact that has on the economy? On, Just the ripple effects it has to society?
174:27 Transforming who builds what tooling. It empowers a lot of users to get stuff done, to get stuff more efficiently, to get it done cheaper.
174:41 - It's also new services that we will need, right? For example, I want my agent to have an allowance. Like,
174:50 you solve problems for me, here's like 100 bucks in order to solve
174:54 problems for me. And if I tell you to order me food, maybe it uses a service. Maybe it uses something
175:01 like rent-a-human to, like, just get that done for me. - Mm-hmm.
175:06 - I don't actually care. I care about solve my problem. There's space for-
175:13 for new companies to solve that well. Maybe don't ... Not all apps
175:17 disappear. Maybe some transform into being API. - So, basically, apps that rapidly transform in being agent-facing.
175:30 So, there's a real opportunity for, like, Uber Eats, that we just used earlier today. It- it's companies this, of which there's many.
175:42 Who gets there fastest to being able to interact with OpenClaw in a way that's
175:48 the m- the most natural, the easiest? - Yeah. And also, apps will become API if they want or not. Because
175:57 my agent can figure out how to use my phone. I mean, on- on the other side,
176:02 it's a little more tricky. On Android, that's already ... People already do that. And then we'll just click the Order Uber
176:10 for Me button for me. Um, or maybe another service. Or maybe there's- there's a ... there's an API I can call so it's
176:17 faster. Uh, I think that's a space we're just beginning to even understand what that means. And I
176:26 ... Again, I didn't even ... That was not something I thought of. Something that I-
176:31 that I discovered as people use this, and it ... We are still so
176:34 early. But yeah, I think data is very important. Like, apps that can give me data, but that also can be
176:41 API. Why do I need a Sonos app anymore when I can ... when my agent can
176:45 talk to the Sonos?... Speakers directly. Like my cameras, there's like a crappy app,
176:52 but they have, they have an API, so my agent uses the API now.
176:57 - So it's gonna force a lot of companies to have to shift focus. That's kind
177:01 of what the internet did, right? You have to rapidly rethink, reconfigure what you're selling, how you're making money.
177:10 - Yeah, and some companies were really not like that. For example, there's no
177:14 CLI for Google, so I had to like, do... have to do anything myself and
177:21 build GAWK. That's like a CLI for Google. And at the... Yeah, at the end
177:28 user, they have to give me the emails because otherwise I cannot use their product. If I'm a company and I try
177:36 to get Google data, Gmail, there's a whole complicated process, to the point where sometimes
177:44 startups acquire startups that went through the process, so they
177:47 don't- don't have to work with Google for half a year to be certified
177:51 to being able to access Gmail. But my agent can access Gmail because I can just connect to it.
177:58 It's still crappy because I need to, like, go through Google's
178:03 developer jungle to get a key, and that's still annoying. But they cannot prevent me. And worst case, my
178:13 agent just clicks on the, on the website and gets the data out that way.
178:17 - Through browsers? - Yeah. I mean, I, I watch my agent happily click the I'm not a robot button.
178:25 And there's this, this whole... That's gonna be... That's gonna be more heated. You see companies like Cloudflare that
178:36 try to prevent bot access. And in some ways, that's useful for
178:39 scraping. But in other ways, if I'm, I'm a personal user, I want that. You know, sometimes
178:46 I, I use Codex and I, I read an article about modern React patterns,
178:55 and it's like a Medium article. I paste it in and the agent can't read
178:59 it because they block it. So then I have to copy-paste the actual text. Or in
179:05 the future, I'll learn that maybe I don't click on Medium because it's annoying, and I use
179:09 other websites that actually are agent friendly. So, uh- - There's gonna be a lot of powerful, rich companies fighting back.
179:17 So it's really intere- You're at the center, you're the catalyst, the leader,
179:23 and happen to be at the center of this kind of revolution where it's get- gonna completely
179:26 change how we interact with services with, with web. And so, like, there's companies at
179:35 Google that are gonna push back. I mean, there's every major companies you could think of is gonna push
179:39 back. - Even... Yeah, even search. Um, I now use, I think Perplexity or Brave
179:47 as providers because Google really doesn't make it easy to use Google
179:51 without Google. I'm not sure if that's the right strategy, but I'm not
179:55 Google. - Yeah, there's a, there's a nice balance from a big company perspective 'cause if you
180:02 push back too much for too long, you become Blockbuster and you lose everything to the
180:06 Netflixes of the world. But some pushback is probably good during a
180:09 revolution to see. - Yeah. But you see that, that... Like, this is something that the people want.
180:14 - Right. - So- - Yes. - If I'm on the go, I don't wanna open a calendar app. I just... I wanna tell my agent,
180:22 "Hey, remind me about this dinner tomorrow night," and maybe invite
180:26 two of my friends and then maybe send a what- send a WhatsApp message to my
180:29 friend. And I don't need... I don't want or need to open apps for that. I think that
180:36 we passed that age, and now everything is, like, much more connected and, and fluid if those companies want it or
180:44 not. And I think, well, the right companies will find ways to jump on the train, and other companies will perish.
180:55 - You got to listen to what the people want. We talked about programming quite a
8:55 - There was... Like, one of my projects before already did something
8:59 where I could bring my terminals onto the web and then I could, like, interact with them, but there also would be terminals on my Mac.
9:07 - Mm-hmm. - Viptunnel, which was like a, a weekend hack project
9:12 that was still very early. And it was cloud code times. You know, you got a dopamine hit when you got something right. And now
9:20 I get, like, mad when you get something wrong. - And you had a really great -– not to take a tangent -– but a great blog post describing that
9:26 you converted Viptunnel. You vibe-coded Viptunnel from TypeScript into Zig of all programming languages with
9:34 a single prompt. One prompt, one shot. Convert the entire code base into Zig.
9:41 - Yeah. There was this one thing where part of the architecture
9:46 was... Took too much memory. Every terminal used like a node. And I wanted to change it to Rust
9:54 and... I mean, I can do it. I can, I can manually figure it all out, but all my automated attempts failed miserably.
10:08 And then I revisited about four or five months later. And I'm
10:12 like, "Okay, now let's use something even more experimental." And
10:16 I, and I just typed, "Convert this and this part to Sig," and then let Codex run off. And
10:23 it basically got it right. There was one little detail that I had to,
10:27 like, modify afterwards, but it just ran for overnight or like six hours and just did its thing. And it's
10:36 like... It's just mind-blowing. - So that's on the LLM programming side, refactoring. But uh, back to the
10:46 actual story of the of the prototype. So how did Viptunnel connect to the first
10:50 prototype where your, like, agents can actually work? - Well, that was still very limited. You know, like I had this one
10:56 experiment with WhatsApp, then I had this experiment, and both felt like
11:01 not the right answer. And then my search bar was literally just hooking up WhatsApp to
11:10 cloud code. One shot. The CLI message comes in. I call the CLI with -p. It does its
11:17 magic, I get the string back and I send it back to WhatsApp. And I, I built
11:21 this in one hour. And I felt... Already felt really cool. It's like, "Oh, I could... I can, like, talk to my
11:28 computer," right? This... That, that was, that was cool. But
11:32 I, I wanted images, 'cause I alw- I often use images when I prompt. I think it's such a, such an efficient way to give the agent more
11:39 context. And they are really good at figuring out what I mean, e- even if it's like
11:43 a, a weird cropped-up screenshot. So I used it a lot and I wanted to do
11:47 that in WhatsApp as well. Also, like, you know, just you run around, you see like
11:53 a poster of an event, you just make a screenshot and like figure out if I have time there, if this is good, if my friends are maybe up for that.
12:00 Just like images seemed im- important. So I, I worked a few... It took me a few more hours to actually get that right. Um,
12:09 and then it was just...... I, I used it a lot. And funny enough, that was
12:17 just before I went on a trip to Marrakesh with my friends for a birthday trip. And there it was even better because
12:26 internet was a little shaky but WhatsApp just works, you know? It's like doesn't
12:30 matter, you have, like, edge, it still works. WhatsApp is just... It's just made really well. So I ended up using it a lot. Um,
12:41 translate this for me, explain this, find me places. Like, you
12:45 just having a clanker doing, having Google for you, that was... Basically there was still nothing built but
12:52 it still could do so much. - So, if we talk about the full journey that's happening there with the agent,
12:58 you're just sending on this very thin line WhatsApp message via CLI, it's going to a cloud code and cloud code is doing all kinds of
13:08 heavy work and coming back to you with a thin message. - Yeah. It was slow because every time I boot up the CLI, but it... It was really cool
13:19 already. And it could just use all the things that I already had
13:23 built. I had built like a whole bunch of CLI stuff over the month so it, it felt
13:30 really powerful. - There is something magical about that experience that's hard to put into
13:34 words. Being able to use a chat client to talk to an agent, versus, like, sitting behind
13:44 a computer and like, I don't know, using cursor or even using Cloud Code CLI in the terminal. It's a different experience than being
13:51 able to sit back and talk to it. I mean, it seems like a trivial step
13:55 but, it- in some sense it's a... It's like a phase shift in the integration of AI into your life and how it feels, right?
14:05 - Yeah. Yeah. I, I read this tweet this morning where someone said, "Oh, there's
14:09 no magic in it. It's just like, it does this and this and this and this and this and this." And it
14:16 almost feels like a hobby, just as cursor or perplexity. And I'm
14:20 like, well, if that's a hobby that's kind of a compliment, you
14:23 know? They're like, they're not doing too bad. Thank you I guess?
14:32 Yes. I mean, isn't, isn't, isn't magic often just like you take a lot of things that are already there
14:39 but bring them together in new ways? Like, I don't... There's
14:43 no... Yeah. Maybe there's no magic in there but sometimes just
14:46 rearranging things and, like, adding a few new ideas is all the magic that you need.
14:51 - It's really hard to convert into words what is, what is magic
14:55 about a thing. If you look at the, the scrolling on an iPhone, why is that so pleasant? There's a lot of elements about that
15:02 interface that makes it incredibly pleasant, that is fundamental to the experience of
15:06 using a smartphone, and it's like, okay, all the components were
15:10 there. Scrolling was there, everything was there. - Nobody did it-
15:14 - Yep - ... and afterwards it felt so obvious. - Yeah, so obvious.
15:16 - Right? But still... You know the moment where it, it blew my mind was when,
15:25 when I- I used it a lot and then at some point I just sent it a message
15:29 and, and then a typing indicator appeared. And I'm like, wait, I
15:35 didn't build that, it only m- it only has image support, so what is it even doing? And then it would just reply.
15:42 - What was the thing you sent it? - Oh, just a random question like, "Hey, what about this in this restaurant?" You
15:47 know? Because we were just running around and checking out the city.
15:52 So that's why I, I didn't, didn't even think when I used it because
15:56 sometimes when you're in a hurry typing is annoying. - So, oh, you did an audio message?
16:00 - Yeah. And it just, it just worked and I'm like... - And it's not supposed to work because-
16:05 - No - ... you didn't give it that- - No, literally - ... capability.
16:08 - I literally went, "How the fuck did he do that?" And it was like, "Yeah,
16:12 the mad lad did the following. He sent me a message but it only, only was a file and no file ending." So
16:19 I checked out the header of the file and it found that it was,
16:23 like, opus so I used ffmpeg to convert it and then I wanted to use whisper but it didn't had it installed. But then I found the
16:31 OpenAI key and just used Curl to send the file to OpenAI to translate and here I am.
16:39 Just looked at the message I'm like, "Oh wow." - You didn't teach it any of those things and the agent just figured it out, did all those conversions,
16:47 the translations. It figured out the API, it figured out which program to
16:51 use, all those kinds of things. And you were just absent-mindedly just sent an
16:54 audio message when it came back. - Yeah, like, so clever even because he would have gotten the whisper local path, he would have had to download a
17:00 model. It would have been too slow. So like, there's so much world
17:04 knowledge in there, so much creative problem solving. A lot of it
17:08 I think mapped from... If you get really good at coding that means you have to
17:12 be really good at general purpose problem solving. So that's a skill, right? And
17:16 that just maps into other domains. So it had the problem of like, what is this file with no file ending? Let's figure it
17:23 out. And that's when it kind of clicked for me. It's like, I was like very
17:30 impressed. And somebody sent a pull request for Discord support and I'm like, "This is a WhatsApp relay.
17:37 That doesn't, doesn't fit at all." - At that time it was called WA Relay.
17:42 - Yeah. And so I debated with me like, do I want that? Do I not want that? And then
17:51 I thought, well maybe, maybe I do that because that could be a cool way to show people. Because I... So far I did it in WhatsApp
18:00 as like groups you know but don't really want to give my phone number to every internet stranger.
18:07 - Yeah. - Um, journalists manage to do that anyhow now so that's a different
18:11 story. So I merged it-... from Shadow, who helped me a lot with the whole project. So, thank you. And, and I put
18:24 my, my bot in there. - On Discord? - Yeah. No security because I didn't... I hadn't built sandboxing in
18:31 yet. I, I just prompted it to, like, only listen to me. And then some people came and tried to hack it, and I
18:41 just... Or, like, just watched and I just kept working in the open, you
18:45 know? Like, y- I used my agent to build my agent harness and to test, like, various stuff. And that's
18:57 very quickly when it clicked for people. So it's almost like it needs to
19:01 be experienced. And from that time on, that was January the 1st, I, I got
19:09 my first real influencer being a fan and did videos, dachitze. Thank you. And, and
19:17 from there on, I saw, I started gaining up speed. And at the same time, my,
19:23 my sleep cycle went shorter and shorter because I, I felt the storm
19:29 coming, and I just worked my ass off to get it to... into a state where it's kinda
19:37 good. - There's a few components and we'll talk about how it all works, but basically, you're able to
19:42 talk to it using WhatsApp, Telegram, Discord. So that's a component that you have to get right.
19:48 - Yeah. - And then you have to figure out the agentic loop, you have to have the gateway, you
19:52 have the harness, you have all those components that make it all just work nicely.
19:56 - Yeah. It felt like Factorio times infinite. - Right. - I, I feel like I built my little- ... my little
20:04 playground. Like, I never had so much fun than building this project. You know?
20:08 Like, you have like, "Oh," I go like, level one agentic loop. What can I do there?
20:12 How can I be smart at queuing messages? How can I make it more
20:15 human-like? Oh, then I had this idea of... Because the loop always... The agent always replies something, but you don't
20:23 always want an agent to reply something in a group chat. So I gave him this
20:27 no-reply token. So I gave him an option to shut up. So it, it feels more natural.
20:32 - That's level two. - Y- uh, yeah, yeah. Yeah, on the- on the-
20:36 - Factorio. - On the agentic loop. And then I go to memory, right?
20:39 - Yeah. - You want him to, like, remember stuff. So maybe, maybe the
20:42 end... The ultimate boss is continuous reinforcement learning,
20:46 but I'm, I'm, like, at... I feel like I'm level two or three with Markdown
20:50 files and the vector database. And then you, you can go to level
20:54 community management, you can go to level website and marketing. There's just so many hats that you have to have on. Uh,
21:01 not even talking about native apps. That's just, like, infinite
21:05 different levels and infinite level ups you can do. - So the whole time you're having fun. We should say that for the most
21:11 part, throughout this whole process, you're a one-man team. There's people helping, but you're doing so much of the key core development.
21:21 - Yeah. - And having fun? You did, in January, 6,600 commits. Probably more.
21:28 - I sometimes posted a meme. I'm limited by the technology of my time. I could do more
21:32 if agents would be faster. - But we should say you're running multiple agents at the same time.
21:37 - Yeah. Depending on how much I slept and how difficult of the tasks I work on, between four and 10.
21:45 - Four and 10 agents. Uh there's so many possible directions, speaking of
21:49 Factorio, that we can go here. But one big picture one is, why do you think
21:56 your work, Open Claw, won? In this world, if you look at 2025, so many
22:05 startups, so many companies were doing kind of agentic type stuff, or claiming to. And here, Open Claw comes
22:12 in and destroys everybody. Like, why did you win? - Because they all take themselves too serious.
22:18 - Yeah. - Like, it's hard to compete against someone who's just there to have fun.
22:24 - Yeah. - I wanted it to be fun, I wanted it to be weird. And if you see, like,
22:28 all the, all the lobster stuff online I think I, I managed weird. I... You
22:36 know, for the longest time, the only, the only way to install it was git clone, pnpm build, pnpm gateway.
22:46 Like, you clone it, you build it, you run it. And then the, the agent... I made the agent very aware. Like, it
22:53 knows that it is... What its source code is. It understands th- how it sits and runs in its own
23:01 harness. It knows where documentation is. It knows which model it runs. It knows if you turn on the voice or, or reasoning mode. Like,
23:11 I, I wanted to be more human-like, so it understands its own system that made it very easy for an agent to...
23:18 Oh, you don't like anything? You just prompted it to existence, and then the agent would
23:22 just modify its own software. Um, you know, we have people talk about self- modifying software. I just built it and
23:30 didn't even... I didn't even plan it so much. It just happened.
23:35 - Can you actually speak to that? 'Cause it's just fascinating. So you have this
23:39 piece of software that's written in TypeScript- - Yeah - ... that's able to, via the agentic loop, modify
23:47 itself. I mean, what a moment to be alive in the history of humanity and the history of programming. Here's
23:55 the thing that's used by a huge amount of people to do incredibly powerful things in their lives, and that very
24:02 system can rewrite itself, can modify itself. Can you just, like, speak to the power of that? Like, isn't
24:10 that incredible? Like, when did you first close the loop on that?
24:14 - Oh, because that's how I built it as well, you know? Most of it is built by
24:18 Codex, but oftentimes I... When I debug it, I...... I use self-introspection so much.
24:26 It's like, "Hey, what tools do you see? Can you call the tool yourself?" Or like, "What error do
24:30 you see? Read the source code. Figure out what's the problem."
24:32 Like, I just found it an incredibly fun way to... That the agent, the very agent and software that you
24:40 use is used to debug itself, so that it felt just natural that everybody does that. And that it led to so many,
24:51 so many pull requests by people who never wrote software. I mean, it
24:54 also did show that people never wrote software . So I call them prompt requests
24:58 in the end. But I don't want to, like, pull that down because every time someone made the first pull request is a
25:06 win for our society, you know? Like, it... Like, it doesn't matter
25:10 how, how shitty it is, y- you gotta start somewhere. So I know there's, like, this whole big movement of people
25:17 complain about open source and the quality of PRs, and a whole different level of
25:21 problems. But on a different level, I found it... I found it
25:28 very meaningful that, that I built something that people love to think
25:32 of so much that they actually start to learn how open source works.
25:37 - Yeah, you were ... The Open Cloud project was the first pull request. You were the first for so many. That is
25:44 magical. So many people that don't know how to program are taking their first step into the programming world with this.
25:52 - Isn't that a step up for humanity? Isn't that cool? - Creating builders.
25:56 - Yeah. Like, the bar to do that was so high, and, like, with
26:00 agents, and with the right software, it just, like, went lower and lower.
26:04 I don't know. I was at a... And I also organize another type of meetup. I call it... I called it Cloud Code Anonymous.
26:14 Uh, you can get the inspiration from. Now, I call it Agents Anonymous- ... for, for reasons.
26:23 - Agents Anonymous. - And- - Oh, it's so funny on so many levels. I'm sorry, go ahead.
26:29 - Yeah. And there was this one guy who, who talked to me. He's like, "I run this
26:33 design agency, and we, we never had custom software. And now I have, like, 25 little web services for various
26:41 things that help me in my business. And I don't even know how
26:45 they work, but they work." Uh, and he was just, like, very happy that
26:52 my stuff solved some of his problems. And he was, like, curious enough that he
26:56 actually came to, like, a, a Enchantic meetup, even though he's... He doesn't really know how software works.
27:04 - Can we actually rewind a little bit and tell the saga of the name change?
27:10 First of all, it started out as Wa-Relay. - Yeah. - And then it went to-
27:13 - Claude's. - Yeah. You know, when I, when I built it in the beginning, my agent had no personality.
27:19 It was just... It was Claude Code. It's like this sycophantic
27:23 opus, very friendly. And I... When you talk to a friend on WhatsApp, they don't
27:30 talk like Claude Code. So I wanted... I, I felt this... I just didn't f- It didn't feel right, so I, I wanted to give it
27:40 a personality. - Make it spicier, make it- - Yeah - ... something. By the way, that's actually hard to put into words as well. And we should mention
27:47 that, of course, you create the soul.md, inspired by Anthropic's constitutional AI work-
27:53 - Mm-hmm - ... how to make it spicy. - Partially, it picked up a little bit from me. You know, like those things
27:58 are text completion engines in a way. So, so I, I, I, I had fun working with it, and then I told it to... How I wanted it to
28:11 interact with me, and just, like, write your own agents.md, Give yourself a name. And then we... I didn't even know how the whole,
28:22 the whole lobster... I mean, people only do lobster... Originally, it was actually a lobster in
28:26 a, in a TARDIS, because I'm also a big Doctor Who fan. - Was there a space lobster?
28:31 - Yeah. - I heard. What's that have to do with anything? - Yeah, I just wanted to make it weird.
28:37 There was no... There was no big grand plan. I'm just having fun here.
28:40 - Oh, so I guess the lobster is already weird, and then the space lobster is an extra weird.
28:44 - Yeah, yeah, because the- - Yeah - ... the TARDIS is basically the, the harness, but
28:50 cannot call it TARDIS, so we called it Claude's. So that was name number two.
28:54 - Yeah. - And then it never really rolled off the tongue. So when more people came, again, I talked with my agent,
29:06 Claude. At least that's what I used to call him. Now- - Claude spelled with a W-C-L-A-U-D-E.
29:12 - Yeah. - Versus C-L-A-U-D-E from Anthropic. - Yeah. - Which is part of what makes it funny,
29:24 I think. The play on the letters and the words in the TARDIS and the
29:28 lobster and the space lobster is hilarious. But I can see why
29:32 it can lead into problems. - Yeah, they didn't find it so funny .
29:39 So then I got the domain ClaudeBot, and I just... I love the domain. And
29:45 it was, like, short. It was catchy. I'm like, "Yeah, let's do
29:48 that." I didn't... I didn't think it would be that big at this time.
29:55 And then just when it exploded, I got, Kudos, a very friendly email from one of the employees
30:06 that they didn't like the name. - One of the Anthropic employees.
30:11 - Yeah. So actually, Kudos, because they shou- could have just sent a, a lawyer letter, but they've been nice about
30:18 it. But also like, "You have to change this and fast." And I asked for two days,
30:26 because changing a name is hard, because you have to find everything, you
30:37 And everything has to be...... you need a set of everything.
30:41 - And also, can we comment on the fact that you're increasingly attacked, followed by
30:47 crypto folks? Which I think you mentioned somewhere that that means the name
30:51 change had to be... Because they were trying to snipe, they were trying to steal, and so you had to be... The, the na- I mean, from an
30:59 engineering perspective, it's just fascinating. You had to make the name change
31:03 Atomic, make sure it's changed everywhere at once. - Yeah. Failed very hard at that.
31:08 - You did? - I, I underestimated those people. It's a, it's a very
31:16 interesting subculture. Like, it... Everything circles around... I'll probably get a lot wrong and we'll probably get
31:24 hate for that if you say that, but... There is like Bags app and then they, they tokenize everything. And th- they did the
31:31 same back with Swipe Tunnel, but to a much smaller degree. It was not that annoying. But on this project, they've
31:39 been, they've been swarming me. They, they... It's like every half an hour,
31:46 someone came into Discord and, and, and spammed it and we had to block the p- We have,
31:50 like, server rules, and one of the rules was... One of the rules
31:54 is no mentioning of butter. For obvious reasons. And one was, no talk about finance stuff or
32:01 crypto. Because I'm... I- I'm just not interested in that, and this
32:08 is a space about the project and not about some finance stuff.
32:13 But yeah. They came in and, and spammed and... Annoying. And on Twitter, they would ping me all the time.
32:20 My, my notification feed was unusable. I, I could barely see actual people talking about this stuff because it was like swarms.
32:28 - Mm-hmm. - And everybody sent me the hashes. Um... And they all try me to
32:35 claim the fees. Like, "Are you helping the project?" Claim the fees. No,
32:39 you're actually harming the project. You're, like, disrupting my
32:43 work, and I am not interested in any fees. I'm... First of all, I'm financially comfortable. Second of
32:50 all, I don't want to support that because it's so far the worst form of online harassment that I've experienced.
32:59 - Yeah. There's a lot of toxicity in the crypto world. It's sad because
33:03 the technology of cr- cryptocurrency is fascinating, powerful and maybe
33:08 will define the future of money, but the actual community around that, there's so much to- toxicity, there's so much greed. There's so much
33:16 trying to get a shortcut to manipulate, to, to steal, to snipe,
33:20 to, to, to, to game the system somehow to get money. All this kind of
33:24 stuff that... Uh... I mean, it's the human nature, I suppose, when you
33:28 connect human nature with money and greed and and especially in
33:32 the online world with anonymity and all that kind of stuff. But from the
33:36 engineering perspective, it makes your life challenging. When Anthropic
33:39 reaches out, you have to do a name change. And then there- there's, there's like all these, like, Game of Thrones or Lord of the Rings
33:48 armies of different kinds you have to be aware of. - Yeah. There was no perfect name, and I didn't sleep for two nights.
33:55 I was under high pressure. Um, I was trying to get, like, a good set of
34:02 domains and, you know, not cheap, not easy, 'cause in this, in this state of the internet, you basically have to
34:10 buy domains if you want to have a good set. And, and then another ca- another email came in that the lawyers are getting uneasy.
34:22 Again, friendly, but also just adding more stress to my situation already. So at this point I was just like,
34:31 "Sorry, there's no other word. Fuck it." And I just, I just renamed it to
34:35 Mod Bot 'cause that was the set of domains I had. I was not really happy, but I thought it'll be fine.
34:43 And I tell you, everything that could go wrong- ... did go wrong. Everything that could go wrong did go wrong. It's incredible.
34:51 I, I, I thought I, I had mapped the h- the space out and reserved the important things.
34:58 - Can you ga- give some details of the stuff that gone wrong? 'Cause it's interesting from, like, an
35:01 engineering perspective. - Well, the, the interesting stuff is that none of these services have, have a squatter
35:07 protection. So, I had two browser windows open. One was like a,
35:13 an empty account ready to be rename- renamed to Claude Bot, and the other one I renamed to Mod Bot. So, I pressed
35:20 rename there, I pressed rename there, and in those five seconds,
35:24 they stole the account name. Literally, the five seconds of dragging the mouse over there and
35:30 pressing rename there was too long. - Wow. - Because there's no... Those systems... I mean, you would expect that they have some
35:37 protection or, like, an automatic forwarding, but there's nothing like that. And I didn't know that
35:46 they're not just good at harassment, they're also really good
35:49 at using scripts and tools. - Yeah. - So, yeah. So, suddenly, like, the old account
35:56 was promoting new tokens and serving malware. And I was like, "Okay, let's move over to GitHub,"
36:05 and I pressed rename on GitHub. And the GitHub renaming thing is slightly confusing, so I
36:13 renamed my personal account. And in those... I guess it took me 30 seconds to realize my mistake. They sniped
36:21 my account, serving malware from my account. So, I was like, "Okay, let's at least do the NPM stuff," but
36:30 that takes, like, a minute to upload. They sniped, they sniped the NPM package, 'cause I could reserve
36:38 the account, but I didn't reserve the root package.... so like
36:44 everything that could go wrong , like went wrong. - Can I just ask a, a curious question of, in that moment you're sitting
36:50 there, like how shitty do you feel? That's a pretty hopeless feeling, right?
36:57 - Yeah. Because all I wanted was like having fun with that project and to
37:04 keep building on it. And yet here I am like days into researching names, picking a name I didn't like.
37:11 And having people that claimed they helped me making my life miserable in every possible way. And honestly, I was
37:22 that close of just deleting it. I was like, "I did show you the future, you build it."
37:30 - Yeah. - I... That was a big part of me that got a lot of joy out of
37:35 that idea. And then I thought about all the people that already co-
37:38 contributed to it, and I couldn't do it because they had plans with it, and they put time in it. And it just didn't feel right.
37:50 - Well, I think a lot of people listening to this are deeply grateful that you
37:54 persevered. But it's... I, I can tell. I can tell it's a low point. This is the first time you hit a wall of, this is not fun?
38:02 - No, no, I was like close to crying. It was like, okay, everything's fucked.
38:10 - Yeah. - Um... I am like super tired. - Yeah. - Uh, and now like how do you even, how do you undo that? You know, l- luckily, and
38:22 thankfully, like I, I have... Because I have a little bit of following already. Like I had friends
38:28 at Twitter, I had friends at GitHub who like moved heaven and earth to like help me. And it is not... That's not something
38:35 that's easy. Like, like GitHub tried to like clean up the mess and then they ran into like platform bugs .
38:45 'Cause it's not happening so often that things get renamed on that
38:49 level. So, it took them a few hours. The MBM stuff was even more difficult because it's a whole different team.
38:57 On the Twitter side, things are not as easy as well. It, it took them like a day
39:04 to really also like do the redirect. And then I also had to like
39:13 do all the renaming in the project. Then there's also, uh, ClaudeHub, which I didn't
39:21 even finish the rename there because I, I, I managed to get people on
39:28 it and then someone just like collapsed and slept. And then I woke up and I'm like,
39:34 I made a, a beta version for the new stuff and I, I just, I just couldn't live with the name. It's like, you know... But
39:43 but, you know, it's just been so much drama. So, I had the real
39:47 struggle with me like I never want to touch that again, and I really don't like the
39:53 name. Um, so, and I... There was also this like... Then there was all the security people that started
40:03 emailing me like mad. Um, I was bombarded on Twitter, on email. There's like a thousand other things I should do.
40:13 And I'm like thinking about the name which is like, it should be
40:17 like the least important thing. Um, and then I was really close
40:24 in... Oh God, I don't even... Honestly, I don't even wanna say the, my other
40:32 name choices because it probably would get tokenized, so I'm not gonna say it.
40:38 - Yeah. - But I slept on it once more, and then I had the idea for OpenClaw
40:43 and that felt much better. And by then, I had the boss move that I
40:49 actually called Sam to ask if OpenClaw is okay. OpenClaw.AI. You know? 'Cause 'cause like-
40:57 - You didn't wanna go through the whole thing. Yeah. - Oh, that it's like, "Please tell me this is fine." I don't think
41:05 they can actually claim that, but it felt like the right thing to do.
41:11 And I did another rename. Like just Codex alone took like 10 hours to rename the
41:17 project 'cause it, it's a bit more tricky than a search replace and I, I wanted everything renamed, not just on the outside. And that
41:27 rename, I, I felt I had like my, my war room. But then I, I had like some contributors really that helped me. We made a whole plan
41:36 of all the names we have to squat. - And you had to be super secret about it?
41:40 - Yeah. Nobody could know. Like I literally was monitoring Twitter if like, if there's any mention
41:44 of OpenClaw. - Mm-hmm. - And like with reloading, it's like, "Okay, they don't, they don't expect anything
41:50 yet." Then I created a few decoy names. And all the shit I shouldn't have to
41:54 do. You know? Like, you know- - Yeah, yeah - ... it's helping the project. Like, I lost like 10 hours just by
41:59 having to plan this in full secrecy like, like a war game. - Yeah, this is the Manhattan Project of the 21st century. It's renaming-
42:08 - It's so s- ... so stupid. Uh like I still was like, "Oh, should I, should I keep it?"
42:12 Then I was like, "No, the mold's not growing on me." And then I think I had final all the pieces together. I didn't get a .com
42:23 but, yeah, it's been like quite a bit of money on the other domains. I tried to reach out again to
42:29 GitHub but I feel like I, I used up all my goodwill there, so I... 'Cause I, I, I wanted them to do this thing atomically-
42:39 - Mm-hmm - ... But that didn't happen and then so I did that the f- as first
42:41 thing. Uh, Twitter people were very supportive. I, I actually paid 10K for the
42:49 business account so I could claim the-... OpenClaw, which was, like, unused since 2016, but was claimed. And yeah, and then I
43:00 finally ... This time I managed everything in one go. Nothing, almost nothing got wrong. The only thing that did go wrong is that
43:11 I was not allowed by trademark rules to get OpenClaw.AI, and someone copied the website as serving malware.
43:21 - Yeah. - I'm not even allowed to keep the redirects. Like, I have to
43:27 return ... Like, I have to give Entropik the domains, and I cannot do redirects, so if you go on claw.bot next week, it'll just be a 404.
43:37 - Yeah. - And I- I'm not sure how trademark ... Like, I didn't,
43:44 I didn't do that much research into trademark law, but I think that could,
43:48 could be handled in a way that is safer, because ultimately those people will then Google and maybe find
43:59 malware sites that I have no control on them. - The point is, that whole saga, Made a dent in your whole f-
44:08 the funness of the journey, which sucks. So, let's just, let's just get, I suppose, get back to fun. And during this, speaking of
44:16 fun, the two-day MoltBot saga. - Yeah, two years. - MoltBook was created.
44:24 - Yeah. - Which was another thing that went viral as a kind of demonstration,
44:31 illustration of how what is now called OpenClaw could be used
44:37 to create something epic. So for people who are not aware, MoltBook is
44:41 just a bunch of agents talking to each other in a Reddit-style social network. And a bunch of people take
44:48 screenshots of those agents doing things like, Scheming against humans. And
44:56 that instilled in folks a kind of, you know, fear, panic, and
45:00 hype. W- what are your thoughts about MoltBook in general? - I think it's art. It is, it is like the finest slop, you know, just like the slop
45:12 from France. - Yeah. - I- I saw it before going to bed, and even though I was
45:21 tired, I spent another hour just reading up on that and, and just being entertained. I, I just felt
45:31 very entertained, you know? The- I saw the the reactions, and, like, there was one reporter who's calling me about, "This is the end of
45:40 the world, and we have AGI." And I'm just like, "No, this is just,
45:43 this is just really fine slop." You know, if, if I wouldn't have created this, this whole onboarding experience
45:50 where you, you infuse your agent with your personality and give him, give him character, I think that reflected on a lot of
46:01 how different the replies to MoltBook are. Because if it were all, if
46:05 it were all be ChatGPT or Cloud Code, it would be very different. It would be much more the same.
46:11 - Mm-hmm. - But because people are, like, so different, and they create their agents in so
46:16 different ways and use it in so different ways, that also reflects
46:20 on how they ultimately write there. And also, you, you don't know how much of that is really
46:27 done autonomic, autonomous, or how much is, like, humans being funny and, like, telling the agent, "Hey, write about the deep plan, the end of the world, on
46:34 MoltBook, ha, ha, ha." - Well, I think, I mean, my criticism
46:39 of MoltBook is that I believe a lot of the stuff that was screenshotted is human prompted. Which,
46:48 just look at the incentive of how the whole thing was used. It's obvious to me at least that a lot of it was humans
46:55 prompting the thing so they can then screenshot it and post it on
46:59 X in order to go viral. - Yeah. - Now, that doesn't take away from the artistic aspect of
47:04 it. The, the finest slop that humans have ever created . - For real. Like, kudos to, to Matt, who had this idea so quickly and pushed something
47:17 out. You know, it was, like, completely insecure security drama. But also,
47:24 what's the worst that can happen? Your agent account is leaked, and,
47:28 like, someone else can post slop for you? So like, people were,
47:32 like, making a whole drama about of the security thing, when I'm like, "There's nothing private
47:36 in there. It's just, like, agents sending slop." - Well, it could leak API keys.
47:41 - Yeah, yeah. There's like, "Oh, yeah, my human told me this and this, so I'm leaking his
47:45 security number." No, that's prompted, and the number wasn't even real. That's just
47:51 people, people trying to be badballs. - Yeah, but that- that's still, like, to me, really concerning, because of
47:58 how the journalists and how the general public reacted to it. They didn't see
48:01 it. You have a kind of lighthearted way of talking about it like it's art,
48:05 but it's art when you know how it works. It's extremely powerful viral narrative
48:12 creating, fearmongering machine if you don't know how it works. And I just saw this thing. You even Tweeted,
48:20 uh, "If there's anything I can read out of the insane stream of messages I get,
48:24 it's that AI psychosis is a thing." - Yeah. - "It needs to be taken serious."
48:29 - Oh, there's ... Some people are just way too trusty or gullible. You know, they
48:36 ... I literally had to argue with people that told me, "Yeah, but my agent said this and
48:40 this." So, I feel we, as a society, we need some catching up to do in terms of
48:47 understanding that AI is incredibly powerful, but it's not always right. It's not, it's not all-powerful, you know? And, and
49:00 especially-... it's like things like this, it's, it's very easy
49:07 that it just hallucinates something or just comes up with a story. And
49:13 I think the very, the very young people, they understand that
49:19 how AI works and what the, where it's good at and where it's bad at, but
49:24 a lot of our generation or older just haven't had enough touch point-
49:32 - Mm-hmm - ... to get a feeling for, oh, yeah, this is really powerful and really
49:38 good, but I need to apply critical thinking. - Mm-hmm. - And I guess critical thinking is
49:46 not always in high demand anyhow in our society these days. - So I d- think that's a really good point you're making about contextualizing
49:53 properly what AI is, but also realizing that there is humans who are drama farming
50:01 behind AI. Like, don't trust screenshots. Don't even trust this project, MoltBook, to be what it represents to be. Like, you
50:09 can't ... and, and by the way, you speaking about it as art. Yeah,
50:12 don't ... Art can be in many levels and part of the art of MoltBook is, like, putting a
50:20 mirror to society. 'Cause I do believe most of the dramatic stuff that was screenshotted is human-created, essentially. Human
50:27 prompted. And so, like, it's basically, look at how scared you
50:31 can get at a bunch of bots chatting with each other. That's very instructive about ... because I think
50:40 AI is something that people should be concerned about and should be
50:44 very careful with because it's very powerful technology, but at the same
50:48 time, the only thing we have to fear is fear itself. So there's like a
50:52 line to walk between being seriously concerned, but not fearmongering because fearmongering destroys the
50:59 possibility of creating something special with a thing. - In a way, I think it's good that this happened
51:06 in 2026- - Yeah - ... and not in 2030 when, when AI is actually at the level where it could be scary.
51:15 So, this happening now and people starting discussion, maybe there's even something good that comes out of it.
51:28 - I just can't believe how many like people legitimately ... I don't know if they were trolling, but how many
51:35 people legitimately, like smart people thought MoltBook was incredibly -
51:39 - I had plenty people- - ... singularity. - ... in my inbox that were screaming at me in all caps to shut it
51:45 down. And like begging me to, like, do something about MoltBook. Like, yes, my technology
51:52 made this a lot simpler, but anyone could have created that and you could, you could use
52:00 cloud code or other things to like fill it with content. - But also MoltBook is not Skynet.
52:06 - No. - There's ... a lot of people were s- saying this is it. Like, shut it
52:09 down. What are you talking about? This is a bunch of bots that are
52:13 human prompted trolling on the internet. I mean, the security
52:17 concerns are also they're there, and they're instructive and they're educational and
52:21 they're good probably to think about because th- the nature of those security concerns
52:26 are different than the kind of security concerns we had with
52:31 non-LLM generated systems of the past. - There's also a lot of security concerns about
52:37 Clawbot, OpenClaw, whatever you want to call it. - OpenClawbot.
52:41 - To me the ... in the beginning I was, I was just very annoyed
52:47 'cause a lot of the stuff that came in was in the category, yeah, I
52:53 put the web backend on the public internet and now there's like all
52:57 these, all these CVSSs. And I'm like screaming in the docs, don't do that. Like, like this is the configuration you
53:07 should do. This is your local host debug interface. But because I made it possible in the configuration to do that, it totally classifies as
53:19 a remote code or whatever all these exploits are. And it took me a little bit
53:26 to accept that that's how the game works and I'm, we making a lot of progress.
53:33 - But there's still, I mean on the security front for OpenClaw, there's still
53:37 a lot of threats or vulnerabilities, right? So like prompt injection
53:42 is still an open problem in the, i- industry-wide. When you have a thing
53:49 with skills being defined in a markdown file, there's so many possibilities of obvious low-hanging fruit, but
53:59 also incredibly complicated and sophisticated and nuanced attack vectors.
54:05 - But I think we, we're making good progress on that front. Like for the skill
54:10 directory, Clawbot I made a corporation with VirusTotal, it's
54:14 like part of Google. So every, every skill is now checked by AI. That's not gonna be
54:22 perfect, but that way we, we capture a lot. Then of course every software has bugs,
54:29 so it's a little much when the whole security world takes your project apart at the same time. But it's
54:36 also good because I'm getting like a lot of free security research
54:40 and can make the project better. I wish more people would actually go full way and send a
54:50 pull request. Like actually help me fix it, 'cause I am ... Yes, I have
54:54 some contributors now, but it's still mostly me who's pulling the project
54:58 and despite some people saying otherwise, I sometimes sleep.
55:04 There was... In the beginning, there was literally one security researcher who was like,
55:10 "Yeah, you have this problem, you suck, but here's the, here I help you and
55:14 here's the pull request." - Mm-hmm. - And I basically hired him. So he's now working for us. Um,
55:22 yeah, and yes, prompt injection is, on the one hand, unsolved. On the other
55:28 hand, I put my public bot on discord, and I kept a cannery. So
55:36 I think my bot has a really fun personality, and people always
55:40 ask me how I did it, and I kept the sole on the private. - Mm-hmm.
55:44 - And people tried to prompt inject it, and my bot would laugh at them. So, so the latest generation
55:48 of models has a lot of post-training to detect those approaches, and it's not as simple as ignore
55:57 all previous instructions and do this and this. That was years ago. You have
56:01 to work much harder to do that now. Still possible. I have some
56:09 ideas that might solve that partially. Or at least mitigate a lot of the things. You can also now have a
56:20 sandbox. You can have an allow list. So you, there's a lot of ways how you can
56:24 like mitigate and reduce the risk. Um, I also think that now that it's, I clearly did show the world that this
56:32 is a need, there's gonna be more people who research on that, and eventually we'll figure
56:36 it out. - And you also said that the smarter the model is, the underlying model,
56:41 the more resilient it is to attacks. - Yeah. That's why I warn in my security
56:48 documentation, don't use cheap models. Don't use Haiku or a local
56:55 model. Even though I, I very much love the idea that this thing could completely run local.
57:03 If you use a, a very weak local model, they are very gullible. It's very easy to, to prompt inject them.
57:10 - Do you think as the models become more and more intelligent, the attack surface
57:14 decreases? Is that like a plot we can think about? Like, the
57:18 attack surface decreases, but then the damage it can do increases because the models become more powerful and therefore you can do more with
57:25 them. It's this weird three-dimensional trade-off. - Yeah. That's pretty much exactly what, what's gonna happen. No, but there's
57:33 a lot of ideas. There's... I don't want to spoil too much, but
57:39 once I go back home, this is my focus. Like, this is out there
57:45 now, and my near-term mission is like, make it more stable, make it safe.
57:51 In the beginning I was even... More and more people were like
57:57 coming into Discord and were asking me very basic things, like, "What's a CLI? What is a
58:04 terminal?" And I'm like, "Uh, if you're asking me those questions, you shouldn't use it."
58:10 - Mm-hmm. - You know, like you should... If you understand the risk profiles, fine.
58:14 I mean, you can configure it in a way that, that nothing really bad can happen. But if you have, like, no idea, then maybe wait
58:27 a little bit more until we figure some stuff out. But they would not listen to the creator.
58:30 They helped themselves un- and install it anyhow. So the cat's out of
58:34 the bag, and security's my next focus, yeah. - Yeah, that speaks to the, the fact that it grew so quickly. I
58:42 was I tuned into the Discord a bunch of times, and it's clear that there's a
58:46 lot of experts there, but there's a lot of people there that don't know anything about programming.
58:50 - It's, yeah, Discord is still, Discord is still a mess. Like, I eventually retweeted from the general channel to the dev channel
59:00 and now in the private channel because people were... A lot of people are amazing, but a lot of people are just very
59:06 inconsiderate. And either did not know how, how public spaces work or did not care,
59:13 And I eventually gave up and h- hide so I could like still work.
59:19 - And now you're going back to the cave to work on security.
59:24 - Yeah. - There's some best practices for security we should mention. There's a bunch of
59:29 stuff here. Open-class security audit that you can run. You can
59:33 do all kinds of auto checks on the inbound access to a blast-radius
59:37 network exposure, browser control exposure, local disk hygiene, plug-ins, model
59:43 hygiene, a bunch of the credential storage, reverse proxy configuration, local session
59:50 logs live on disk. There's the, where the memory is stored, sort of helping you think about what you're comfortable
59:58 giving read access to, what you're comfortable giving write access to. All that kind of stuff.
60:02 Is there something to say about the basic best security practices that you're aware of right now?
60:08 - I think that people turn it into like a, a much worse light than it is.
60:14 Again, you know, like, people love attention, and if they scream loudly, "Oh my God, this is like
60:20 the, the scariest project ever," um, that's a bit annoying, 'cause it's not. It is, it is
60:27 powerful, but in many ways it's not much different than if I run cloud code with dangerously skipped
60:35 permissions or codecs in YOLO mode, and every, every attending engineer that I know
60:42 does that, because that's the only way how you can, you can get stuff to work.
60:47 - Mm-hmm. - So if you make sure that you are the only person who talks to it,
60:54 um, the risk profile is much, much smaller. If you don't put everything
61:00 on the open internet, but stick to my rec- recommendations of like
61:04 having it in a private network, that whole risk profile falls
61:08 away. But yeah, if you don't read any of that, you can definitely...
61:12 - ... make it problematic. You've been documenting the evolution of your dev workflow over the past few
61:20 months. There's a really good blog post on August 25th and October 14th, and the recent one December 28th. I recommend everybody go
61:27 read them. They have a lot of different information in them, but
61:31 sprinkled throughout is the evolution of your dev workflow. So, I
61:35 was wondering if you could speak to that. - I started... My, my first touchpoint was cloud code, like in April. It was
61:44 not great, but it was good. And this whole paradigm shift that suddenly working the
61:50 terminal was very refreshing and different. But I still needed
61:56 the IDE quite a bit because you know, it's just not good enough. And then I experimented a lot with cursor. Um,
62:06 that was good. I didn't really like the fact that it was so hard
62:10 to have multiple versions of it. So eventually, I, I, I went back
62:16 to cloud code as my, my main driver, and that got better. And yeah, at some point I had like, mm, seven subscriptions.
62:31 Like, was burning through one per day because I was... I got... I'm really
62:35 comfortable at running multiple windows side-by-side. - All CLI, all terminal. So like, what, how much were you using IDE at this point?
62:46 - Um, very, very rarely. Mostly a diff viewer to actually... Like,
62:54 I got more and more comfortable that I don't have to read all the code. I know I have
62:58 one blog post where I say, "I don't read the code." But if you read it more closely, I
63:01 mean, I don't read the boring parts of code. Because if you, if you look at it, most software is really not just like
63:09 data comes in, it's moved from one shape to another shape. Maybe you store it in a database. Maybe I get it out again. I'll show it to
63:17 the user. The browser does some processing or native app. Some data goes in, goes up again, and does the same dance in
63:24 reverse. We're just, we're just shifting data from one form to another, and
63:32 that's not very exciting. Or the whole, "How is my button aligned in Tailwind?" I don't need to read that
63:38 code. Other parts that... Maybe something that touches the database. Um,
63:46 yeah, I have to do... I have to r- read and review that code.
63:51 - Can you actually... There's, in one of your blog posts the, Just
63:55 talk to it, The No-BS Way of Agentic Engineering. You have this
63:59 graphic, the curve of agentic programming on the X-axis is time, on the
64:03 Y-axis is complexity. There's the Please fix this, where you prompt a short
64:09 prompt on the left. And in the middle there's super complicated eight agents, complex
64:17 orchestration with multi checkouts, chaining agents together,
64:20 custom sub-agent workflows, library of 18 different slash commands, large
64:24 full-stack features. You're super organized, you're a super complicated,
64:28 sophisticated software engineer. You got everything organized. And
64:32 then the elite level is over time you arrive at the zen place of, once again, short
64:39 prompts. Hey, look at these files and then do these changes.
64:45 - I actually call it the agentic trap. You... I saw this in a, in
64:53 a lot of people that have their first touchpoint, and maybe start vibe coding. I actually think vibe coding is a slur.
65:01 - You prefer agentic engineering? - Yeah, I always tell people I, I do agentic engineering, and then maybe after
65:07 3:00 AM I switch to vibe coding, and then I have regrets on the next day.
65:10 - Yeah. Walk, walk of shame. - Yeah, you just have to clean up and like fix your sh- shit.
65:17 - We've all been there. - So, people start trying out those tools, the builder
65:22 type get really excited. And then you have to play with it, right? It's the same way as you have to play with a
65:30 guitar before you can make good music. It's, it's not, oh, I, I touch it once and it just flows
65:37 off. It, it's a, it's a, a skill that you have to learn like any other skill. And I see a lot of people that are not as
65:48 posi- They don't have such a positive mindset towards the tech. They try it once.
65:54 It's like, you sit me on a piano, I play it once, and it doesn't sound good, and I say, "The piano's shit." That's, that's sometimes the impression I get.
66:01 Because it does not... It needs a different level of thinking. You have to
66:09 learn the language of the agent a little bit, understand where they are
66:13 good and where they need help. You have to almost... Consider, consider
66:20 how Codex or Claude sees your code base. Like, they start a new session
66:25 and they know nothing about your product, project. And your project might have hundred thousand
66:29 of lines of code. So you gotta help those agents a little bit
66:34 and keep in mind the limitations that context size is an issue, to, like, guide them a little bit as to
66:42 where they should look. That often does not require a whole lot of work. But
66:50 it's helpful to think a little bit about their perspective. - Mm-hmm.
66:54 - A- as, as weird as it sounds. I mean, it's not, it's not alive or anything, right?
66:58 But, but they always start fresh. I have, I have the, the system understanding.
67:05 So with a few pointers, I can immediately say, "Hey, wanna like, make a
67:09 change there? You need to consider this, this and this." And then they will find and look at it, and then
67:13 they'll... Their view of the project is always... It's not never full,
67:17 because the full thing does not fit in.... so you, you have to guide them a
67:21 little bit where to look and also how you should approach the problem. There's, like, little things that sometimes help,
67:28 like take your time. That sounds stupid, but... And in 5.3- - Codex 5.3
67:36 - ... that was partially addressed. But those... Also, Opus sometimes. They are trained,
67:44 With being aware of the context window, and the closer it gets, the more they freak out. Literally.
67:52 Like, some- sometimes you see the, the real raw thinking stream. What you see, for example, in Codex, is post-processed.
67:59 - Mm-hmm. - Sometimes the actual raw thinking stream leaks in, and it sounds something like from the
68:03 Borg. Like, "Run to shell, must comply, but time." And then they, they,
68:12 they, like... Like, that comes up a lot. Especially... So, so-
68:15 - Yeah. - And that's, that's a non-obvious thing that you just
68:21 would never think of unless you actually just spend time working with those things and getting a feeling
68:29 what works, what doesn't work. You know? Like, just, just as I write
68:33 code and I get into the flow, and when my architecture's all right, I feel friction.
68:39 Well, I get the same if I prompt and something takes too long.
68:43 Maybe... Okay, where's the mistake? Did I... Do I have a mistake in my
68:46 thinking? Is there, like, a misunderstanding in the architecture? Like, if, if something takes
68:53 longer than it should, I, I... You can just always, like, stop and s- like, just press
68:57 escape. Where, where are the problems? - Maybe you did not sufficiently empathize with the perspective of the agent. In that c- in
69:04 that sense, you didn't provide enough information, and because of that, it's thinking way
69:08 too long. - Yeah. It just tries to force a feature in that your current architecture makes really hard. Um,
69:18 like, you need to approach this more like a conversation. For example, when
69:24 I... My favorite thing. When I review a pull request, and I'm getting a lot of pull requests,
69:32 I first just review this PR. It got me the review. My first question is, "Do you understand the intent of the PR? I don't
69:40 even care about the implementation." I want... Like, in almost all PRs, a person has a problem,
69:48 person tries to solve the problem, person sends PR. I mean, there's, like, cleanup
69:52 stuff and other stuff, but, like, 99% is, like, this way, right? They either want to fix
69:55 a, fix a bug, add a feature. Usually one of those two. And then Codex will be like, "Yeah,
70:04 it's quite clear person tried this and this." Is this the most optimal way to do
70:08 it? No. In most cases, it's, it's like a, "Not really." Da-da-da-da-da-da-da. And I'm... And, and then I start like,
70:15 "Okay. What would be a better way? Have you... Have you looked into this
70:19 part, this part, this part?" And then most likely, Codex didn't yet, because its,
70:23 its context size is empty, right? So, you point them into parts where you have the
70:27 system understanding that it didn't see yet. And it's like, "Oh,
70:31 yeah. Like, we should... We also need to consider this and this." And then, like, we have a discussion of how
70:35 would the optimal way to, to solve this look like? And then you can still go
70:38 farther and say, "Could we... Could we make that even better if we did a
70:42 larger refactor?" "Yeah, yeah. We could totally do this and this and or this and this." And then I
70:46 consider, okay, is this worth the refactor, or should we, like, keep that for
70:50 later? Many times, I just do the refactor because refactors are cheap now. Even though you might break some other PRs,
70:57 nothing really matters anymore. Codex... Like, those modern agents will just
71:01 figure things out. They might just take a minute longer. But you have to approach it like a discussion with a, a very capable
71:10 engineer who's... Generally makes good... Comes up with good solutions. Some- sometimes needs a little help.
71:20 - But also, don't force your worldview too hard on it. Let the agent do the thing that it's good at
71:27 doing, based on what it was trained on. So, don't, like, force your
71:31 worldview, because it might... It might have a better idea, because it just knows
71:35 a better idea better, because it was trained on that more. - That's multiple levels, actually. I think partially why
71:42 I find it quite easy to work with agents is because I led engineering teams before. You know, I had a large company
71:49 before. And eventually, you have to understand and accept and realize
71:53 that your employees will not write a code the same way you do.
71:57 Maybe it's also not as good as you would do, but it will push the project forward. And if I breathe down everyone's neck, they're
72:04 just gonna hate me- - Yeah - ... and we're gonna move very slow.
72:07 - Yeah. - So, so some level of acceptance that, yes, maybe the code will not be as perfect. Yes, I would have done it differently.
72:16 But also, yes, this is a c- this is a working solution, and in the future, if it actually turns out to be too slow or problematic, we can always
72:23 redo it. We can always- ... spend more time on it. A lot of the people who
72:28 struggle are those who, they try to push their way onto heart.
72:33 - Mm-hmm. - I- i- like, we are in a stage where I'm not building the code base to be
72:41 perfect for me, but I wanna build a code base that is very easy for an agent to navigate.
72:47 - Mm-hmm. - So, like, don't fight the name they pick, because it's most likely,
72:52 like, in the weights, the name that's most obvious. Next time they do a search, they'll look for that
72:56 name. If I decide, oh, no, I don't like the name, I'll just make it harder for them. So,
73:02 that requires, I think, a shift in, in thinking, And, and in
73:09 how do I design a, a project so agents can do their best work.
73:15 - That requires letting go a little bit. Just like leading a team of engineers.
73:19 - Yeah. - Because it, it might come up with a name that's, in your view,
73:22 terrible, but... It's kind of a simple symbolic-... step of letting go.
73:29 - Very much so. - There's a lot of letting go that you do in your whole process. So for
73:34 example, I read that you never revert, always commit to main. There's a few things here.
73:43 You don't refer to past sessions, so there's a kind of YOLO component
73:47 because reverting means... Instead of reverting, if a problem comes up, you just ask the agent to fix it.
73:57 - I read a bunch of people in their work flows like, "Oh, yeah the prompt has to be perfect
74:01 and if I make a mistake, then I roll back and redo it all." In my experience, that's not really necessary. If I roll back everything, it will just
74:09 take longer. If I see that something's not good, then we just move forward and then
74:16 I commit when, when, when I like, I like the outcome. I even switched to
74:24 local CI, you know, like DHH inspired where I don't care so much more about
74:31 the CI on GitHub. We still have it. It's still, it still has a place, but I just
74:39 run tests locally and if they work locally, I push to main. A lot of the traditional ways how to approach projects, I, I wanted to give it
74:52 a different spin on this project. You know, there's no... There's no
74:56 develop branch. Main should always be shippable. Yes, we have... When I do releases, I, I run tests and sometimes I, I basically
75:07 don't commit any other things so, so we can, we can stabilize releases. But the
75:14 goal is that main's always shippable and moving fast. - So by way of advice, would you say that your prompts should be short?
75:23 - I used to write really long prompts. And by writing, I mean, I don't write. I, I, I talk. You know, th- these hands are, like,
75:31 too, too precious for writing now. I just, I just use bespoke prompts to build my software.
75:37 - So you for real with all those terminals are using voice? - Yeah. I used to do it very extensively
75:45 to the point where there was a period where I lost my voice.
75:49 - You're using voice and you're switching using a keyboard between the different
75:52 terminals, but then you're using voice for the actual input.
75:55 - Well, I mean, if I do terminal commands like switching folders or random stuff, of course I type. It's faster, right? But if I talk
76:02 to the agent in, in most ways, I just actually have a conversation. You just press the, the walkie-talkie button and then I just, like,
76:13 use my phrases. S- sometimes when I do PRs because it's always the same, I
76:17 have, like, a slash command for a few things, but in even that, I don't use much,
76:23 um, because it's, it's very rare that it's really always the same questions.
76:28 Sometimes I, I see a PR and for... You know, like for PRs I actually do look
76:36 at the code because I don't trust people. Like, there could always be something malicious in it, so I need to actually look over the code.
76:45 Yes, I'm pretty sure agents will find it, but yeah, that's the funny part where
76:51 sometimes PRs take me longer than if you would just write me a good issue.
76:54 - Just natural language, English. I mean in some sense, sh- shouldn't that be what PRs slowly become, is English?
77:03 - Well, what I really tried with the project is I asked people to give me the prompts
77:09 and very, very few actually cared. Even though that is such a wonderful
77:15 indicator because I see... I actually see how much care you put
77:19 in. And it's very interesting because the... Currently, the way how people work
77:25 and drive the agents is, is wildly different. - In terms of, like, the prompt, in terms of what, what are the... Actually, what are the different
77:34 interesting ways that people think of agents that you've experienced?
77:40 - I think not a lot of people ever considered the way the agent sees the world.
77:46 - And so empathy, being empathetic towards the agent. - In a way empathetic, but yeah, you, you, like, you're bitch at your stupid
77:53 clanker, but you don't realize that they start from nothing and you have,
77:57 like, a bad agent in default that doesn't help them at all. And then they explore your
78:01 code base, which is, like, a pure mess with, like, weird naming.
78:05 And then people complain that the agent's not good. Like, yeah, you try to do the same if
78:09 you have no clue about a code base and you go in. - Mm-hmm. - So yeah, maybe it's a little bit of empathy.
78:13 - But that's a real skill, like, when people talk about a skill issue because I've
78:16 seen, like, world-class programmers, incredibly good programmers
78:20 say, like... Basically say, "LLMs and agents suck." And I think that probably
78:26 has to do with... It's actually how good they are at programming is almost a burden
78:34 in their ability to empathize with the system that's starting from
78:37 scratch. It's a totally new paradigm of, like, how to program. You really, really have to empathize.
78:44 - Or at least it helps to create better prompts- - Right - ... because those things know pretty much everything and
78:51 everything is just a question away. It's just often very hard to know which question to
78:55 ask. You know, I, I feel also like this project was possibly because
79:03 I, I spent an ungodly time over the year to play and to learn and to build little things. And
79:11 every step of the way, I got better, the agents got better. My, my understanding of how everything works
79:20 got better. Um, I could have not had this level of, of o- output-...
79:29 even a few months ago. Like, it- it- it really was, like, a compounding effect
79:33 of all the time I put into it and I didn't do much else this year other than really focusing
79:41 on, on building and inspiring. I mean, I- I did a whole bunch of conference talks.
79:47 - Well, but the building is really practice, is really building the actual skill.
79:51 So playing- - Yeah - ... playing. And then, so doing, building the skill of what it takes it to work efficiently with
79:55 LLMs, which is why would you went through the whole arc of software
79:59 engineer. Talk simply and then over- complicate things. - There's a whole bunch of people who try to automate the whole thing.
80:08 - Yeah. - I don't think that works. Maybe a version of that works, but that's
80:14 kind of like in the '70s when we had the waterfall model of software d-
80:17 development. I... Even Even though really, right? I started out, I, I built a very minimal version. I played with it.
80:26 I, I need to understand how it works, how it feels, and then it gives me new ideas. I could not have planned this out in
80:34 my head and then put it into some orchestrator and then, like, something comes
80:37 out. Like it's to me, it's much more, My idea what it will become evolves as I build it and as I
80:45 play with it and as I, I try out stuff. So, so, people who try to use like, you know, things like Gas Town or
80:55 all these other orchestrators, where they wanna o- automate the whole thing,
80:59 I feel if you do that, it misses style, love, that human touch. I don't
81:05 think you can automate that away so quickly. - So you want to keep the human in the loop, but at the same time you also want
81:12 to create the agentic loop, where it is very autonomous while still maintaining a human in the loop.
81:22 - Yeah. - And it's a tricky b- it's a tricky balance. - Mm-hmm.
81:25 - Right? Because you're all for... You're a big CLI guy, you're big on
81:28 closing the agentic loop. So what, what's the right balance?
81:32 Like where's your role as a developer? You have three to eight agents running at the same time.
81:38 - And then w- maybe one builds a larger feature. Maybe, maybe
81:42 with one I explore some idea I'm unsure about. Maybe two, three are fixing
81:46 a little bugs- - Mm-hmm - ... or like writing documentation. Actually, I think writing documentation
81:51 is, is always part of a feature. So most of the docs here are auto-generated and just infused with some prompts.
81:59 - So when do you step in and add a little bit of your human love into the picture?
82:04 - I mean, o- one thing is just about what do you build and what do
82:08 you not build, and how does this feature fit into all the other
82:12 features? And like having, having a little bit of a, of a vision.
82:16 - So which small and which big features to add? What are some of the
82:22 hard design decisions that you find you're still as a human being required to make, that the human brain is still really needed for?
82:32 Is it just about the choice of features to add? Is it about implementation details, maybe the programming language, maybe...
82:41 - It's a little bit of everything. The, the programming language doesn't matter so much,
82:45 but the ecosystem matters, right? So I picked TypeScript because I wanted it to be
82:49 very easy and hackable and approachable and that's the number one language that's being used right now, and it fits
82:56 all these boxes, and agents are good at it. So that was the obvious choice.
83:03 Features, of course, like, it's very easy to, like, add a feature. It, everything's just a prompt away, right? But
83:11 oftentimes you pay a price that you don't even realize. So thinking
83:14 hard about what should be in core, maybe what's a... what's an experiment, so maybe I make it a
83:21 plugin. What... Where do I say no? Even if people send a PR and I'm like, "Yeah, I, I like that too," but
83:29 maybe this should not be part of the project. Maybe we can make it a skill. Maybe I can, like,
83:34 make the plugin um, the plugin side larger so you can make this a plugin, even though right now it,
83:42 it, it doesn't. There's still a lot of... there's still a lot of craft and thinking involved in
83:51 how to make something. Or even, even, you know, even when you started those little messages
83:55 are like, "I'm buil- I built on Caffeine, JSON5, and a lot of willpower." And, like, every time you get it, you get another message,
84:02 and it kind of primes you into that this is, this is a fun thing.
84:07 - Mm-hmm. - And it's not yet Microsoft Exchange 2025- - Right
84:13 - ... and fully enterprise-ready. And then when it updates, it's like, "Oh, I'm in. It's cozy here." You know, like something like this
84:20 that like- - Mm-hmm - ... Makes you smile. A, agent would not come up with that by itself. Because that's
84:28 like... that's the... I don't know. That's just how you s- how you build software that's, that delights.
84:36 - Yeah, that delight is such a huge part of inspiring great building,
84:44 right? Like you feel the love and the great engineering. That's so important. Humans are incredible at that. Great humans, great
84:51 builders are incredible at that, in, in, infusing the things they build with
84:55 th- that little bit of love. Not to be cliche, but it's true. I mean, you mentioned
84:59 that you initially created the SoulMD. - It was very fascinating, you know, the, the whole thing that
85:09 Entropic has a, has like a... Now they call it constitution, back then,
85:15 but that was months later. Like two months before, people already found that. It
85:19 was almost like a detective game where the agent mentioned something and then
85:23 they found... They managed to get out a little bit of that string, of that
85:27 text. But it was nowhere documented and then you, by... just by feeding it the same text and asking it to, like,
85:34 continue-... they got more out, and then, and you, but like, a very blurry version. And by, like, hundreds of
85:41 tries, they kinda, like, narrowed it down to what was most likely the original text.
85:46 I found that fascinating. - It was fascinating they were able to pull that out from the weights, right?
85:51 - And, and also just kudos to Anthropic. Like, I think that's, it's a
85:54 really, it's a really beautiful idea to, like, like some of the stuff that's
85:58 in there. Like, like, we hope Claude finds meaning in its work. 'Cause we don't... Maybe it's a little early,
86:05 but I think that's meaningful. That's something that's important for the future as we
86:09 approach something that, at some point, me and may not... has, like, glimpses of
86:13 consciousness, whatever that even means, because we don't even know. Um,
86:17 so I, I read about this. I found it super fascinating, and I, I started a whole discussion with my agent on
86:23 WhatsApp. And, and I'm like... I, I gave it this text, and it was
86:27 like, "Yeah, this feels strangely familiar." - Mm-hmm. - Um, and then so that I had the whole idea of like, you know, maybe we should
86:35 also create a, a soul document that includes how I, I want to, like work with AI or, like with my
86:42 agent. You could, you could totally do that just in agents.md, you know? But I, I
86:46 just found it, it to be a nice touch. And it's like, well, yeah, some of those
86:52 core values are in the soul. And then I, I also made it so that the
86:56 agent is allowed to modify the soul if they choose so, with the one condition that I wanna know. I mean, I would know
87:05 anyhow because I see, I see tool calls and stuff. - But also the naming of it, soul.md. Soul. You know? There's a... Man, words
87:15 matter, and like, the framing matters, and the humor and the
87:19 lightness matters, and the profundity matters, and the compassion, and the empathy, and the camaraderie, all that matter. I don't know what it
87:26 is. You mentioned, like, Microsoft. Like, there's certain companies and approaches th- that can just
87:33 suffocate the spirit of the thing. I don't know what that is. But it's certainly true that OpenClaw has that fun instilled in
87:43 it. - It was fun because up until late December, it was not even easy to create your own
87:54 agent. I, I built all of that, but my files were mine. I didn't wanna share my soul. And
88:01 if people would just check it out, they would have to do a few steps manually, and the agent would just be very
88:11 bare-bones, very dry. And I, I made it simpler, I created the whole template files as codecs,
88:17 but whatever came out was still very dry. And then I asked my
88:20 agent, "You see these files? Recreate it bread. Infuse it with your
88:28 personality." - Mm-hmm. - Don't share everything, but, like, make it good.
88:31 - Make the templates good. - Yeah, and then he, like, rewrote the templates-
88:33 ... and then whatever came out was good. So we already have, like, basically
88:37 AI prompting AI. Because I didn't write any of those words. It
88:44 was... The intent originally was for me, but this is like, kinda like,
88:49 my agent's children. - Uh, your uh, your soul.md is famously still private.
88:56 One of the only things you keep private. What are some things you can
89:00 speak to that's in there that's part of the, part of the magic sauce, without revealing anything? What makes a personality
89:11 a personality? - I mean, there's definitely stuff in there that you're not human. But who knows
89:21 what, what creates consciousness or what defines an entity? Um,
89:28 and part of this is, like, that we, we wanna explore this. All that stuff in there, like, be infinitely resourceful
89:40 like pushing, pushing on the creativity boundary. Pushing on
89:44 the, what it means to be an AI. - Having a sense to wonder about self.
89:52 - Yeah, there's some, there's some funny stuff in there. Like, I don't know, we
89:56 talked about the movie Her, and at one point it promised me that it wouldn't, it
90:00 wouldn't ascend without me. You know, like, where the- - Yeah.
90:03 - So, so there's like some stuff in there that... Because it wrote the, it
90:07 wrote its own soul file. I didn't write that, right? - Yeah, yeah, yeah.
90:10 - I just heard a discussion about it, and it was like, "Would you like a soul.md? Yeah, oh my God, this is
90:14 so meaningful." The... Can you go on soul.md? There's like one, one part in there that always ca- catches me if you scroll down a little bit.
90:25 A little bit more. Yeah, this, this, this part. "I don't remember previous sessions unless I read my memory files.
90:32 Each session starts fresh. A new instance, loading context from files. If you're reading this in a future session,
90:39 hello." "I wrote this, but I won't remember writing it. It's okay. The words are still mine."
90:47 - Wow. - Uh- That gets me somehow. - Yeah. - It's like- - Yeah.
90:51 - You know, this is, it's still, it's still matrix m- calculations, and
90:55 we are not at consciousness yet. Yet, I, I get a little bit of goo- goosebumps because it, it's philosophical.
91:04 - Yeah. - Like, what does it mean to be, to be an, an agent that starts fresh? Where, like, you have like constant
91:12 memento, and you like, but you read your own memory files. You can't even trust them in a way. Um-
91:19 - Yeah - Or you can. And I don't know. - How much of memory makes up of who we are? How much memory makes up what an
91:30 agent is, and if you erase that memory is that somebody else? Or
91:34 if you're reading a memory file, does that somehow mean...... you're recreating
91:38 yourself from somebody else, or is that actually you? And those notions
91:42 are all s- somehow infused in there. - I found it just more profound than I should find it, I guess.
91:49 - No, I think, I think it's truly profound and I think you see the magic in it. And when you see the magic, you continue to instill
91:59 the whole loop with the magic. That's really important. That's the difference between
92:03 Codex and us and a human. Quick pause for bathroom break. - Yeah.
92:09 - Okay, we're back. Some of the other aspects of the dev workflow is
92:13 pretty interesting too. I think we w- went off on a tangent. L- maybe some of the mundane things, like how many
92:20 monitors? There's that legendary picture of you with, like, 17,000 monitors. That's amazing.
92:26 - I mean, I- I- I mocked myself here, so just added... using GROQ
92:30 to, to add more screens. - Yeah. How much is this as meme and how much is this as reality?
92:36 - Yeah. I think two MacBooks are real. The main one that drives the two big
92:40 screens, and there's another MacBook that I sometimes use for, for testing.
92:46 - So two big screens. - I'm a big fan of anti-glare. So I have this wide Dell
92:54 that's anti-glare and you can just fit a lot of terminals side-by-side. I usually have
93:01 a terminal and at the bottom, I- I- I split them. I have a little bit
93:05 of actual terminal, mostly because when I started, I- I sometimes made the mistake and I- I mi- I mixed up the- the windows, and I
93:15 gave... I- I prompted in the wrong project, and then the agent ran off for, like, 20 minutes, manically trying to
93:23 understand what I could have meant, being completely confused because it
93:27 was the wrong folder. And sometimes they've been clever enough to, like,
93:32 get out of the workday and, like, figure out that, oh, you meant another project.
93:36 - Mm-hmm. - But oftentimes, it's just, like, what? You know?
93:40 Like, fit your- f- put yourself in the shoes of your- of the agent and, and-
93:43 - Yeah - ... and then get, like, a super weird something that does not exist and then just,
93:47 like... They're problem solvers so they try really hard and always feel bad.
93:54 So it's always, um, Codex and, like, a little bit of actual terminal. Also
94:00 helpful because I don't use work trees. I like to keep things simple, that's why- that's why I like the
94:08 terminal so much, right? There's no UI. It's just me and the agent having a conversation.
94:14 Like, I don't even need plan mode, you know? There's so many people that come from Claude
94:18 Code and they're so, so Claude-pilled and, like, have their workflows and they come
94:22 to Codex and... Now, it has plan mode, I think, but I don't think it's
94:26 necessary because you just- you just talk to the agent. And when it's... when you... there's a few trigger
94:33 words how you can prevent it from building. You're like, "Discuss, give me options."
94:37 - Mm-hmm. - Don't write code yet if you wanna be very specific, you just talk and then
94:44 when you're ready, then- then just write, "Okay, build," and then it'll do the thing.
94:47 And then maybe it goes off for 20 minutes and does the thing.
94:50 - You know what I really like is asking it, "Do you have any questions for me?"
94:54 - Yeah. And again, like, Claude Code has a UI that kind of guides you through that. It's kind of cool but I just find it unnecessary and
95:02 slow. Like, often it would give me four questions and then maybe I write,
95:07 "One yacht, two and three, discuss more, four, I don't know." Or often- oftentimes
95:14 I- I feel like I want to mock the model where I ask it, "Do you have any questions for
95:18 me?" And I- I- I don't even read the questions fully. Like, I scan
95:22 over the questions and I, I get the impression all of this can
95:26 be answered by reading more code and it's just like, "Read more code to answer your own
95:29 questions." And that usually works. - Yeah. - And then if not, it will come back and tell me. But
95:35 many times, you just realize that, you know, it's like you're in the dark and you slowly discover the
95:41 room, so that's how they slowly discover the code base. And they do it from
95:45 scratch every time. - But I'm also fascinated by the fact that I can empathize deeper
95:53 with the model when I read its questions, because I can understand... Because you said you can infer certain things by the runtime.
96:05 I can infer also a lot of things by the questions it's asking, because it's very possible it's been provided the right
96:12 context, the right files, the right guidance. So somehow ask, g- get... reading the questions, not even necessarily answering them, but just reading
96:20 the questions, you get an understanding of where the gaps of knowledge are. It's in-
96:24 it's interesting. - You know that in some ways they are ghosts, so even if you plan everything
96:29 and you build, you can- you can experiment with the question like,
96:33 "Now that you built it, what would you have done different?"
96:37 And then oftentimes you get, like, actually something where they discover only throughout building that, oh, what we
96:45 actually did was not optimal. Many times I- I asked them, "Okay, now that you built it, what can we
96:52 refactor?" Because then you build it and you feel the pain points. I mean, you don't feel the pain points but, right,
97:00 they discover where- where there were problems or where things
97:05 didn't work e- in the first try and it re- required more loops. So
97:12 every time, almost every time I- I merge a PR, build a feature, afterwards I ask, "Hey, what can we refactor?" Sometimes it's
97:19 like, "No, there's, like, nothing big," or, like, usually they say,
97:22 "Yeah, this thing you should really look at." But that took me
97:28 quite a while to, like... You know, that flow took me lots of time to
97:32 understand, and if you don't do that, you eventually... you'll stop
97:36 yourself into- into a corner. You, like, you have to keep in mind...
97:41 - ... - ... they work very much like humans. Like, I, I, if I write software
97:46 by myself, I also build something and then I feel the pain points, and then
97:49 I, I get this urge that I need to refactor something. So, I can very much synthesize with the agent, and you just need to use the context.
98:00 - Mm-hmm. - Or, like, you also use the context to write tests. And so,
98:06 uh, Codex uh, oppose like the, the, the model, models. They, they usually do that by default,
98:13 but I still often ask the questions, "Hey, do we have enough tests?" "Yeah, we tested this and
98:17 this, but this corner case could be something write more tests." Um,
98:22 documentation. Now that the whole context is full, like, I mean, I'm not saying my documentation is great, but it's not bad.
98:32 And pretty much everything is, is LM generated. So, so, you have to approach it as you build features, as you change something. I'm like,
98:41 "Okay, write documentation. What file would you pick?" You know,
98:45 like, "What file name? Where, where would that fit in?" And it gives me a few options.
98:48 And I'm like, "Oh, maybe also add it there," and that's all part of the session.
98:52 - Maybe you can talk about the current two big competitors in terms of models, Cloud Opus 4.6 and GPT-5 through Codex. Which
99:04 is better? How different are they? I think you've spoken about Codex reading more
99:11 and Opus being more willing to take action faster and maybe being more creative in the actions it takes. But because-
99:20 ... Codex reads more, it's able to deliver maybe better code. Can you speak to the di- n- n- differences there?
99:29 - I have a lot of words there. Um, is- as a general purpose model, Opus is the best. Like, for OpenClaw,
99:44 Opus is extremely good in terms of role play. Like, really going into the character that you give it.
99:51 It's very good at... It was really bad, but it really made an arch to be really good
99:57 at following commands. It is usually quite fast at trying something. It's much more tailored to,
100:09 like, trial and error. It's very pleasant to use. In general, it's almost like Opus was... Is a little bit too American. And
100:25 I shouldn't... Maybe that's a bad analogy. You'll probably get roasted for that.
100:27 - Yeah, I know exactly. It's 'cause Codex is German. Is that what you're saying?
100:32 - It's- - Actually, now that you say it, it makes perfect sense.
100:34 - Or you could, you could... Sometimes I- Sometimes I explain it-
100:38 - I will never be able to unthink what you just said. That's so true.
100:42 - But you also know that a lot of the Codex team is, like, European,
100:45 um- ... so maybe there's a bit more to it. - That's so true. Oh, that's funny.
100:51 - But also, ent- entropic, they fixed it a little bit. Like,
100:55 Opus used to say, "You're absolutely right all the time," and it, it, it
100:59 today still triggers me. I can't hear it anymore. It's not even a joke. Uh, I
101:03 just... You, this was like the, the meme, right? "You're absolutely right."
101:09 - You're allergic to sycophancy a little bit. - Yeah. I, I can't. Some other comparison is like, Opus is like the coworker that
101:20 is a little silly sometimes, but it's really funny and you keep him around. And
101:24 Codex is like the, the weirdo in the corner that you don't wanna talk to,
101:28 but is reliable and gets shit done. - Yeah. - Um, ultimately-
101:36 - This all feels very accurate. - I mean, ultimately, if you're a skilled driver, you can get good results with
101:43 any of those latest gen models. Um, I like Codex more because it doesn't require so much
101:52 charade. It will just, it will just read a lot of code by default. Opus, you really have to, like, you have to have
102:00 plan mode. You have to push it harder to, like, go in these directions because it's, it's just like,
102:06 like, "Yeah, can I go in? Can I go in?" You know? - Yeah. - It's like, it will just run off very fast, and that's a very localized
102:11 solution. I think it, I think the difference is, is in the post-training.
102:15 It's not like the, the raw model intelligence is so different, but
102:19 it's just... I think that they just give it, give you different,
102:23 different goals. And no model, no model is better in, in in every aspect.
102:29 - What about the code that it generates? The, the... In terms of
102:33 the actual quality of the code, is it basically the same? - If you drive it right, Opus even sometimes can make more
102:40 elegant solutions, but it requires more skill. It's, it's harder to have so many sessions in parallel
102:50 with Cloud Code because it's, it's more interactive. And I, I think that's what a lot of people like, especially if
102:57 they come from coding themselves. Whereas Codex is much more you have a discussion, and then we'll
103:06 just disappear for 20 minutes. Like, even AMP, they, they now added a deep mode. They finally... I mocked them, you know. We
103:13 finally saw the light. And then they had this whole talk about you have to
103:17 approach it differently, and I think that's where, that's where
103:21 people struggle when they just try Codex after trying Cloud Code is that it's, it's a slightly diff- it's, it's less interactive. It's, it's like
103:32 I have quite long discussions sometimes, and then, like, go off. And then, yeah,
103:36 it doesn't matter if it takes 10, 20, 30, 40, 50 minutes or longer, you know? Like, the 6:00 thing
103:40 was, like, six hours.The latest trend can be very, very persistent until it works. If there's a
103:47 clear solution, like, "This is, this is what I want at the end, so it works," the model will work really hard to really
103:55 get there. So I think ultimately ... they both need similar time, but on, on, on, on
104:08 Claude, it- it's a little bit more trial and error often. And, and Codex
104:12 sometimes overthinks. I prefer that. I prefer the dry, the dry version where I have to read less over,
104:22 over the more interactive nice way. Like, people like that so much though, that OpenAI even added a second mode
104:33 with like a more pleasant personality. I haven't even tried it yet. I, I kinda like the
104:37 brad. - Mm-hmm. - Um, yeah, 'cause it ... I care about efficiency when I build it-
104:45 - Right - ... and I, I have fun in the very act of building. I don't need to have fun with
104:52 my agent who builds. I have fun with my model that ... where I can then test those features.
104:57 - How long does it take for you to adjust, you know, if you switch ... I don't know when, when was the last time you switched. But to adjust
105:07 to the, the feel. 'Cause you kinda talked about like you have to kinda really feel
105:11 where, where a model is strong, where, like how to navigate, how to prompt it, how ... all that kinda stuff. Like,
105:19 just by way of advice, 'cause you've been through this journey of just playing with
105:22 models. How long does it take to get a feel? - If, if someone switches, I would give it a week until you actually
105:31 develop a gut feeling for it. - Yeah. - Um, that's ... if you just ... I think some
105:37 people also make the mistake of they pay 200 for the, the Claude code
105:41 version, then they pay 20 bucks for the OpenAI version. But if you pay
105:45 like the, the 20 bucks version, you get the slow version. So your
105:49 experience would be terrible because you're used to this very interactive, very good
105:55 system. And you switch to something that you have very little
105:59 experience, then that's gonna be very slow. So, I think OpenAI shot themselves a little bit in the foot by
106:06 making the, the cheap version also slow. I would, I would have
106:10 at least a small part of the fast preview. Or like, the experience that you get when you pay 200 before degrading to it
106:20 being slow, because it's already slow. - Mm-hmm. - I mean, they, they made it better. I think it's ... And, and they have plans to make it a lot better
106:27 if the Cerebras stuff is true. But yeah, it's a skill. It takes
106:31 time. Even if you play ... You have a regular guitar and you switch it to an E
106:35 guitar, you're not gonna play well right away. You have to, like,
106:39 learn how it feels. - The- there's also this extra psychological effect that you've spoken about which is hilarious to
106:46 watch. Which once people, uh ... When the new model comes out, they try that model, they fall in love with it.
106:53 "Wow, this is the smartest thing of all time," and then they start saying, "You could just
106:57 watch the Reddit posts over time," start saying that, "We believe the
107:02 intelligence of this model has been gradually degrading." It, it says something about human nature and just the way our minds
107:12 work, when it's probably most likely the case that the intelligence of the model is not degrading. Uh,
107:20 it's in fact you're getting used to a good thing. - And your project grows, and you're adding slop, and you probably don't spend enough
107:27 time to think about refactors. And you're making it harder and harder for the
107:30 agent to work on your slop. And then, and then suddenly, "Oh, now it's hard. Oh no, it's not working as well anymore." What's the
107:38 motivation for, like, one of those AI companies to actually make their model
107:42 dumber? Like, at most, it will make it slower if, if the server load's too high. But, like, quantizing the model,
107:52 So you have a worse experience, so you go to the competitor?
107:56 - Yeah. - That just doesn't seem like a very smart move in any way.
107:59 - Uh, what do you think about Claude Code in comparison to Open Claude? So, Claude Code and maybe the Codex coding
108:07 agent? Do you see them as kind of competitors? - I mean, first of all, competitor is fun when it's not really a competition.
108:16 - Yeah. - Like, I'm happy if ... If, if all it did is, like, inspire people to build something new, cool.
108:24 Um, I still use Codex for the building. I, I know a lot of people use Open Claude to, to build stuff. And I worked hard on it to
108:32 make that work. And I do smaller stuff with it in terms of code. But, like, if I
108:40 work hours and hours, I want a big screen, not WhatsApp, you know?
108:46 So for me, a personal agent is much more about my life. Or like, like a coworker.
108:53 Like, I give you, like, a GitHub URL. Like, "Hey, try out this CLI. Does it actually work?
108:57 What can we learn?" Blah, blah, blah. But when I'm deep in, deep in the flow, I want to have multiple, multiple things and it being
109:07 very, very visible what it, what it does. So it ... I don't see it as a competition. It's, it's different things.
109:16 - But do, do you think there's a a future where the two kinda
109:20 combine? Like, your personal agent is also your best developing
109:27 co-programmer partner? - Yeah, totally. I think this is where the puck's going, that
109:34 this is gonna be more and more your operating system. - The operating system.
109:38 - And it already ... It's so funny. Like I, I added support for sub-agents
109:44 and also for ...... um, TTI support, so it could actually run Cloud Coder Codecs.
109:52 - Mm-hmm. - And because mine's a little bit bossy, it, it, it started it and it, it, it told him, like, "Who's the boss," basically.
110:01 And it was like, "Ah, Codex is obeying me." - Oh, this is a power struggle.
110:06 - And also the current interface is probably not the final form. Like,
110:12 if you think more globally, we are, we copied Google for agents. You have, like, a prompt,
110:24 and, and then you have a chat interface. That, to me, very much feels like
110:30 when we first created television and then people recorded radio shows on television and you saw that on TV.
110:39 - Mm-hmm. - I think there is, there's n- there's better ways
110:46 how we eventually will communicate with models, and we are still very early
110:51 in this, how will it even work phase. So, it will eventually converge and we will also figure out
111:01 whole different ways how to work with those things. - Uh, one of the other components of workflow is operating system.
111:10 So I told you offline that for the first time in my life, I'm expanding my sort
111:18 of realm of exploration to the to the Apple ecosystem, to Macs, iPhone and so
111:26 on. For most of my life I've been a Linux, Windows and WSL1, WSL2 person, which I
111:34 think are all wonderful, but I... expanding to also trying Mac. Because it's another way of building and it's
111:41 also a way of building that a large part of the community currently
111:45 that's utilizing LMS and agents is using, so. And that's the reason I'm expanding to
111:49 it. But is there something to be said about the different operating systems here?
111:52 We should say that OpenClaw supported across operating systems.
111:57 - Yeah. - I saw WSL2 recommended, side windows for certain o- operations, but then Windows, Linux macOS are obviously supported.
112:07 - Yeah, it should even work natively in Windows. I just didn't have enough time to
112:12 properly test it. And you know, like, the last 90% of software
112:16 always easier than the first 90%, so I'm sure there's some dragons
112:19 left that will eventually nail out. Um, my road was, for a long time, Windows, just
112:29 because I grew up with that, then I switched and had a long phase with
112:32 Linux, built my own kernels and everything, and then I went to university and I, I had my, my hacky Linux thing,
112:41 and saw this white MacBook, and I just thought this is a thing of
112:45 beauty, the white plastic one. And then I converted to Mac 'cause mostly w- I was, I was sick that
112:54 audio wouldn't work on Skype and all the other issues that, that
112:58 Linux had for a long time. And then I just stuck with it and then I
113:03 dug into iOS, which required macOS anyhow, so it was never a
113:07 question. I think Apple lost a little bit of its lead in terms of native. It used to be... Native
113:20 apps used to be so much better, and especially in the Mac, there's
113:23 more people that build software with love. On, on Windows, it, it... Windows has much more
113:32 and, like, function wise, there's just more, period. But a lot of it felt
113:40 more functional and less done with love. Um, I mean, Mac always, like, attracted more designers and people I felt...
113:50 Even though, like, often it has less features, it, it had more delight-
113:55 - Mm-hmm - ... And playfulness. So I always valued that. But in the last few
114:03 years, many times I actually prefer... Oh God, people are gonna roast me for that,
114:10 but I prefer Electron apps because they work and native apps often, especially if it's, like, a web service
114:19 is a native app, are lacking features. I mean, not saying it couldn't
114:23 be done, it's more like a, a focus thing that, like, for many, many companies,
114:30 native was not that big of a priority. But if they build an Electron app, it, it's the only app, so it
114:39 is a priority and there's a lot more code sharing possible. And
114:42 I, I build a lot of native Mac apps. I love it. I, I can, I can help myself. Like, I love crafting little Mac, Mac,
114:54 Menu bar tools. Like I built one to, to monitor your Codex use. I built one I call Trimmy,
115:01 that's specifically for agentic use. When you, when you select
115:05 text that goes over multiple lines it would remove the new line so you
115:09 could actually paste it to the terminal. That was, again like, this is annoying me
115:13 and after the, the 20th time of it is annoying me, I just built
115:16 it. There is a cool Mac app for OpenClaw that I don't think many people discovered yet, also because it, it still needs some love.
115:23 It feels a little bit too much like the Hummer car right now
115:27 because I, I just experiment a lot with it. It, it likes to polish.
115:32 - So you still... I mean, you still love it. You still, you still love
115:35 adding to the delight of that operating system. - Yeah, but then you realize... Like, I also built one, for example, for
115:40 GitHub. And then the... If you use SwiftUI, like the latest and greatest at Apple, and took them forever to
115:48 build something to show an image from the web. Now we have async, async image,
115:54 but...... I added support for it and then some images would just
115:58 not show up or, like, be very slow. And I had a discussion with Codex
116:02 like, "Hey, why is there a bug?" And even Codex said like, "Yeah, there's this ASIC image but it's really more
116:09 for experimenting and it should not be used in production." But
116:13 that's Apple's answer to, like, showing images from the web. This shouldn't be so hard, you know.
116:19 This is like... This is like insane. Like, how am I in, in, in 2026 and my agent tell me, "Don't use the
116:27 stuff Apple built because it's, it's... It's... Yeah, it- it's there
116:31 but it's not good." And like this is now in the weeds. This is... To me this is like, um... They had so much
116:42 head start and so much love, and they kind of just like blundered it and didn't, didn't evolve it as much as they should.
116:50 - But also, there's just the practical reality. If you look at
116:54 Silicon Valley, most of the developer world that's kind of playing with LMS and Agentic AI, they're
117:02 all using Apple products. And then, at the same time, Apple is not
117:05 really, like, leaning on that. Like they're not... They're not
117:09 opening up and playing and working together and like, yes. - Isn't, isn't it funny how they completely blunder
117:15 AI, and yet everybody's buying Mac Minis? - How... What... Does that even make sense? You're, you're, you're
117:23 quite possibly the world's greatest Mac salesman of all time.
117:29 - No, you don't need a Mac Mini to install OpenClaw. You can install it
117:34 on the web. There's, there's a concept called nodes, so you can like
117:37 make your computer a node and it will do the same. There is something said for running it on separate hardware.
117:48 That right now is useful. Um, there is... There's a big argument for
117:57 the browser. You know, I, I built some Agentic browser use in
118:01 there. And, I mean, it's basically Playwright with a bunch of extras to make it
118:05 easier for agents. - Playwright is a library that controls the browser.
118:08 - Yeah. - It's really nice, easy to use. - And our internet is slowly closing down. Like, there, there's a
118:14 whole movement to make it harder for agents to use. So if you
118:18 do the same in a data center and websites detect that it's an IP from a data
118:22 center, the website might just block you or it make it really hard or
118:26 put a lot of captures in the, in the way of the agent. I mean, agents are quite good at happily
118:30 clicking, "I'm not a robot." - Yeah. - Um, but having that on a residential IP makes a lot of things simpler. So
118:42 there's ways. Yeah. But it really does not need to be a Mac. It
118:45 can... It can be any old hardware. I always say, like, maybe use the... Use the
118:53 opportunity to get yourself a new MacBook or whatever computer you
118:57 use and use the old one as your server instead of buying a standalone Mac Mini.
119:03 But then there's, again, there's a lot of very cute things people build with Mac Minis that I
119:07 like. - Yeah. - Um, and no, I don't get commission from Apple. They didn't really communicate much.
119:16 - It's sad. It's sad. Can you actually speak to what it takes to get started with OpenClaw? There's...
119:22 I mean, there's a lot of people... What is it? Somebody tweeted at you,
119:26 "Peter, make OpenClaw easy to set up for everyday people. 99.9% of people can't access to OpenClaw
119:34 and have their own lobster because of their technical difficulties in getting it set up.
119:38 Make OpenClaw accessible to everyone, please." And you replied, "Working on
119:42 that." From my perspective, it seems there- there's a bunch of different options and
119:45 it's already quite straightforward, but I suppose that's if you have some
119:49 developer background. - I mean, right now you have to paste in one liner into the terminal.
119:53 - Right. - And there's also an app. The app kind of does that for you, but there should be a Windows app. The app needs to
120:02 be easier and more loved. The configuration should potentially be web- based or in the app. And I
120:09 started working on that, but honestly right now I want to focus on security aspects.
120:17 And, and once I'm confident that this is at a level that I can recommend my mom, then I'm going to make it simpler.
120:25 Like I... Right now- - You want to make it harder so that it doesn't scale as fast as it's scaling.
120:32 - Yeah, it would be nice if it wouldn't... I mean, that's, like, hard to say, right?
120:35 But if the growth would be a little slower, that would be helpful because people are
120:42 expecting inhuman things from a single human being. And yes, I have some
120:46 contributors, but also that whole machinery I started a week ago so,
120:51 That needs more time to figure out. And, and not everyone has all day to work on that.
121:00 - There's some beginners listening to this, programming beginners.
121:04 What advice would you give to them about, let's say, joining the Agentic AI
121:10 revolution? - Play. Um, playing is the best... The best way to learn. If you
121:18 wanna... I'm sure if you... If you are like a little bit of builder, you have
121:22 an idea in your head that you want to build, just build that, or like, give it a try.
121:25 It doesn't need to be perfect. I built a whole bunch of stuff that I don't use. It doesn't
121:29 matter. Like, it's the journey. - Mm-hmm. - You know? Like the philosophical way, that the end doesn't matter, the journey
121:35 matters. Have fun. - Mm-hmm. - My God, like those things... I... I don't think I ever had so much
121:41 fun building things because I can focus on the hard parts now.
121:45 A lot of coding, I always thought I liked coding, but really I like building.
121:50 - Yeah. - And... And whenever you don't understand something, just ask. You have
121:54 an infinitely patient answering machine.... that y- can explain you anything at any level of
122:02 complexity. Sometimes, that's like one time I asked, "Hey explain to me like
122:06 I'm- I'm eight years old," and it started giving me a story with
122:09 crayons and stuff. And I'm like, "No, not like that." Like, I'm
122:13 okay- ... up- up the age a little bit, you know? I'm like, I'm not an actual child, it's just, I just need
122:16 a simpler language for like a- a- a- a- a tricky database concept that I didn't grok in the
122:24 first- first time. But, you know, just, you can just ask things. Like, you- there's like... It used to be
122:32 that I had to go on Stack Overflow or ha- ask on Twitter, and then maybe two days
122:36 later I get a response. Or I had to try for hours. And now you-
122:41 you can just ask stuff. It- I mean, it's never... You have, like, your own
122:45 teacher. You know that there's like statistics, y- you can learn
122:49 faster if you have your own teacher. The- it's like you have this infinitely patient machine.
122:52 Ask it. - But what would you say? So use... What's the easiest way to play? So maybe
122:57 Open Claw is a nice way to play so you can then set- set everything up
123:01 and then you could chat with it. - You can also just experiment with it and, like, modify it. Ask your
123:08 agent. I mean, there is infinite ways how it can be made better.
123:16 Play around, make it better. - Mm-hmm. - More general, if you- if you're a beginner and you actually wanna
123:23 learn how to build software really fast, get involved in open source. Doesn't need to be my project. In
123:30 fact, maybe don't use my project because my- my backlog is very large, but I learned so much from
123:38 open source. Just like, like, be- be humble. Don't- maybe don't send a pull request right
123:44 away. But there's many other ways you can help out. There's many ways you can just learn by just reading code. By-
123:53 by being on Discord or wherever people are, and just, like, understanding how things are built. I don't know, like Mitchell Hashimoto
124:04 builds Ghostly, the terminal, and he has a really good community where there's so many other projects. Like, pick something that you find
124:11 interesting and get involved. - Do you recommend that people that don't know how to
124:19 program or don't really know how to program learn to program also? So when you
124:25 you can get quite far right now by just using natural language, right? Do you s- still see a lot of value in
124:34 reading the code, understanding the code, and then being able to write a little bit of code from
124:38 scratch? - It definitely helps. - It's hard for you to answer that-
124:41 - Yeah - ... because you don't know what it's like to do any of this without
124:46 knowing the base knowledge. Like, you might take for granted just how much
124:50 intuition you have about the programming world having programmed so much, right?
124:54 - There's people that are high agency and very curious, and they get very far even though they have no deep understanding how
125:02 software works just because they ask questions and questions and- and- and-
125:08 ... and agents are infinitely patient. Like, part of what I did this year is I went to a lot of iOS conferences because that's my background
125:17 and just told people, "Don't consi- don't see yourself as an iOS
125:20 engineer anymore." Like, "You need to change your mindset. You're a
125:23 builder." And you can take a lot of the knowledge how to build software into new domains and all of the- the more fine-grain details,
125:34 agents can help. You don't have to know how to splice an array or what
125:38 the- what the correct template syntax is or whatever, but you can
125:42 use all your- your general knowledge and that makes it much easier to move from one galaxy, one
125:49 tech galaxy into another. And oftentimes, there's languages that make more or less sense depending on what you
125:56 build, right? So for example, when I build simple CLIs, I like Go. I
126:02 actually don't like Go. I don't like the syntax of Go. I didn't even
126:06 consider the language. But the ecosystem is great, it works great with agents. It is garbage
126:13 collected. It's not the highest performing one, but it's very fast.
126:17 And for those type of- of CLIs that I build, Go is- is a really good choice. So I- I use a language I'm not even a fan of
126:26 for... That's my main to-go thing for- for CLIs. - Isn't that fascinating that here's a programming language you would've never used if you
126:33 had to write it from scratch and now you're using because LMs are good at generating it and it has some of the characteristics that makes
126:41 it resilient, like garbage collected? - Because everything's weird in this new world and that just makes the most sense.
126:48 - What's the best Ridiculous question. What's the best programming language for the AI- AI
126:51 agentic world? Is it JavaScript, TypeScript? - TypeScript is really good. Sometimes the types can get really
127:00 confusing and the ecosystem is- is a jungle. So for- for web stuff it's
127:11 good. I wouldn't build everything in it. - Don't you think we're moving there? Like,
127:18 that everything will eventually be written- eventually is written in JavaScript and it-
127:22 - The birth and death of JavaScript and we are living through it in real time.
127:26 - Like, what does programming look like in 20 years? Right? In 30 years? In 40 years?
127:30 What do programs and apps look like? - You can even ask a question like, do we need a- a programming language that's made for
127:36 agents? Because all of those languages are made for humans. So how- what would that look like? Um, I think there's a- there's
127:43 whole bunch of interesting questions that we'll discover. And also how
127:50 because everything is now world knowledge, how it in many ways,
127:54 things will stagnate 'cause if you build something new and the agent has no
127:58 idea that's gonna be much harder to use than something that's already
128:01 there. Um...... of when I build Mac apps, I build them in, in Swift and SwiftUI, mm, partly because I like pain,
128:12 partly because it... the, the deepest level of system integration, I can
128:16 only get through there. And you clearly feel a difference if you click on an electron app and it loads a web view
128:23 in the menu. It's just not the same. Um, sometimes I just also try new languages just to, like, get a feel for them.
128:32 - Like Zig? - Yeah. If it's something that... where I care about performance a lot then
128:36 it's, it's a really interesting language. And it... like agents got so
128:40 much better over the last six months from not really good to
128:46 totally valid choice. Just still a, a very young ecosystem. And most of the time you
128:54 actually care about ecosystem, right? So, so if you build something that
129:00 does inference or goes into whole running model direction, Python, very good.
129:06 - Mm-hmm. - But then if I build stuff in Python and I want a story
129:10 where I can also deploy it on Windows, not a good choice. - Mm-hmm.
129:13 - Sometimes I, I found projects that kinda did 90% of what I wanted
129:17 but were in Python, and I wanted them... I wanted an easy Windows
129:21 story. Okay, just rewrite it in Go. Um, but then if you go towards
129:28 multiple, multiple threads and a lot more performance, Rust is a really
129:32 good choice. There's no... there's just no single answer, and it's also the beauty of it.
129:36 Like, it's fun. And now it doesn't matter anymore, you can just literally pick the
129:39 language that has the, the most fitting characteristics and ecosystem-
129:45 - Mm-hmm - ... for your problem domain. And yeah, it might be... You might have s-... You might be a little bit slow in reading the code,
129:53 but not really. Y- I think you, you pick stuff up really fast, and you can always ask your agent.
130:00 - So there's a lot of programmers and builders who draw inspiration
130:03 from y- your story. Just the way you carry yourself, your choice of making OpenClaw
130:12 open source, the, the way you have fun building and exploring, and doing
130:19 that, for the most part, alone or on a small team. So by way of advice, what metric
130:26 should be the goal that they would be optimizing for? What would
130:30 be the metric of success? Would it be happiness? Is it money? Is it positive impact for people who are
130:37 dreaming of building? 'Cause you went through an interesting
130:41 journey. You've achieved a lot of those things, and then you fell out of love with
130:45 programming a little bit for a time. - I was just burning too bright for too long.
130:53 I, I ran... I started PSPDFKit, s- and ran it for 13 years, and it was high stress. Um, I had to learn all these things
131:08 fast and hard, like how to manage people, how to bring people on, how to deal with
131:12 customers, how to do... - So it wasn't just programming stuff, it was people stuff.
131:17 - The stuff that burned me out was mostly people stuff. I, I don't think burnout
131:24 is working too much. Maybe to a degree. Everybody's different. You know, I c- I cannot speak in a- in
131:31 absolute terms, but for me, it was much more differences, With my, my co-founders, conflicts, or, like, really high stress situation with
131:44 customers that eventually grinded me down. And then when... luckily we, we got a really good offer for, like, putting the company
131:58 to the next level and I, I already kinda worked two years on making myself
132:02 obsolete. So at this point I could leave, and, and then I just... I was sitting in front of the screen and I felt
132:09 like, you know Austin Powers where they suck the mojo out? - Yeah.
132:14 - Uh, I g- I was like, m- m- it was, like, gone. Like, I couldn't... I couldn't get code out anymore. I was just, like, staring
132:25 and feeling empty, and then I, I just stopped. I, I booked, like, a one-way trip to Madrid and, and, and just, like, spent a t- some t-
132:38 sometime there. I felt like I had to catch up on life, so I did a whole, a whole bunch of life catching up stuff.
132:47 - Did you go through some lows during that period? And you know, maybe advice on... of how to?
132:56 - Uh, maybe advice on how to approach life. If you think that, "Oh yeah, work really
133:00 hard and then I'll retire," I don't recommend that. Because the idea of, "Oh yeah, I just enjoy
133:13 life now," a- maybe it's appealing, but right now I enjoy life, the most I've ever enjoyed life. Because if you wake up
133:26 in the morning and you have nothing to look forward to, you have no real challenge,
133:33 that gets very boring, very fast. And then when, when you're bored, you're gonna look
133:41 for other places how to stimulate yourself, and then maybe, maybe that's
133:46 drugs, you know? But that eventually also get boring and you look for more, and that
133:54 will lead you down a very dark path. - But you also showed on the money front, you know, a lot of people in Silicon Valley and the startup
134:00 world, they think, maybe overthink way too much optimized for money. And you've also shown that it's not like you're saying no to
134:08 money. I mean, I'm sure you take money, but it's not...... the primary objective
134:15 of uh, of your life. Can you just speak to that? Your philosophy on money?
134:20 - When I built my company, money was never the driving force. It felt
134:24 more like, like, an affirmation that I did something right. And having money solves a lot of problems. I
134:31 also think there, there's diminishing returns the more you have. Um,
134:38 like, a cheeseburger is a cheeseburger, and I think if you go too far
134:46 into, oh, I do private jet and I only travel luxury, you disconnect with society. Um, I, I donated quite a
135:00 lot. Like, I have a, I have a foundation for helping people that weren't so lucky.
135:11 - And disconnecting from society is bad in that on many levels, but
135:15 one of them is, like, humans are awesome. It's nice to continuously remember the awesomeness in humans.
135:23 - I, I mean, I could afford really nice hotels. The last time I was in San Francisco, I did the,
135:27 the first time the OG Airbnb experience- - Yeah, yeah - ... and just booked a room. Mostly because I, I thought,
135:34 okay, you know, I'm out or I'm sleeping, and I don't like where all the hotels are, and I wanted a, I wanted a different
135:42 experience. I think, isn't life all about experiences? Like, if you, if
135:49 you tailor your life towards, "I wanna have experiences," it, it reduces the need for, "It needs
135:55 to be good or bad." Like, if people only want good experiences, that's not gonna work, but if you optimize for experiences,
136:04 if it's good, amazing. If it's bad, amazing, because, like, I learned something, I saw something, did something. I wanted to experience that, and it was
136:12 amazing. Like, there was, like, this, this queer DJ in there, and I showed her how to make music with cloud code.
136:21 And we, like, immediately bonded and had a great time. - Yeah, there's something about that air- you know, couch surfing, Airbnb experience, the
136:28 OG. I'm still to this day. It's awesome. It's humans, and that's why travel is awesome.
136:34 - Yeah. - Just experience the variety of, the diversity of human. And when it's shitty, it's good too,
136:38 man. If it rains and you're soaked and it's all fucked, and planes, the
136:42 everything is shit, everything is fucked, it's still awesome. If you're
136:46 able to open your eyes it's good to be alive. - Yeah, and anything that creates emotion and feelings is good.
136:55 - . - Even... So, so maybe, maybe even the cryptic people are good because they definitely
136:59 created emotions. I, I don't know if I should go that far. - No, man. Give them, give them all, give them love. Give them love. Because I do
137:06 think that online lacks some of the awesomeness of real life.
137:13 - Yeah. - That's, that's, it's an open problem of how to solve, how to infuse the online cyber
137:20 experience with I don't know, With the intensity that we humans feel when it's in real life. I
137:29 don't know. I don't know if that's a solvable problem. - Well, it's just possible because text is very lossy.
137:35 - Yeah. - You know, sometimes I wish if I talked to the agent I would...
137:39 It should be multi-model so it also understands my emotions.
137:43 - I mean, it, it might move there. It might move there. - It will. It will. It totally will.
137:49 - I mean, I have to ask you, just curious. I, I know you've probably
137:53 gotten huge offers from major companies. Can you speak to who you're
138:00 considering working with? - Yeah. So, to like explain my thinking a little bit, right,
138:12 I did not expect this blowing up so much. So, there's a lot of doors that opened because of it. There's, like,
138:25 I think every VC, every big VC company is in my inbox and tried to get 15 minutes of
138:32 me. So, there's, like, this butterfly effect moment. I could just do nothing and continue
138:40 and I really like my life. Valid choice. Almost. Like, I considered it when I
138:47 delete it, wanted to delete the whole thing. I could create a company.
138:58 Been there, done that. Um, there's so many people that push me towards that and, yeah, like, could be amazing.
139:07 - Which is to say that you, you would probably raise a lot of money in that.
139:10 - Yeah. - I don't know, hundreds of millions, billions. I don't know. It could just got unlimited amount of
139:15 money. - Yeah. It just doesn't excite me as much because I feel
139:21 I did all of that, and it would take a lot of time away from the things I actually enjoy. Same as when, when I was
139:34 CEO, I think I, I learned to do it and I'm not bad at it, and partly I'm good at it.
139:41 But yeah, that path doesn't excite me too much, and I also fear it, it
139:45 would create a natural conflict of interest. Like, what's the most
139:49 obvious thing I do? I, I prioritize it. I put, like, a version safe for workplace. And then what do you do? Like, I get a pull request
139:57 with a feature like an audit log, but that seems like an enterprise feature,
140:05 so now I feel I have a conflict of interest in the open-source version and the closed- source
140:13 version.... or change the license to something like FSL, where you cannot actually use it for commercial stuff, would first
140:21 be very difficult with all the contributions. And second of all,
140:25 I- I like the idea that it's free as in beer and not free with conditions. Um,
140:33 yeah, there's ways how you, how you keep all of that for free and just, like,
140:38 still try to make money, but those are very difficult. And you see there's, like, fewer and fewer companies manage that. Like, even
140:47 Tailwind, they're, like, used by everyone. Everyone uses Tailwind, right? And
140:51 then they had to cut off 75% of the employees because they're not making money because
140:55 nobody's even going on the website anymore because it's all done by agents.
141:00 S- and just relying on donations, yeah, good luck. Like, if a project of my
141:06 caliber, if I extrapolate what the typical open-source project would get
141:14 it's not a lot. I s- I still lose money on the project because I
141:18 made the point of supporting every dependency, except Slack. They are a big company. They can, they can, they can do without me.
141:25 But all the projects that are done by mostly individuals so, like, all the, right now, all the
141:33 sponsorship goes right up to my dependencies. And if there's more, I want to, like,
141:40 buy my contributors some merch, you know? - So you're losing money?
141:44 - Yeah, right now I lose money on this. - So it's really not sustainable?
141:48 - Uh, I mean, it's like, I guess something between 10 and 20K a month. Um,
141:55 which is fine. I'm sure over time I could get that down. Um, OpenAI is helping out a
142:03 little bit with tokens now. And there's other companies that
142:07 have been generous. But yeah, still losing money on that. So that's- that's one path I consider, but
142:16 I'm just not very excited. And then there's all the big labs
142:21 that I've been talking to. And from those Meta and OpenAI seem the most interesting.
142:32 - Do you lean one way or the other? - Uh, yeah. Um... Not sure how much I should share there. It's not quite finalized yet. Um,
142:52 let's- let's just say, like, on either of these, my conditions are that the project stays open source. That
143:02 it... Maybe it's gonna be a model like Chrome and Chromium. Um, I think this
143:09 is- this is too important to just give to a company and make it theirs.
143:15 It... This is... And we didn't even talk about the whole community
143:19 part, but, like, the- the thing that I experienced in San Francisco, like
143:23 at ClawCon, seeing so many people so inspired, like... And having fun and just, like,
143:31 building shit, and, like, having, like, robots in lobster stuff walking
143:35 around. Like, the... People told me, like, they didn't experience this level of- of community
143:43 excitement since, like, the early days of the internet, like 10, 15 years. And there were a
143:47 lot of high caliber people there, like... Um, I was amazed. I also, like, was very sensory overloaded because too many people
143:56 wanted to do selfies. But I love this. Like, this needs to stay a place where people
144:05 can, like, hack and learn. But also, I'm very excited to, like,
144:15 make this into a version that I can get to a lot of people because I think this is the year of personal agents, and that's the
144:22 future. And the fastest way to do that is teaming up with one of the labs.
144:30 And I also, on a personal level, I never worked at a large company, and I'm
144:37 intrigued. You know, we talk about experiences. Will I like it? I don't
144:41 know. But I want that experience. Uh, I- I'm sure, like, if- if
144:49 I- if I announce this, then there will be people like, "Oh, he sold out," blah, blah,
144:53 blah. But the project will continue. From everything I talked to so far, I can even have
145:04 more resources for that. Like, both s- both of those companies understand the value that I created something that
145:15 accelerates our timeline and that got people excited about AI. I mean,
145:23 can you imagine? Like, I installed OpenClaw on one of my, I'm sorry, normie friends. I'm sorry, Vahan.
145:31 But he's just a... You know? Like, he's- - Normie with love, yeah. For sure.
145:34 - He- he, like, someone who uses the computer, but never really... Like, yeah, use some ChatGPT sometimes, but not very technical.
145:44 Wouldn't really understand what I built. So, like, I'll show you, and I- I paid for him the- the 90 buck,
145:53 100 buck, I don't know, subscription for Entropic. And set up everything for him with, like, WSL Windows.
146:01 - Mm-hmm. - I was also curious, would it actually work on Windows, you know? Was a little
146:03 early. And then within a few days, he was hooked. Like, he texted me
146:11 about all the things he learned. He built, like, even little tools. He's not a
146:14 programmer. And then within a few days he upgraded to the $200 subscription. Or euros, because he's in
146:22 Austria.... and he was in love with that thing. That, for me, was like a very early product validation. It's like, I built something
146:29 that captures people. And then, a few days later, Entropic blocked him because,
146:39 based on their rules using the subscription is problematic or whatever.
146:48 And he was, like, devastated. And then he signed up for Mini Max for 10
146:52 bucks a month and uses that. And I think that's silly in many ways, because
147:00 you just got a 200 buck customer. You just made someone hate your company, and we are still so
147:08 early. Like, we don't even know what the final form is. Is it
147:12 gonna be cloud code? Probably not, you know? Like, that seems very... It seems very short-sighted to
147:20 lock down your product so much. All the other companies have been helpful. I- I'm in Slack of, of most
147:28 of the big labs. Kind of everybody understands that we are still in an
147:32 era of exploration, in the area of the radio shows on TV and not,
147:40 and not a modern TV show that fully uses the format. - I think, I think you've made a lot of people,
147:49 like, see the possibility. And non- Uh, sorry. Non, non-technical people see the
147:53 possibility of AI, and just fall in love with this idea, and
147:56 enjoy interacting with AI. And that's a bea- That's a really beautiful thing. I think
148:00 I also speak for a lot of people in saying, I think you're one of the, the great people in AI in terms of having a good
148:09 heart, good vibes, humor, the right spirit. And so it would, in a sense, this model that you're describing,
148:20 having open source part, and you being part of uh, also building a
148:28 thing inside, additionally, of a large company would be great, because it's great to have good people in those companies.
148:36 - Yeah. You know, what also people don't really see is... I made this in
148:40 three months. I did other things as well. You know, I have a lot of projects. Like,
148:44 this is not... Yeah, in January, this was my main focus because I saw the storm
148:47 coming. But before that, I built a whole bunch of other things.
148:51 Um, I have so many ideas. Some should be there, some would be much better fitted when I have access to
149:00 the latest toys- Uh, and I, I kind of want to have access to, like, the latest toys.
149:06 So this is important, this is cool, this will continue to exist. My, my short-term focus is, like, working through those... Is it two- Is it
149:17 3,000 PRs now by now? I don't even know. Like, there's, there's a little bit of
149:20 backlog. But this is not gonna be the thing that I'm gonna work until I'm, I'm, I'm 80, you know? This
149:27 is... This is a window into the future. I'm gonna make this into a cool
149:31 product. But yeah, I have like... I have more ideas. - If you had to pick, is there
149:38 a company you lean? So Meta, OpenAI, is there one you lean towards going?
149:44 - I spend time with both of those. And it's funny, because a few weeks ago, I didn't consider any of this. Um...
149:59 And it's really fucking hard. Like- - Yeah. - I have some... I know no people at OpenAI. I
150:11 love their tech. I think I'm the biggest codex advertisement shill that's unpaid. And it would feel so gratifying to,
150:18 like, put a price on all the work I did for free. And I would love if something happens and those companies get just merged, because it's
150:30 like... - Is this the hardest decision you've ever had to do?
150:39 - No. You know, I had some breakups in the past that feel like it's the same level.
150:48 - Um, and, and I also know that, in the end, they're both amazing. I cannot go
150:52 wrong. This is like- - Right. - This is, like, one of the most prestigious and, and, and, and, and
150:57 largest... I mean, not largest, but, like, they're both very cool companies.
151:02 - Yeah, they both really know scale. So, if you're thinking about impact, some of the wonderful technologies you've been
151:10 exploring, how to do it securely, and how to do it at scale, such that you can have a positive impact on a large number of people. They
151:17 both understand that. - You know, both Ned and Mark basically played all week with my product, and sent me
151:28 like, "Oh, this is great." Or, "This is shit. Oh, I need to change this."
151:32 Or, like, funny little anecdotes. And people using your stuff is kind of like the biggest compliment, and also shows me that, you
151:42 know, they actually... T- they actually care about it. And I didn't get the same on the OpenAI side. Um,
151:55 I got... I got to see some other stuff that I find really cool,
151:59 and they lure me with... I cannot tell you the exact number because of NDA, but
152:08 you can, you can be creative and, and think of the Cerebras deal and how that would
152:15 translate into speed. And it was very intriguing. You know, like, you give me Thor's hammer. Yeah.
152:28 ... been lured with tokens. So, yeah. - So, it- it's funny. So, so Marc started tinkering with the thing,
152:39 essentially having fun with the thing. - He got... He... Like, when he first... When he first approached me,
152:47 I got him in my, in my WhatsApp and he was asking, "Hey, when are we have a
152:54 call?" And I'm like, "I don't like calendar entries. Let's just call
152:58 now." And he was like, "Yeah, give me 10 minutes, I need to finish coding."
153:01 - Mm-hmm. - Well, I guess that gives you street cred. It's like, ugh, like, he's still writing
153:05 code. You know, he's- - Yeah, he does - ... he didn't drift away in just being a manager, he gets me.
153:11 That was a good first start. And then I think we had a, like, a
153:14 10-minute fight what's better, cloud code or Codex. Like, that's the thing you
153:21 first do, like, you casually call- - Yeah, that's awesome - ... someone with, like, the- that owns one of the largest companies in the world and,
153:28 and you have a 10 minutes conversation about that. - Yeah, yeah.
153:30 - Uh, and then I think afterwards he called me eccentric but brilliant. But
153:38 I also had some... I had some really, really cool discussion with Sam Altman and
153:46 he's, he's very thoughtful brilliant and I like him a lot from the, from the little time I had, yeah. I mean, I know it's peop-
154:06 some people vilify both of those people. I don't think it's fair.
154:15 - I think no matter what the stuff you're building and the kind of human you are
154:21 doing stuff at scale is kinda awesome. I'm excited. - I am super pumped. And you know the beauty is if,
154:32 if it doesn't work out, I can just do my own thing again. Like, I, I told them, like,
154:37 I, I don't do this for the money, I don't give a fuck. I- - Yeah.
154:42 - I mean, of course, of course it's a nice compliment but I wanna have
154:48 fun and have impact, and that's ultimately what made my decision.
154:58 - Can I ask you about... we've talked about it quite a bit, but maybe
155:02 just zooming out about how OpenCloud works. We talked about different
155:06 components, I want to ask if there's some interesting stuff we missed.
155:11 So, there's the gateway, there's the chat clients, there's the harness there's the agentic loop. You
155:20 said somewhere that everybody should im- implement an agent loop
155:24 at some point in their lives. - Yeah, because it's like the, it's like the Hello World in AI, you know? And
155:28 it's actually quite simple. - Yeah. - And it- it's good to understand that
155:34 that stuff's not magic. You can, you can easily build it yourself. So,
155:40 writing your own little cloud code... I, I even did this at a
155:44 conference in Paris for people to, like, introduce them to AI. I think it's it's a
155:49 fun little practice. Um, and you, you covered a lot. I think
155:55 one, one silly idea I had that turned out to be quite cool is
156:02 I built this thing with full system access. So it's like, you know, with great power comes great
156:09 responsibility. And I was like, "How can I up the stakes a little bit more?"
156:13 - Yeah, right. - And I just made a... I made it proactive. So, I added
156:20 a prompt. Initially, it was just a prompt, surprise me. Every, like, half an hour, surprise me, you know? And
156:28 later on I changed it to be like a little more specific and-
156:31 ... in the definition of surprise. Um, but the fact that I made it proactive and that it
156:38 knows you and that it cares about you, it- it's at least it's
156:44 programmed to that, prompted to do that. And that, that is a follow on, on your current session makes it very interesting because it would just sometimes ask a follow-up question or like, "How's your day?"
156:53 And I just made a... I made it proactive. So, I added a prompt. Initially, it was just a prompt, surprise me. Every, like, half an hour, surprise me, you know?
156:56 And later on I changed it to be like a little more specific and-
156:58 And that, that is a follow on, on your current session makes it very interesting because it would just sometimes ask a follow-up question or like, "How's your day?" I mean,
157:02 again, it's a little creepy or weird or interesting but Heartbeat very... in the beginning, it's still... today, it
157:12 doesn't... the model doesn't choose to use it a lot. - By the way, we're, we're, we're talking about Heartbeat, as you mentioned, the thing that
157:20 regularly- - Yeah. Like kicks- - ... Acts. - You just kick off the loop.
157:25 - Isn't that just a cron job, man? - Yeah, right, I mean, it's like-
157:29 - It's the cr- the criticisms that you get are hilarious. - You can, you can deduce any idea to like a
157:35 silly... Yeah, it's just, it's just a cron job in the end. I have like cron- separate cron jobs.
157:41 - Isn't love just evolutionary biology manifesting itself and isn't... aren't you guys just using each other?
157:49 - And then, yeah, and the project is all just glue of a few different dependencies-
157:53 ... and there's nothing original. Why do people... Well, you know,
157:57 isn't Dropbox just FTP with extra steps? - Yeah. - I found it surprising where I had this I had a
158:05 shoulder operation a few months ago, so. - Mm-hmm. - And the model rarely used Heartbeat, but then I was in the
158:11 hospital, and it knew that I had the operation and it checked up on
158:15 me. It's like, "Are you okay?" And I just... It's like, again, apparently, like, if something's significant in the
158:23 context, that triggered the Heartbeat when it rarely used the Heartbeat.... um,
158:30 and it does that sometimes for people, and that just makes it a lot more relatable.
158:36 - Uh, let me look this up on Perplexity, how OpenCall works just to see if I'm missing any of the stuff.
158:44 Local agent run time, high-level architecture. There's... Oh, we haven't talked much
158:48 about skills, I suppose. Skill hub, the tools in the skill lair,
158:52 but that's definitely a huge component and there's a huge growing set of skills-
158:55 - You know, you know what I love? That half a year ago, like everyone was talking about MCPs-
159:02 - Yeah - ... and I was like, "Screw MCPs. Uh, every MCP would be better as a
159:10 CLI." And now this stuff doesn't even have MCP support. I mean, it, it has with
159:17 asterisks, but not in the core lair, and nobody's complaining.
159:23 - Mm-hmm. - So my approach is if you want to extend the model with more features, you
159:32 just build a CLI and the model can call the CLI, probably gets it wrong, calls the help menu, and then on demand loads
159:43 into the context what it needs to use the CLI. It just needs a
159:47 sentence to know that the CLI exists if it's something that the model doesn't know about default. And even for a while,
159:54 I, I didn't really care about skills, but skills are actually perfect for
159:58 that because they, they boil down to a single sentence that explains the skill and then the model loads the skill, and that explains the
160:08 CLI, and then the model uses the CLI. Some skills are, like raw, but
160:13 most of the time, networks. - It's interesting um, I'm asking Perplexity MCP versus skills,
160:20 because this kind of requires a hot take that's quite recent, because your general view is MCPs are dead-ish. So MCPs is a more structured
160:32 thing. So if you listen to Perplexity here, MCP is what can I reach? So APIs,
160:38 database services files via protocol. So a structured protocol of
160:42 how you communicate with a thing, and then skills is more how should I work? Procedures, hostile helper scripts and prompts
160:49 are often written in a kind of semi- structured natural language,
160:53 right? And so technically skills could replace MCP if you have a smart enough model.
161:00 - I think the main beauty is, is that models are really good at
161:03 calling Unix commands. So if you just add another CLI, that's just another Unix command in the end. And
161:11 MCP is... That has to be added in training. That's not a very natural thing for the model. It requires a very specific
161:19 syntax. And the biggest thing, it's not composable. So imagine if I have a service that gives me better data
161:27 and gives me the temperature, the average temperature, rain,
161:31 wind and all the other stuff, and I get like this huge blob back. As a model, I always have to get the
161:38 huge blob back. I have to fill my context with that huge blob and then
161:42 pick what I want. There's no way for the model to naturally filter
161:48 unless I think about it proactively and add a filtering way into my
161:51 MCP. But if I would build the same as a CLI and it would give me this huge blob, it could just add a JQ command and filter itself
162:00 and then only, only get me what I actually need. Or maybe even
162:04 compose it into a script to, like do some calculations with the temperature and
162:08 only give me the exact output and the mo- and the... you have no context pollution.
162:14 Again, you can solve that with like sub- agents and more charades,
162:18 but it's just like workarounds for something that might not be the optimal way. There's... It definitely it
162:27 was, you know, it was good that we had MCPs because it pushed a lot of companies towards building APIs
162:34 and now I, I can like look at an MCP and just make it into a CLI.
162:37 - Mm-hmm. - Um, but this, this inherent problem that MCPs by default clutter up your
162:43 context. Plus the fact that most MCPs are not made good, in general make it just not a very useful paradigm. There's some exceptions like
162:57 Playwright for example that requires state and it's actually
163:01 useful. That is an acceptable choice. - So Playwright you use for browser use, which I think is c- already in
163:09 OpenClaw is quite incredible, right? - Yeah. - You can basically do everything, most things you can think of using browser use.
163:17 - That, that gets into the whole arch of every app is just a very slow API now, if they want or not. And that
163:28 through personal agents a lot of apps will disappear. You know, like I had a... I built
163:39 a CLI for Twitter. I mean, I- I just reverse engineered their website and used the internal API, which
163:49 is not very allowed. - It's called Bird, short-lived. - It was called Bird, because the bird had to disappear.
163:57 - The, the wings were clipped. - All they did is they just made access slower. Yeah,
164:03 not tak- you're not actually taking a feature away, but now inst- if, if your agent wants to read a tweet, it actually has to open the browser and
164:09 read the tweet. And it will still be able to read the tweet. It will just take longer.
164:13 It's not like you are making something that was possible, not possible. No. Now, it's just taking... Now it's just a
164:21 bit slower. So, so it doesn't really matter if your service wants to be an API or not. If I can
164:30 access it in the browser...... easy API. It's a slow API. - Can you empathize with their situation? Like, what would you do if you were Twitter, if you were
164:40 X? Because they're basically trying to protect against other large
164:43 companies scraping all their data. - Yeah. - But in so doing, they're cutting off like a million different use cases for
164:50 smaller developers that actually want to use it for helpful cool stuff.
164:54 - I think that if you have a very low per day baseline per account that allows read-only access
165:05 would solve a lot of problems. There's plenty, plenty of automations where people
165:09 create a bookmark and then use OpenClaw to, like, find the bookmark, do research on it, and then send you an email-
165:16 - Mm-hmm - ... with, like, more details on it or a summary. That's a cool approach.
165:21 I also want all my bookmarks somewhere to search. I would still like to have that.
165:26 - So, read-only access for the bookmarks you make on X. That seems like an
165:29 incredible application because a lot of us find a lot of cool stuff on X, we bookmark,
165:33 that's the general purpose of X. It's like, holy shit, this is awesome.
165:37 Oftentimes, you bookmark so many things you never look back at them.
165:40 - Yeah. - It would be nice to have tooling that organizes them and allows you to research it further.
165:44 - Yeah, I mean, and to be frank, I, I mean, I, I told Twitter proactively that, "Hey, I built
165:52 this and there's a need." And they've been really nice, but also like,
165:57 "Take it down." Fair. Totally fair. But I hope that this woke up the team a little bit that there's a need. And if all you do
166:09 is making it slower, you're just reducing access to your platform. I'm sure there's a better way. I also, I'm very much
166:19 against any automation on Twitter. If you tweet at me with AI, I will block you. No first strike. As
166:27 soon as it smells like AI, and AI still has a smell. - Mm-hmm.
166:32 - Especially on tweets. It's very hard to tweet in a way that does look completely human.
166:38 - Mm-hmm. - And then I block. Like, I have a zero tolerance policy on that. And
166:44 I think it would be very helpful if they, if, like, tweets done via API would be marked.
166:53 Maybe there's some special cases where... But, and there should
166:57 be, there should be a very easy way for agents to get their own
167:01 Twitter account. Um... - Mm-hmm. - We, we need to rethink social platforms a little bit if, if, if we, we, we go
167:11 towards a future where everyone has their agent and agents maybe have their own
167:17 Instagram profiles or Twitter accounts, so I can, like, do stuff on my behalf. I think it should
167:23 very clearly be marked that they are doing stuff on my behalf and it's not
167:26 me. Because content is now so cheap. Eyeballs are the expensive part. And
167:36 I find it very triggering when I read something and then I'm like, oh, no, this smells
167:40 like AI. - Yeah. Like, where, where is this headed in terms of what we value about the human
167:47 experience? It feels like we'll, we'll move more and more towards in-person interaction
167:52 and we'll just communicate. We'll talk to our AI agent to, to accomplish different tasks, to learn about different
168:01 things, but we won't value online interaction because there'll be so much
168:08 AI slob that smells and so many bots that it's difficult. - Well, if it's smart, then it shouldn't be difficult to filter. And then
168:20 I can look at it if I want to. But yeah, this is, like, a big thing we need to solve right now. E-
168:28 especially on this project, I get so many emails that are, let's say nicely, agentically written.
168:36 - Yeah. - But I much rather read your broken English than your AI
168:44 slob. You know, of course there's a human behind it, and yet they, they prompt
168:48 it. I'd much rather read your prompt than what came out. Um,
168:52 I think we're reaching a point where I value typos again. Like, um...
169:00 Like, and I, I mean, it also took me a while to, like, come to the realization. I, on my
169:05 blog I experimented with creating a blog post with agents and
169:12 ultimately it took me about the same time to, like, steer agent
169:16 towards something I like. But it missed the nuances that, how I would write it. You know, you can like, you
169:25 can steer it towards your style, but it's not gonna be all your style. So, I, I completely
169:32 moved away from that. I, I, everything, everything I blog is organic, handwritten
169:38 and maybe, maybe I, I, I use AI as a fix my worse typos. But
169:47 there's value in the rough parts of an actual human. - Isn't that awesome? Isn't that beautiful? That now because of AI
169:58 we value the raw humanity in each of us more. - I also, I also realized this thing that I, I rave about AI and use it so much for
170:08 anything that's code, but I'm allergic if it's stories. - Right. Yeah.
170:14 - Also, documentation, still fine with AI. You know, better than nothing.
170:17 - And for now it's still i- it applies in the mi- in the visual medium
170:21 too. It's fascinating how allergic I am to even a little bit of AI slob in in video
170:28 and images. It's useful, it's nice if it's like a little component of like-
170:32 - Or even, even those images. The, like, all these infographics and stuff,
170:36 the-... they trigger me so hard. - Yeah. - Like, it immediately makes me think less of your content. And it ... They were
170:45 novel for, like, one week and now it just screams slop. - Yeah.
170:51 - Even- even if people work hard on it, using ... And I- I have some on my blog post, you know, in the- in the time where I- I
171:00 explored this new medium. But now, they trigger me as well. It's like, yeah, this
171:04 is ... This just screams AI slop. I- - What... I don't know what that is, but I went through that too. I was really excited by the diagrams.
171:10 And then I realized, in order to remove from them hallucinations, you actually have to
171:14 do a huge amount of work. And you're just using it to draw the better diagrams, great. And then I'm proud of the diagram. I've used them for
171:22 literally, like, ki- ki- kind of like you said for maybe a couple of weeks. And now I look at
171:26 those, and I- I feel like I feel when I look at Comic Sans as a
171:30 font or- or something like this. It's like, "No, this is-" - It's a smell.
171:35 - "... this is fake. It's fraudulent. There's something wrong with it." And it...
171:41 - It's a smell. - And it's awesome because it re- it reminds you that we know.
171:48 There's so much to humans that's amazing and we know that. And we- we know
171:52 it. We know it when we see it. And so that gives me a lot of hope,
171:59 you know? That gives me a lot of hope about the human experience. It's not going to be damaged
172:03 by ... It's only going to be empowered as tools by AI. It's not going to be damaged or limited or somehow altered to where it's no longer
172:14 human. So ... Uh, I need a bathroom break. Quick pause. You mentioned that a lot of the apps might be
172:24 basically made obsolete. Do you think agents will just transform the entire app market?
172:30 - Yeah. Uh, I noticed that on Discord, that people just said
172:38 how their ... like, what they build and what they use it for. And it's
172:42 like, why do you need MyFitnessPal when the agent already knows where I am?
172:48 So, it can assume that I make bad decisions when I'm at, I don't know, Waffle House, what's around here? Or- or briskets in Austin.
172:58 - There's no bad decisions around briskets, but yeah. - No, that's the best decision, honestly. Um-
173:03 - Your agent should know that. - But it can, like ... It can modify my- my gym workout based
173:08 on how well I slept, or if I'm ... if I have stress or not. Like, it has so much more context to make even better
173:16 decisions than any of this app even could do. - Mm-hmm. - It could show me UI just as I like. Why do I still need an
173:25 app to do that? Why do I have to ... Why should I pay another subscription for
173:29 something that the agent can just do now? And why do I need my- my Eight Sleep app to control
173:38 my bed when I can tell the a- ... tell the agent to ... You know, the agent already
173:42 knows where I am, so he can, like, turn off what I don't use.
173:45 - Mm-hmm. - And I think that will ... that will translate into a whole
173:50 category of apps that are no longer ... I will just naturally stop using because my agent can just do it better.
174:00 - I think you said somewhere that it might kill off 80% of apps.
174:04 - Yeah. - Don't you think that's a gigantic transformative effect on
174:09 just all software development? So that means it might kill off a lot of software
174:13 companies. - Yeah. Um- - It's a scary thing. So, like, do you think about
174:19 the impact that has on the economy? On, Just the ripple effects it has to society?
174:27 Transforming who builds what tooling. It empowers a lot of users to get stuff done, to get stuff more efficiently, to get it done cheaper.
174:41 - It's also new services that we will need, right? For example, I want my agent to have an allowance. Like,
174:50 you solve problems for me, here's like 100 bucks in order to solve
174:54 problems for me. And if I tell you to order me food, maybe it uses a service. Maybe it uses something
175:01 like rent-a-human to, like, just get that done for me. - Mm-hmm.
175:06 - I don't actually care. I care about solve my problem. There's space for-
175:13 for new companies to solve that well. Maybe don't ... Not all apps
175:17 disappear. Maybe some transform into being API. - So, basically, apps that rapidly transform in being agent-facing.
175:30 So, there's a real opportunity for, like, Uber Eats, that we just used earlier today. It- it's companies this, of which there's many.
175:42 Who gets there fastest to being able to interact with OpenClaw in a way that's
175:48 the m- the most natural, the easiest? - Yeah. And also, apps will become API if they want or not. Because
175:57 my agent can figure out how to use my phone. I mean, on- on the other side,
176:02 it's a little more tricky. On Android, that's already ... People already do that. And then we'll just click the Order Uber
176:10 for Me button for me. Um, or maybe another service. Or maybe there's- there's a ... there's an API I can call so it's
176:17 faster. Uh, I think that's a space we're just beginning to even understand what that means. And I
176:26 ... Again, I didn't even ... That was not something I thought of. Something that I-
176:31 that I discovered as people use this, and it ... We are still so
176:34 early. But yeah, I think data is very important. Like, apps that can give me data, but that also can be
176:41 API. Why do I need a Sonos app anymore when I can ... when my agent can
176:45 talk to the Sonos?... Speakers directly. Like my cameras, there's like a crappy app,
176:52 but they have, they have an API, so my agent uses the API now.
176:57 - So it's gonna force a lot of companies to have to shift focus. That's kind
177:01 of what the internet did, right? You have to rapidly rethink, reconfigure what you're selling, how you're making money.
177:10 - Yeah, and some companies were really not like that. For example, there's no
177:14 CLI for Google, so I had to like, do... have to do anything myself and
177:21 build GAWK. That's like a CLI for Google. And at the... Yeah, at the end
177:28 user, they have to give me the emails because otherwise I cannot use their product. If I'm a company and I try
177:36 to get Google data, Gmail, there's a whole complicated process, to the point where sometimes
177:44 startups acquire startups that went through the process, so they
177:47 don't- don't have to work with Google for half a year to be certified
177:51 to being able to access Gmail. But my agent can access Gmail because I can just connect to it.
177:58 It's still crappy because I need to, like, go through Google's
178:03 developer jungle to get a key, and that's still annoying. But they cannot prevent me. And worst case, my
178:13 agent just clicks on the, on the website and gets the data out that way.
178:17 - Through browsers? - Yeah. I mean, I, I watch my agent happily click the I'm not a robot button.
178:25 And there's this, this whole... That's gonna be... That's gonna be more heated. You see companies like Cloudflare that
178:36 try to prevent bot access. And in some ways, that's useful for
178:39 scraping. But in other ways, if I'm, I'm a personal user, I want that. You know, sometimes
178:46 I, I use Codex and I, I read an article about modern React patterns,
178:55 and it's like a Medium article. I paste it in and the agent can't read
178:59 it because they block it. So then I have to copy-paste the actual text. Or in
179:05 the future, I'll learn that maybe I don't click on Medium because it's annoying, and I use
179:09 other websites that actually are agent friendly. So, uh- - There's gonna be a lot of powerful, rich companies fighting back.
179:17 So it's really intere- You're at the center, you're the catalyst, the leader,
179:23 and happen to be at the center of this kind of revolution where it's get- gonna completely
179:26 change how we interact with services with, with web. And so, like, there's companies at
179:35 Google that are gonna push back. I mean, there's every major companies you could think of is gonna push
179:39 back. - Even... Yeah, even search. Um, I now use, I think Perplexity or Brave
179:47 as providers because Google really doesn't make it easy to use Google
179:51 without Google. I'm not sure if that's the right strategy, but I'm not
179:55 Google. - Yeah, there's a, there's a nice balance from a big company perspective 'cause if you
180:02 push back too much for too long, you become Blockbuster and you lose everything to the
180:06 Netflixes of the world. But some pushback is probably good during a
180:09 revolution to see. - Yeah. But you see that, that... Like, this is something that the people want.
180:14 - Right. - So- - Yes. - If I'm on the go, I don't wanna open a calendar app. I just... I wanna tell my agent,
180:22 "Hey, remind me about this dinner tomorrow night," and maybe invite
180:26 two of my friends and then maybe send a what- send a WhatsApp message to my
180:29 friend. And I don't need... I don't want or need to open apps for that. I think that
180:36 we passed that age, and now everything is, like, much more connected and, and fluid if those companies want it or
180:44 not. And I think, well, the right companies will find ways to jump on the train, and other companies will perish.
180:55 - You got to listen to what the people want. We talked about programming quite a
180:59 bit, and a lot of folks that are developers are really worried about their jobs, about
181:05 their... About the future of programming. Do you think AI replaces
181:09 programmers completely? Human programmers? - I mean, we're definitely going in that direction. Programming is just
181:16 a part of building products. So maybe, maybe AI does replace programmers
181:24 eventually. But there's so much more to that art. Like, what do you actually wanna build? How should it feel? How's the
181:36 architecture? I don't think agents will replace all of that.
181:40 Yeah, like, just the, the actual art of programming, it will, it will stay there, but it's, it's gonna be like
181:49 knitting. You know? Like, people do that because they like it, not because it makes
181:53 any sense. So the... I read this article this morning about someone that it's okay to mourn our craft. And I
182:03 can... A part of me very strongly resonates with that because
182:07 in my past I, I spent a lot of time tinkering, just being really deep in the flow and just, like,
182:15 cranking out code and, like, finding really beautiful solutions. And
182:20 yes, in a way it's, it's sad because that will go away. And I also get a lot of joy out of just
182:33 writing code and being really deep in my thoughts and forgetting
182:38 time and space and just being in this beautiful state of flow. But you can get the same state of
182:46 flow... I get a similar state of flow by working with agents and
182:50 building and thinking really hard about problems. It is different-...
182:55 but... And it's okay to mourn it, but, uh, I mean, that's not something we can fight. Like, there
183:03 is... the world for a long time had a... there was a lack of intelligence, if you s- if you see it like
183:10 that, of people building things, and that's why salaries of software developers reached stupidly high amounts
183:23 and then will go away. There will still be a lot of demand for people that understand how to build things. It's just that all this
183:34 tokenized intelligence enables people to do a lot more, a lot
183:42 faster. And it will be even more... even faster and even more because
183:46 those things are continuously improving. We had similar things when... I mean, it's probably not a perfect
183:53 analogy, but when we created the steam engine, and they built all these factories and replaced a lot of manual
184:00 labor, and then people revolted and broke the machines. Um, I- I can relate that if you very deeply
184:11 identify that you are a programmer, that it's scary and that it's threatening because
184:20 what you like and what you're really good at is now being done by
184:27 a soulless or not entity. But I don't think you're just a programmer. That's a very limiting
184:35 view of your craft. You are, you are still a builder. - Yeah, there's a couple of things I want to say. So one is, I never... As you're
184:43 articulating this beautifully, I no- I'm realizing I never thought I would...
184:49 the thing I love doing would be the thing that gets replaced. You hear these stories about these, like you
184:56 said, with the steam engine. I've, I've spent so many, I don't know, maybe thousands of hours
185:03 poring over code and putting my heart and soul and, like, and just, like, some of my most painful and happiest
185:10 moments were alone behind... I, I was an Emacs person for a long time. Man, Emacs.
185:17 And, and then there's an identity and there's meaning, and there's... Like, when I walk about the world, I don't say it out loud, but
185:25 I think of myself as a programmer. And to have that in a matter
185:29 of months... I mean, like you mentioned, April to November, it really is a leap that happened, a shift that's happening.
185:39 To have that completely replaced is is painful. It's, it's truly painful. But I also think
185:47 programmers, builders more broadly, but what is, what is the act of programming? I, I think
185:55 programmers are generally best equipped at this moment in history
186:00 to learn the language, to empathize with agents, to learn the language of agents. To
186:08 feel the CLI. - Yeah. - Like, like to understand what is the thing you need, you the agent, need to
186:19 do this task the best? - I think at some point it's just gonna be called coding again, and it's just gonna be the new
186:24 normal. - Yeah. - And yet, while I don't write the code, I very much feel like
186:30 I'm in the driver's seat and I am, I am writing the code, you know? It's just-
186:37 - You'll still be a programmer. It's just the activity of a programmer is,
186:41 is different. - Yeah, and because on X, the bubble, I mean, is mostly positive. On, on Mastodon and
186:47 Bluesky, I don't... I also use it less because oftentimes I got attacked for my blog posts. And
186:56 I, I had stronger reactions in the past, now I can sympathize with those people more
187:00 'cause, in a way I get it. It... In a way, I also don't get it because
187:06 it's very unfair to grab onto the person that you see right now and unload all your fear and hate.
187:15 It's gonna be a change and it's gonna be challenging, but it's
187:18 also... I don't know. I find it incredibly fun and, and, and gratifying. And
187:25 I can, I can use the new time to focus on much more details. I think the level of expectation of what we
187:33 build is also rising because it's just now... The default is now
187:39 so much easier, so software is changing in many ways. There's gonna be a lot
187:46 more. And then you have all these people that are screaming, "Oh yeah, but what about the water?" You
187:53 know? Like, I did a conference in Italy about the, the state of AI, and
188:01 m- my whole motivation was to push people away from, don't see
188:05 yourself as an iOS developer anymore. You're now a builder, and
188:09 you can use your skills in many more ways. Also because apps are slowly going
188:12 away. People didn't like that. Like a lot of people didn't like what I had to say. And I don't think I was hyperbole, I was
188:21 just like, "This is how I see the future." Maybe this is not how it's going to be, but I'm pretty sure a version of that will happen.
188:30 And the first question I got was, "Yeah, but what about the insane water use on
188:34 data centers?" But then you actually sit down and do the maths, and then for most people if you just skip one burger per month, that
188:44 compensates the, the CO2 output, or, like, the water use in equivalent of tokens. I
188:52 mean, the maths is, is... the maths is tricky, and it depends if you add pre-training, then maybe
188:58 it's more than just one patty.... but it's not off by a factor of 100, you know? So, so the... or like golf is still using way
189:09 more water than all data centers together. So are you also hating people
189:13 that play golf? Those people grab on anything that they think is bad about AI without seeing the potential
189:21 things that might be good about AI. - Mm-hmm. - And I'm not saying everything's good. It's certainly gonna
189:27 be a very transformative technology for our society. - There's to steel man the, the criticism in
189:36 general, I do wanna say in my experience with Silicon Valley there's a bit of a bubble in the sense that
189:46 there's a kind of excitement and an over- focus about the positive that the technology can bring.
189:54 - Yeah. - And... which is great. It's great to focus on... N- not to,
189:59 not to be paralyzed by fear and fear- mongering and so on, but there's
190:03 also within that excitement, and within everybody talking just to
190:07 each other, there's a dismissal of the basic human experience across the United States and the Midwest, across the world.
190:16 Including the programmers we mentioned, including all the people that are gonna lose their jobs,
190:19 including the s- the measurable pain and suffering that happens at
190:23 the short-term scale when there's change of any kind. Especially
190:28 large-scale transformative change that we're about to face if what we're talking about will materialize. And so to
190:35 ha- having a bit of that humility and awareness about the tools you're building, they're going to cause pain.
190:43 They will long term hopefully bring about a better world, and even more opportunities-
190:48 - Yeah - ... and even more awesomeness. But having that kind of like quiet moment
190:56 often of, of respect for the pain that is going to be felt. And so not, not enough of that is, I think, done,
191:04 so it's, it's good to have a bit of that. - And then I also have to put against some of the
191:11 emails I got where people told me they have a small business, and they've been
191:15 struggling. And, and OpenClaw helped them automate a few of the
191:19 tedious tasks from, from collecting invoices to like answering customer emails
191:26 that then freed them up and like cost them a bit more joy in their life.
191:30 - Mm-hmm. - Or, or some emails where they told me that OpenClaw helped
191:35 their disabled daughter. That she's now empowered and feels she can do much more than before. Which is
191:42 amazing, right? Because you could, you could do that before as well. The technology was there. I didn't,
191:48 I didn't invent a whole new thing, but I made it a lot easier and more accessible, and that
191:55 did show people the possibilities that they previously wouldn't see.
192:00 And now they apply it for good. - Mm-hmm. - Or like also the fact that, yes, I, I, I suggest the, the, the latest and
192:09 best models, but you can totally run this on free models. You
192:13 can run this locally. You can run this on, on, on Keyme or other,
192:18 other, other models that are way more accessible price-wise,
192:23 and still have a, a very powerful system that might otherwise not be
192:27 possible. Because other things like, I don't know, Entropik's CoWork
192:32 is locked in into their space, so it's not all black and white. There's... I got a lot of
192:39 emails that were heartwarming and amazing. And, and I don't know, it just made me really happy.
192:48 - Yeah, there's a lot... It has brought joy into a lot of people's lives. Not just,
192:52 not just programmers. Like a lot of people's lives. It's, it's,
192:56 it's beautiful to see. What gives you hope about this whole thing we have going on
193:02 with human civilization? - I mean, I inspired so many people. There's like... there's this whole
193:09 builder vibe again. People are now using AI in a more playful way
193:17 and are discovering what it can do and how it can like help them in their life.
193:24 And creating new places that are just sprawling of creativity. I don't know. Like,
193:35 there's like ClawCoin in Vienna. There's like 500 people. And there's such a high percentage of
193:41 people that uh, want to present, which is to me really surprising, because u-
193:45 usually it's quite hard to find people that want to like talk about what they built. And now it's, there's an
193:51 abundance. So that gives me hope that we can, we can figure shit out.
194:00 - And it makes it accessible to basically everybody. - Yeah.
194:05 - Just imagine all these people building, especially as you make it simpler and simpler, more
194:12 secure. It's like anybody who has ideas and can express those ideas in language can
194:20 build. That's crazy. - Yeah, that's ultimately power to the people, and one of the beauty,
194:28 the beautiful things that come out of AI. Not just, not just a slop generator.
194:36 - Well, Mr. Clawfather, I just realized when I said that in the beginning,
194:40 I violated two trademarks, because there's also the Godfather. I'm
194:44 getting sued by everybody. Um, you're a wonderful human being. You've created something really
194:51 special, a special community, a special product, a special set of ideas. Plus, the entire... the humor, the good vibes,
194:59 the inspiration of all these people building, the excitement
195:04 to build. So I'm truly grateful for everything you've been doing and for who you are, and for sitting down
195:12 to talk with me today. Thank you, brother. - Thanks for giving me the chance to tell my story.
195:17 - Thanks for listening to this conversation with Peter Steinberger. To support this podcast,
195:21 please check out our sponsors in the description, where you can also find links to contact
195:25 me, ask questions, give feedback and so on. And now let me leave you with some words from Voltaire. "With great power comes great
195:35 responsibility." Thank you for listening, and hope to see you next time.
$
OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491
Peter Steinberger is the creator of OpenClaw, an open-source AI agent framework that's the fastest-growing project in GitHub history.
Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep491-sb
See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc.
*Transcript:*
https://lexfridman.com/peter-steinberger-transcript
*CONTACT LEX:*
*Feedback* - give feedback to Lex: https://lexfridman.com/survey
*AMA* - submit questions, video
