// transcript — 687 segments
0:00 Intro
0:02 There's a way to tune your clawed code so you get absolute superpowers. Some of
0:07 my smartest [music] friends are using this tool and it's called Last 30 Days.
0:11 It just came out and I had the founder Matt Van Horn come on the podcast to
0:16 show how to use it. What last 30 days does is it uses trending data on places
0:21 like Axe and Reddit as the starting point for your prompts on cloud code.
0:26 The cool part is you end up getting way more out of cloud code because the
0:29 prompts are way more dialed and optimized. [music] Enjoy the episode.
0:33 Let me know if you think this is cool [music] >> I feel like I'm in I know I'm in for a
0:49 treat. Matt Van Horn is on the podcast and he's going to demo something very
0:54 special. met by the end of this episode. What are people going to learn?
0:58 >> They will have a new superpower in cloud code. It's called slashlast30 days and
1:04 uh it's going to for any topic you could possibly imagine from best prompting
1:09 techniques for a particular tool to how should I use claudebot to what are the
1:14 greatest rap songs right now. It's going to search X, Twitter, and the web for
1:18 only things that have happened in the last 30 days and give you a great result
1:22 and help you be expert. >> My smartest friends are using this right
1:26 now, so I had to ask Matt to come on and showcase how to use this. So, thank you,
1:31 Matt. Uh, I say we just get right into it. >> All right, yolo. Let's uh let's go. I
1:41 can just uh dive in. So we are in cloud code and so this is a skill for cloud
1:46 code and so once you install it you just type in last 30 days. So research any
1:53 topic from the last 30 days. So uh can be anything. So let's say I want
2:01 um let's kick one off right here. Um most popular rap songs. So, let's kick that one off. And
2:11 so, and then I'll I'll show some ones some pre-en ones that I've done. So,
2:14 using the last 30 days school, let's research most popular rap songs. Let me
2:17 research some popular rap songs from the last 30 days. And so, right here, it's
2:21 Reddit reading what Redditors are saying. X reading the timeline. And then
2:27 it's also running a a web search. So, so is the is the thinking like is the
2:31 reason why you created this is there's so much good data on the internet and
2:34 you want to use that as like the jumping off point. Is that why
2:39 >> the the reason that I built this tool is I feel like everything is moving so
2:46 quickly in AI and it's nearly impossible to keep up with the conversation on X on
2:54 Reddit on GitHub and I wanted a tool that gives you superpowers
3:00 to be able to just become expert on any topic based on what's happening on
3:05 Reddit X and the web. very very quickly because the prompts are changing so
3:09 quickly. What's happened to like look at you know Maltbot? What's happened there?
3:15 It's changing so quickly. And so being able to just become expert at something
3:18 it kind of reminds me in the Matrix when he gets plugged in and it's like I know
3:22 kung fu, right? So to be able to do that for any topic very very quickly is is
3:28 why I built this and honestly so I could use it for myself. Looks like our our
3:32 rap research. So, I found one Reddit thread, 19 expost, and the most popular rap
3:39 songs right now. So, it kind of dug in and into hip-hop heads, some expos, Billboard data, some
3:47 Spotify data, hip-hop heads, hip-hop all day, rap complex, and created this uh
3:52 this reply. >> So, someone in the comment section is going to be like, well, I could have
3:58 used perplexity or chat GBT to do that. Why should I use this?
4:04 >> Yeah. So, absolutely could, but it's uh and again I'm I'm not fully up to date
4:08 on which which tools have what access but to set up last 30 days. So you need
4:14 cloud code, right? So you're already paying for cloud code, right? So you
4:18 need that account. You need a open AI key because OpenAI has a deal with
4:23 Reddit that gives you that Reddit access. Uh I'm actually see looking into Reddit API
4:29 keys if that's an option to make the the product better. That's what I was um
4:33 that's one of the things my my weekend projects. And then the the third is you
4:39 can't search X using your X account. You have to use an XAI key. And so last 30
4:45 days pulls in all these API keys and pulls it all together if that makes
4:51 sense. So this is one that I did right before we started. So I
4:55 said last 30 days highest performing cold email frameworks for ICP output
5:02 three email variants subject line. So it went on X went on Twitter found
5:09 Reddit threads expost web pages and what what's interesting how I use this tool
5:14 is I often don't even read what it says. Like sure it's interesting to see what
5:17 it learned but mostly I just wanted it to write a good email. So I said, "Can
5:20 you write me some cold emails for getting on Greg Eisenberg's podcast?"
5:25 Uh, sorry I spelled your name wrong. Good target startup ideas later is all
5:27 about unconventional startup ideas, community building unique relevance. You
5:31 know, what's your angle? What's your credibility signal? Any connection
5:34 points to Greg mutual follows. What timely uh talk about AI tools I'm
5:40 working on and I once made a smart oven. Uh, smart oven details gold unexpected
5:44 founders who shipped hardware and software. That's your hook. Here are
5:48 three email variants. subject smart oven AI tools not the P Greg I went from building smart oven to
5:53 building AI tools that might sound weird until you hear it turns out the hardest
5:56 part of hardware taught me exactly what's broken and how people interact
6:00 with AI I'm building the fixed what is smart oven taught me about AI so
6:05 anyway I literally did nothing here I barely gave it any context and it had
6:12 become expert in different cold emailing concepts the three Ps framework praise
6:16 picture push ADA attention I've never read any of these in my life
6:20 intentionbased data trigger framework and then it just did it for me if that
6:24 makes sense without even me having to read the research. Yeah. I think what's
6:29 really cool about this is these frameworks. I mean, ADA has been around
6:34 for forever like you know OG David Oglevie days, but other than you know
6:39 the other ones like it feels like those are timely frameworks
6:43 like these emails like I'm reading these emails and they're I get a lot of cold
6:47 email and these are actually ones that would break through the noise. So that's
6:51 what's really cool about this. >> And I didn't even give it much effort. I
6:55 literally said, "I'm a former smart oven >> You couldn't have given it like less
7:01 context. [laughter] >> And it did. Okay. >> Yeah.
7:06 >> And then another one I I just tried last 30 days how to get X followers. I
7:12 researched people actually saying um and what I learned reply is the number one
7:16 growth strategy. Multiple X power users all credit becoming a reply guy as the
7:22 fastest path to growth. shows 380 thoughtful replies, daily engagement, 40
7:25 before posting. Find larger accounts. Reply first consistently. Post at least
7:31 1x a day, 5 days a week. U Reddit threads, X threads, Instagram marketing,
7:36 Facebook ads, and share your vision. What's your account and what's
7:41 followers? Uh, I'm M. Van Horn and I made an AI tool for claw code. I don't
7:46 know what it's going to do now. [laughter] >> That's the exciting part. one one other
7:53 let's spin up another last 30 days window. Sorry for all the all the
7:58 windows. So all right so last 30 days of research cla bot so I can
8:18 build an enterprise competitor to make money and not give it away for free.
8:22 Maybe I'm giving too much context but I should have just said research clav um
8:27 actually I'm going to do that. I mean, what do you recommend when you're, you
8:31 know, is it shorter prompts, medium prompts, long prompts? Like, how do
8:34 getting the most out of >> I think it's a it's kind of you prime
8:38 the engine by first doing just research cloudbot top use cases. So, what my my
8:45 vision for this thread of what we're going to do is it's going to learn about
8:50 Cloudbot, right? Check X, check Reddit, and then what I'm going to do is
8:54 actually load one of my favorite tools, Compound Engineering. Uh, shout out to
8:58 to Kevin Rose for introducing me to it. And compound engineering is when I'm
9:03 starting a new project is where I do my planning. And so what I want to do here
9:10 is take this research on from last 30 days from Cloudbot and use that as the
9:19 starting point to kick off a enterprise version of Cloudbot and to build it. So
9:25 let's let's let this run and see how our other stuff is going. All right. I'm in
9:30 I'm in A tool for cloud code. A tools are hot right now. Here's a growth
9:33 playbook for you. Morning. 40 minutes for your first post. Reply to tip from
9:37 anthropic. Alex Burch Labs. Cloud power users. Add genuine value. Share a tip.
9:41 Ask a smart question. Content type tool demo. Build X today. What it does. Cloud
9:46 code tip. Five things I learned X. Build in public. Ship. By the way, I I'm not
9:49 going to do this. I think this was just I I actually asked chat GPT for
9:53 suggestions of what I should demo on your show and it gave me this one but
9:58 quickest wins. Pin your best demo tweet. Show the tool in action. Bio formula I
10:02 build tool one logo shipping AI tools for cloud code. >> That's not bad.
10:07 >> No, it's it's not bad. [laughter] So [snorts] all right, let's close this
10:11 one out. All right, we got cloud code best use overnight coding agent
10:14 management daily briefings and morning. Do you want to dive deeper? Okay. So now
10:18 I'm going to call in another tool. Let's plan throw in workflows plan. Take the
10:26 context above about Claudebot and propose an enterprise version that could make a lot of money.
10:39 Just do that. [laughter] Keep it simple. >> Yeah. I mean this is this is
10:42 interesting, right? like [laughter] you know you're taking a trend open
10:46 source trend and you're like okay how can I build a product that you know gets
10:52 me paid and it's interesting because there's so many trends that are
10:56 happening right now on X as like a data source and Reddit as a data source that
11:00 you know if this if this is actually good is going to help give you ideas and sort of
11:07 eventually a PRD right >> yep anyway so it's still making our plan
11:13 it's researching still doing an analysis. So, we'll we'll we'll see where that
11:17 Best Practices for Webdesign Research
11:18 adds up. But anyway, do you want you want to try a last 30-day prompt? Yeah.
11:22 I mean, I'm curious like I mean you you're seeing all the all these prompts.
11:25 So, like the people that listen to this show, these are people these are
11:29 founders, solopreneurs, people building businesses. Um they're always looking
11:35 for unfair advantages. Um like what sort of what sort of prompt do you would you
11:41 suggest that type of person someone who wants to build a business? Um you know
11:49 is it competitive research? Is it um you know what what are some ideas you have?
11:55 Good good question. Um I mean I think it's it's very good for competitive
11:59 research. I think it's very good for looking at at hot topics. Um, you know,
12:04 I think it's, you know, we could probably do something like, um, you
12:11 know, let's try it for your your web design idea, right? You kind of describe
12:17 poke.com, which I missed, right? So, um, what webpage designs are getting the
12:24 most love right now? Yeah, because that's interesting because if we, you
12:29 know, knowing the trend is just so important. And uh it also helps like give, you know,
12:37 when I'm when I'm designing something, it helps give me ideas. It gives it's so
12:41 overwhelming sometimes when you have like a blank page and you don't know
12:43 where to start. >> Yep. >> We've all been there. And I kind of my
12:51 kind of the the the thing that inspired this tool was I used to before I would
12:57 uh jump into any type of vibe coding session, I would go on chat GBT 5.2 and
13:04 I would just say in thinking mode and be like go uh this was for design. I would say go
13:12 on Reddit and research Nana Banana Pro best uh prompting techniques and I would do
13:18 that at the beginning and then I wouldn't even read what it would say and
13:21 then I would just be like, "Okay, give me a prompt." And so it's it's kind of
13:27 this like hack of, you know, learning kung fu, learning the prompts that
13:30 everyone's using without actually needing to read about it, if that makes
13:35 sense. you kind of just like, okay, let's just trust the the the mind crowd,
13:42 the the world that's out there and and use those best practices. All right,
13:44 Building an Enterprise Moltbot Clone Live
13:48 we've got a plan. Let's uh I guess let's open the plan in uh BBEdit. All right,
13:54 >> so this is the plan for for the audio listeners. So this is the plan for the
13:59 the mobile >> enterprise commercial SAS platform. So again, this is not a business plan. This
14:08 is actually a plan to how to build the software architecture and I could just
14:12 then tell compound engineering to just go build it. So transform maltbot form
14:17 cloud into an open source AI into a multi-tenant enterprise SAS platform
14:21 called moltra enterprise or cloud cloud. This capitalizes on the viral growth.
14:26 GitHub stars B2B adoption enterprise AI assistance are 30 blah blah blah. Why
14:31 enterprise can't use multipart today? No multi-tensity, no RBA security
14:36 vulnerabilities, no audit logging. Uh there move wise required by service
14:39 now for market validation. proposed solution of fully managed
14:44 enterprisegrade maltbot commercial I really don't like the name
14:50 cloud cloud come on come on lm cloud architecture slack web hook discord
14:55 gateway team hooks message radar core technical components and so kind of it
15:00 used last 30 days as the basis for its expertise to learn about maltbot and
15:08 then it did the rest and now I could just say okay build it.
15:11 >> So, >> dude, that's crazy. It also that that was a pretty that was pretty dialed.
15:20 >> I'm impressed. >> And and LLMs knew nothing about
15:26 Claudebot because [laughter] it didn't exist before didn't exist till very very
15:32 recently. All right. Phase one only build a multi-tenant foundation postcress but
15:41 has full MVP. Uh I guess let's uh let's build the proof of concept. So So
15:46 basically, you know, it asked if you wanted to do an MVP. It asked if you wanted to do a
15:50 bigger build. Is that is that what I saw? >> Yeah, exactly. And so I just said,
15:55 "Okay, let's just do the the demo." >> Yeah. >> I typed something else. New repo. Also,
16:02 can you please give it a better name? Greg, you're you're you live with this
16:05 world. What should we call this thing? >> I mean, why not? I actually don't hate
16:12 uh Claude Cloud. It just Well, >> well, they got in trouble for using
16:15 claw. >> Exactly. So, why not just call it molt cloud?
16:19 >> But then we're getting in trouble with malt. [laughter]
16:24 >> Mold. Um, okay. So, why don't we call it >> Should we go lobster? [laughter]
16:30 >> Um, I was going to say, why don't we call it red lobster?
16:32 >> [laughter] >> But I think that >> All right. All right. Let's go with
16:38 Let's go with lead Speak. >> Let's go with Red Lava. >> That's cool.
16:43 >> This describes a 40week product surveys. What scope do you want to do? I don't
16:47 know. Just Just do it. >> Just do it. Why are you asking me so
16:51 many questions? >> All right. Um All right. What do we ask
16:55 here? What web designs are getting the most love right now? >> Um the most love Shopify winter edition.
17:01 Yes, I remember. I saw that one. That was really >> That was really awesome. That was
17:04 beautiful. >> Look, 3,000 likes, 320 retweets, the YC landing page. So, what's funny is I saw
17:11 this. I didn't actually look at it yet, but I saw that it was getting love and
17:14 hype. Praise for scrappy humble beginnings. Shop web landing all like
17:21 jewelry, craft works, design, all agents reported back. Reddit didn't
17:25 find anything topic too visual for text discussions, but found uh what tool do
17:29 you want to use to create designs? So it's like, okay, now that I've
17:32 researched these things, what do you what do you want to design in, right?
17:38 Um, so it's kind of it's kind of pushing you to be like, okay, I've now learned
17:42 this expertise. Now, let's do something with it. >> So So this could literally pump out a
17:43 Generating Figma Prompts and Nano Banana Images
1:39 What Is "Last 30 Days"
1:41 can just uh dive in. So we are in cloud code and so this is a skill for cloud
1:46 code and so once you install it you just type in last 30 days. So research any
1:53 topic from the last 30 days. So uh can be anything. So let's say I want
2:01 um let's kick one off right here. Um most popular rap songs. So, let's kick that one off. And
2:11 so, and then I'll I'll show some ones some pre-en ones that I've done. So,
2:14 using the last 30 days school, let's research most popular rap songs. Let me
2:17 research some popular rap songs from the last 30 days. And so, right here, it's
2:21 Reddit reading what Redditors are saying. X reading the timeline. And then
2:27 it's also running a a web search. So, so is the is the thinking like is the
2:31 reason why you created this is there's so much good data on the internet and
2:34 you want to use that as like the jumping off point. Is that why
2:39 >> the the reason that I built this tool is I feel like everything is moving so
2:46 quickly in AI and it's nearly impossible to keep up with the conversation on X on
2:54 Reddit on GitHub and I wanted a tool that gives you superpowers
3:00 to be able to just become expert on any topic based on what's happening on
3:05 Reddit X and the web. very very quickly because the prompts are changing so
3:09 quickly. What's happened to like look at you know Maltbot? What's happened there?
3:15 It's changing so quickly. And so being able to just become expert at something
3:18 it kind of reminds me in the Matrix when he gets plugged in and it's like I know
3:22 kung fu, right? So to be able to do that for any topic very very quickly is is
3:28 why I built this and honestly so I could use it for myself. Looks like our our
3:32 rap research. So, I found one Reddit thread, 19 expost, and the most popular rap
3:39 songs right now. So, it kind of dug in and into hip-hop heads, some expos, Billboard data, some
3:47 Spotify data, hip-hop heads, hip-hop all day, rap complex, and created this uh
3:52 this reply. >> So, someone in the comment section is going to be like, well, I could have
3:58 used perplexity or chat GBT to do that. Why should I use this?
4:04 >> Yeah. So, absolutely could, but it's uh and again I'm I'm not fully up to date
4:08 on which which tools have what access but to set up last 30 days. So you need
4:14 cloud code, right? So you're already paying for cloud code, right? So you
4:18 need that account. You need a open AI key because OpenAI has a deal with
4:23 Reddit that gives you that Reddit access. Uh I'm actually see looking into Reddit API
4:29 keys if that's an option to make the the product better. That's what I was um
4:33 that's one of the things my my weekend projects. And then the the third is you
4:39 can't search X using your X account. You have to use an XAI key. And so last 30
4:45 days pulls in all these API keys and pulls it all together if that makes
4:51 sense. So this is one that I did right before we started. So I
4:55 said last 30 days highest performing cold email frameworks for ICP output
5:02 three email variants subject line. So it went on X went on Twitter found
5:09 Reddit threads expost web pages and what what's interesting how I use this tool
5:14 is I often don't even read what it says. Like sure it's interesting to see what
5:17 it learned but mostly I just wanted it to write a good email. So I said, "Can
5:20 you write me some cold emails for getting on Greg Eisenberg's podcast?"
5:25 Uh, sorry I spelled your name wrong. Good target startup ideas later is all
5:27 about unconventional startup ideas, community building unique relevance. You
5:31 know, what's your angle? What's your credibility signal? Any connection
5:34 points to Greg mutual follows. What timely uh talk about AI tools I'm
5:40 working on and I once made a smart oven. Uh, smart oven details gold unexpected
5:44 founders who shipped hardware and software. That's your hook. Here are
5:48 three email variants. subject smart oven AI tools not the P Greg I went from building smart oven to
5:53 building AI tools that might sound weird until you hear it turns out the hardest
5:56 part of hardware taught me exactly what's broken and how people interact
6:00 with AI I'm building the fixed what is smart oven taught me about AI so
6:05 anyway I literally did nothing here I barely gave it any context and it had
6:12 become expert in different cold emailing concepts the three Ps framework praise
6:16 picture push ADA attention I've never read any of these in my life
6:20 intentionbased data trigger framework and then it just did it for me if that
6:24 makes sense without even me having to read the research. Yeah. I think what's
6:29 really cool about this is these frameworks. I mean, ADA has been around
6:34 for forever like you know OG David Oglevie days, but other than you know
6:39 the other ones like it feels like those are timely frameworks
6:43 like these emails like I'm reading these emails and they're I get a lot of cold
6:47 email and these are actually ones that would break through the noise. So that's
6:51 what's really cool about this. >> And I didn't even give it much effort. I
6:55 literally said, "I'm a former smart oven >> You couldn't have given it like less
7:01 context. [laughter] >> And it did. Okay. >> Yeah.
7:06 >> And then another one I I just tried last 30 days how to get X followers. I
7:12 researched people actually saying um and what I learned reply is the number one
7:16 growth strategy. Multiple X power users all credit becoming a reply guy as the
7:22 fastest path to growth. shows 380 thoughtful replies, daily engagement, 40
7:25 before posting. Find larger accounts. Reply first consistently. Post at least
7:31 1x a day, 5 days a week. U Reddit threads, X threads, Instagram marketing,
7:36 Facebook ads, and share your vision. What's your account and what's
7:41 followers? Uh, I'm M. Van Horn and I made an AI tool for claw code. I don't
7:46 know what it's going to do now. [laughter] >> That's the exciting part. one one other
7:53 let's spin up another last 30 days window. Sorry for all the all the
7:58 windows. So all right so last 30 days of research cla bot so I can
8:18 build an enterprise competitor to make money and not give it away for free.
8:22 Maybe I'm giving too much context but I should have just said research clav um
8:27 actually I'm going to do that. I mean, what do you recommend when you're, you
8:31 know, is it shorter prompts, medium prompts, long prompts? Like, how do
8:34 getting the most out of >> I think it's a it's kind of you prime
8:38 the engine by first doing just research cloudbot top use cases. So, what my my
8:45 vision for this thread of what we're going to do is it's going to learn about
8:50 Cloudbot, right? Check X, check Reddit, and then what I'm going to do is
8:54 actually load one of my favorite tools, Compound Engineering. Uh, shout out to
8:58 to Kevin Rose for introducing me to it. And compound engineering is when I'm
9:03 starting a new project is where I do my planning. And so what I want to do here
9:10 is take this research on from last 30 days from Cloudbot and use that as the
9:19 starting point to kick off a enterprise version of Cloudbot and to build it. So
9:25 let's let's let this run and see how our other stuff is going. All right. I'm in
9:30 I'm in A tool for cloud code. A tools are hot right now. Here's a growth
9:33 playbook for you. Morning. 40 minutes for your first post. Reply to tip from
9:37 anthropic. Alex Burch Labs. Cloud power users. Add genuine value. Share a tip.
9:41 Ask a smart question. Content type tool demo. Build X today. What it does. Cloud
9:46 code tip. Five things I learned X. Build in public. Ship. By the way, I I'm not
9:49 going to do this. I think this was just I I actually asked chat GPT for
9:53 suggestions of what I should demo on your show and it gave me this one but
9:58 quickest wins. Pin your best demo tweet. Show the tool in action. Bio formula I
10:02 build tool one logo shipping AI tools for cloud code. >> That's not bad.
10:07 >> No, it's it's not bad. [laughter] So [snorts] all right, let's close this
10:11 one out. All right, we got cloud code best use overnight coding agent
10:14 management daily briefings and morning. Do you want to dive deeper? Okay. So now
10:18 I'm going to call in another tool. Let's plan throw in workflows plan. Take the
10:26 context above about Claudebot and propose an enterprise version that could make a lot of money.
10:39 Just do that. [laughter] Keep it simple. >> Yeah. I mean this is this is
10:42 interesting, right? like [laughter] you know you're taking a trend open
10:46 source trend and you're like okay how can I build a product that you know gets
10:52 me paid and it's interesting because there's so many trends that are
10:56 happening right now on X as like a data source and Reddit as a data source that
11:00 you know if this if this is actually good is going to help give you ideas and sort of
11:07 eventually a PRD right >> yep anyway so it's still making our plan
11:13 it's researching still doing an analysis. So, we'll we'll we'll see where that
11:18 adds up. But anyway, do you want you want to try a last 30-day prompt? Yeah.
11:22 I mean, I'm curious like I mean you you're seeing all the all these prompts.
11:25 So, like the people that listen to this show, these are people these are
11:29 founders, solopreneurs, people building businesses. Um they're always looking
11:35 for unfair advantages. Um like what sort of what sort of prompt do you would you
11:41 suggest that type of person someone who wants to build a business? Um you know
11:49 is it competitive research? Is it um you know what what are some ideas you have?
11:55 Good good question. Um I mean I think it's it's very good for competitive
11:59 research. I think it's very good for looking at at hot topics. Um, you know,
12:04 I think it's, you know, we could probably do something like, um, you
12:11 know, let's try it for your your web design idea, right? You kind of describe
12:17 poke.com, which I missed, right? So, um, what webpage designs are getting the
12:24 most love right now? Yeah, because that's interesting because if we, you
12:29 know, knowing the trend is just so important. And uh it also helps like give, you know,
12:37 when I'm when I'm designing something, it helps give me ideas. It gives it's so
12:41 overwhelming sometimes when you have like a blank page and you don't know
12:43 where to start. >> Yep. >> We've all been there. And I kind of my
12:51 kind of the the the thing that inspired this tool was I used to before I would
12:57 uh jump into any type of vibe coding session, I would go on chat GBT 5.2 and
13:04 I would just say in thinking mode and be like go uh this was for design. I would say go
13:12 on Reddit and research Nana Banana Pro best uh prompting techniques and I would do
13:18 that at the beginning and then I wouldn't even read what it would say and
13:21 then I would just be like, "Okay, give me a prompt." And so it's it's kind of
13:27 this like hack of, you know, learning kung fu, learning the prompts that
13:30 everyone's using without actually needing to read about it, if that makes
13:35 sense. you kind of just like, okay, let's just trust the the the mind crowd,
13:42 the the world that's out there and and use those best practices. All right,
13:48 we've got a plan. Let's uh I guess let's open the plan in uh BBEdit. All right,
13:54 >> so this is the plan for for the audio listeners. So this is the plan for the
13:59 the mobile >> enterprise commercial SAS platform. So again, this is not a business plan. This
14:08 is actually a plan to how to build the software architecture and I could just
14:12 then tell compound engineering to just go build it. So transform maltbot form
14:17 cloud into an open source AI into a multi-tenant enterprise SAS platform
14:21 called moltra enterprise or cloud cloud. This capitalizes on the viral growth.
14:26 GitHub stars B2B adoption enterprise AI assistance are 30 blah blah blah. Why
14:31 enterprise can't use multipart today? No multi-tensity, no RBA security
14:36 vulnerabilities, no audit logging. Uh there move wise required by service
14:39 now for market validation. proposed solution of fully managed
14:44 enterprisegrade maltbot commercial I really don't like the name
14:50 cloud cloud come on come on lm cloud architecture slack web hook discord
14:55 gateway team hooks message radar core technical components and so kind of it
15:00 used last 30 days as the basis for its expertise to learn about maltbot and
15:08 then it did the rest and now I could just say okay build it.
15:11 >> So, >> dude, that's crazy. It also that that was a pretty that was pretty dialed.
15:20 >> I'm impressed. >> And and LLMs knew nothing about
15:26 Claudebot because [laughter] it didn't exist before didn't exist till very very
15:32 recently. All right. Phase one only build a multi-tenant foundation postcress but
15:41 has full MVP. Uh I guess let's uh let's build the proof of concept. So So
15:46 basically, you know, it asked if you wanted to do an MVP. It asked if you wanted to do a
15:50 bigger build. Is that is that what I saw? >> Yeah, exactly. And so I just said,
15:55 "Okay, let's just do the the demo." >> Yeah. >> I typed something else. New repo. Also,
16:02 can you please give it a better name? Greg, you're you're you live with this
16:05 world. What should we call this thing? >> I mean, why not? I actually don't hate
16:12 uh Claude Cloud. It just Well, >> well, they got in trouble for using
16:15 claw. >> Exactly. So, why not just call it molt cloud?
16:19 >> But then we're getting in trouble with malt. [laughter]
16:24 >> Mold. Um, okay. So, why don't we call it >> Should we go lobster? [laughter]
16:30 >> Um, I was going to say, why don't we call it red lobster?
16:32 >> [laughter] >> But I think that >> All right. All right. Let's go with
16:38 Let's go with lead Speak. >> Let's go with Red Lava. >> That's cool.
16:43 >> This describes a 40week product surveys. What scope do you want to do? I don't
16:47 know. Just Just do it. >> Just do it. Why are you asking me so
16:51 many questions? >> All right. Um All right. What do we ask
16:55 here? What web designs are getting the most love right now? >> Um the most love Shopify winter edition.
17:01 Yes, I remember. I saw that one. That was really >> That was really awesome. That was
17:04 beautiful. >> Look, 3,000 likes, 320 retweets, the YC landing page. So, what's funny is I saw
17:11 this. I didn't actually look at it yet, but I saw that it was getting love and
17:14 hype. Praise for scrappy humble beginnings. Shop web landing all like
17:21 jewelry, craft works, design, all agents reported back. Reddit didn't
17:25 find anything topic too visual for text discussions, but found uh what tool do
17:29 you want to use to create designs? So it's like, okay, now that I've
17:32 researched these things, what do you what do you want to design in, right?
17:38 Um, so it's kind of it's kind of pushing you to be like, okay, I've now learned
17:42 this expertise. Now, let's do something with it. >> So So this could literally pump out a
17:49 Figma design. Is that possible? [laughter] >> Uh, I I haven't spent enough time in
17:53 Figma in in the AI era. All my Figma time was preai, so I don't know what
17:59 like prompting Figma even means in 2026, >> but uh I don't I could just say one and
18:05 see what happens. [laughter] >> Now, what are you designing? All right.
18:09 Now, it's it's diving in. >> Um a portfolio landing page for productivity app is
18:17 fine. What are the suggestions? >> It says, "What are you designing a SAS
18:23 landing page, a portfolio site, an e-commerce homepage for skincare brand
18:28 agency website hero?" And so I keep coming back to this, but
18:33 it's it is impressive how little you're giving like how short your prompts are.
18:38 >> Mhm. Here we go. Here's a prompt for Figma. Apparently, that's a thing.
18:42 >> Wow. >> 2026 forward. Design that feels warm and
18:46 human, not cold sass up layout. Anti-grid composition hero section use
18:50 asymmetrical balance headline left product screenshots floating right at
18:54 angle sections below flow organic with varied spacing not rigid 12 column
18:59 uniformity typography. One oversized display headline paired with small body
19:03 text. Add a single handdrawn underline or circle accent on the keyword.
19:07 Consider variable font like Satoshi or General Sons. Nature distilled warm
19:12 cream background. charcoal text, one muted accent, avoid harsh pure white or
19:17 saturated blues, glass morphism element, feature card. I want to see this. It's
19:22 too bad I don't have Figma AI set up. [laughter] But um so >> that's pretty cool.
19:30 >> Let's see if it can do this for like a nano banana prompt. Like, hey, you know,
19:34 I wonder if it can do that. Can you make this a nano banana prompt?
19:41 >> Yep. But I didn't ask it to become expert in >> a nano banana.
19:47 >> But I didn't ask it to become a nano banana prompt expert. So, who knows
19:50 what's gonna happen. Back to Telegram and let's see what happens. We should get three images.
20:02 Your maltbot is typing. That's a cool feeling, like watching your maltbot
20:07 type. [laughter] Someone needs to visualize that with like a lobster on a keyboard.
20:19 >> Why did you choose Telegram over iMessage or Discord or something else?
20:26 Cuz my LM told me to. Uh because I'm I'm on a a $4 a month shell account uh for
20:32 my I don't have the Mac Mac Mini setup. And so, um, the, um,
20:40 you can't use iMessage or WhatsApp, I believe, when you're just on a shell. I
20:44 think you need to run some sort of Mac software. And so, Telegram seems to be the default
20:50 where you when you don't have that. >> And for people who don't know what a
20:53 shell account is, can can you just expand on that? >> Yeah, sorry. Uh, by the way, this is our
21:00 Cloudbot competitor that it's building. I don't know what is going on or what
21:05 it's building. [laughter] It's adding a demo script showing tenant
21:09 isolation. Then oh, it already set up TypeScript and NodeJS. That's nice.
21:14 >> Wow, that's cool. [laughter] Here we go. >> Oh yeah,
21:20 >> not bad. I bet it' be nicer in Figma. >> Yeah, that was pretty solid. And again,
21:25 that was by using slast 30 days. What are the hottest webpage designs that the
21:29 world is excited about right now? what's trendy and then okay now use that to for
21:35 my SAS productivity app. I like how it like circled random words flow
21:39 effortlessly more. >> That's awesome. Yeah, I like that. It
21:43 feels like handdrawn. >> It's uh it's pretty wild. It's it's wild
21:49 times. It's uh the rate at what we're we're able to build things and how fast
21:54 Advice for Non-Engineers Getting Started with Claude Code
3:29 Live Demo: Most Popular Rap Songs
3:32 rap research. So, I found one Reddit thread, 19 expost, and the most popular rap
3:39 songs right now. So, it kind of dug in and into hip-hop heads, some expos, Billboard data, some
3:47 Spotify data, hip-hop heads, hip-hop all day, rap complex, and created this uh
3:52 this reply. >> So, someone in the comment section is going to be like, well, I could have
3:58 used perplexity or chat GBT to do that. Why should I use this?
4:04 >> Yeah. So, absolutely could, but it's uh and again I'm I'm not fully up to date
4:08 on which which tools have what access but to set up last 30 days. So you need
4:14 cloud code, right? So you're already paying for cloud code, right? So you
4:18 need that account. You need a open AI key because OpenAI has a deal with
4:23 Reddit that gives you that Reddit access. Uh I'm actually see looking into Reddit API
4:29 keys if that's an option to make the the product better. That's what I was um
4:33 that's one of the things my my weekend projects. And then the the third is you
4:39 can't search X using your X account. You have to use an XAI key. And so last 30
4:45 days pulls in all these API keys and pulls it all together if that makes
4:47 Cold Email Frameworks Demo
4:51 sense. So this is one that I did right before we started. So I
4:55 said last 30 days highest performing cold email frameworks for ICP output
5:02 three email variants subject line. So it went on X went on Twitter found
5:09 Reddit threads expost web pages and what what's interesting how I use this tool
5:14 is I often don't even read what it says. Like sure it's interesting to see what
5:17 it learned but mostly I just wanted it to write a good email. So I said, "Can
5:20 you write me some cold emails for getting on Greg Eisenberg's podcast?"
5:25 Uh, sorry I spelled your name wrong. Good target startup ideas later is all
5:27 about unconventional startup ideas, community building unique relevance. You
5:31 know, what's your angle? What's your credibility signal? Any connection
5:34 points to Greg mutual follows. What timely uh talk about AI tools I'm
5:40 working on and I once made a smart oven. Uh, smart oven details gold unexpected
5:44 founders who shipped hardware and software. That's your hook. Here are
5:48 three email variants. subject smart oven AI tools not the P Greg I went from building smart oven to
5:53 building AI tools that might sound weird until you hear it turns out the hardest
5:56 part of hardware taught me exactly what's broken and how people interact
6:00 with AI I'm building the fixed what is smart oven taught me about AI so
6:05 anyway I literally did nothing here I barely gave it any context and it had
6:12 become expert in different cold emailing concepts the three Ps framework praise
6:16 picture push ADA attention I've never read any of these in my life
6:20 intentionbased data trigger framework and then it just did it for me if that
6:24 makes sense without even me having to read the research. Yeah. I think what's
6:29 really cool about this is these frameworks. I mean, ADA has been around
6:34 for forever like you know OG David Oglevie days, but other than you know
6:39 the other ones like it feels like those are timely frameworks
6:43 like these emails like I'm reading these emails and they're I get a lot of cold
6:47 email and these are actually ones that would break through the noise. So that's
6:51 what's really cool about this. >> And I didn't even give it much effort. I
6:55 literally said, "I'm a former smart oven >> You couldn't have given it like less
7:01 context. [laughter] >> And it did. Okay. >> Yeah.
7:04 Growing an X Following Using Recent Data
7:06 >> And then another one I I just tried last 30 days how to get X followers. I
7:12 researched people actually saying um and what I learned reply is the number one
7:16 growth strategy. Multiple X power users all credit becoming a reply guy as the
7:22 fastest path to growth. shows 380 thoughtful replies, daily engagement, 40
7:25 before posting. Find larger accounts. Reply first consistently. Post at least
7:31 1x a day, 5 days a week. U Reddit threads, X threads, Instagram marketing,
7:36 Facebook ads, and share your vision. What's your account and what's
7:41 followers? Uh, I'm M. Van Horn and I made an AI tool for claw code. I don't
7:46 know what it's going to do now. [laughter] >> That's the exciting part. one one other
7:49 Researching Moltbot to Build a Competitor
7:53 let's spin up another last 30 days window. Sorry for all the all the
7:58 windows. So all right so last 30 days of research cla bot so I can
8:18 build an enterprise competitor to make money and not give it away for free.
8:22 Maybe I'm giving too much context but I should have just said research clav um
8:26 Best Practices for Last 30 days
8:27 actually I'm going to do that. I mean, what do you recommend when you're, you
8:31 know, is it shorter prompts, medium prompts, long prompts? Like, how do
8:34 getting the most out of >> I think it's a it's kind of you prime
8:38 the engine by first doing just research cloudbot top use cases. So, what my my
8:45 vision for this thread of what we're going to do is it's going to learn about
8:50 Cloudbot, right? Check X, check Reddit, and then what I'm going to do is
8:54 actually load one of my favorite tools, Compound Engineering. Uh, shout out to
8:58 to Kevin Rose for introducing me to it. And compound engineering is when I'm
9:03 starting a new project is where I do my planning. And so what I want to do here
9:10 is take this research on from last 30 days from Cloudbot and use that as the
9:19 starting point to kick off a enterprise version of Cloudbot and to build it. So
9:25 let's let's let this run and see how our other stuff is going. All right. I'm in
9:26 Growing an X Following Using Recent Data Results
9:30 I'm in A tool for cloud code. A tools are hot right now. Here's a growth
9:33 playbook for you. Morning. 40 minutes for your first post. Reply to tip from
9:37 anthropic. Alex Burch Labs. Cloud power users. Add genuine value. Share a tip.
9:41 Ask a smart question. Content type tool demo. Build X today. What it does. Cloud
9:46 code tip. Five things I learned X. Build in public. Ship. By the way, I I'm not
9:49 going to do this. I think this was just I I actually asked chat GPT for
9:53 suggestions of what I should demo on your show and it gave me this one but
9:58 quickest wins. Pin your best demo tweet. Show the tool in action. Bio formula I
10:02 build tool one logo shipping AI tools for cloud code. >> That's not bad.
10:07 >> No, it's it's not bad. [laughter] So [snorts] all right, let's close this
10:11 one out. All right, we got cloud code best use overnight coding agent
10:14 management daily briefings and morning. Do you want to dive deeper? Okay. So now
10:18 I'm going to call in another tool. Let's plan throw in workflows plan. Take the
10:26 context above about Claudebot and propose an enterprise version that could make a lot of money.
10:39 Just do that. [laughter] Keep it simple. >> Yeah. I mean this is this is
10:42 interesting, right? like [laughter] you know you're taking a trend open
10:46 source trend and you're like okay how can I build a product that you know gets
10:52 me paid and it's interesting because there's so many trends that are
10:56 happening right now on X as like a data source and Reddit as a data source that
11:00 you know if this if this is actually good is going to help give you ideas and sort of
11:07 eventually a PRD right >> yep anyway so it's still making our plan
11:13 it's researching still doing an analysis. So, we'll we'll we'll see where that
11:18 adds up. But anyway, do you want you want to try a last 30-day prompt? Yeah.
11:22 I mean, I'm curious like I mean you you're seeing all the all these prompts.
11:25 So, like the people that listen to this show, these are people these are
11:29 founders, solopreneurs, people building businesses. Um they're always looking
11:35 for unfair advantages. Um like what sort of what sort of prompt do you would you
11:41 suggest that type of person someone who wants to build a business? Um you know
11:49 is it competitive research? Is it um you know what what are some ideas you have?
11:55 Good good question. Um I mean I think it's it's very good for competitive
11:59 research. I think it's very good for looking at at hot topics. Um, you know,
12:04 I think it's, you know, we could probably do something like, um, you
12:11 know, let's try it for your your web design idea, right? You kind of describe
12:17 poke.com, which I missed, right? So, um, what webpage designs are getting the
12:24 most love right now? Yeah, because that's interesting because if we, you
12:29 know, knowing the trend is just so important. And uh it also helps like give, you know,
12:37 when I'm when I'm designing something, it helps give me ideas. It gives it's so
12:41 overwhelming sometimes when you have like a blank page and you don't know
12:43 where to start. >> Yep. >> We've all been there. And I kind of my
12:51 kind of the the the thing that inspired this tool was I used to before I would
12:57 uh jump into any type of vibe coding session, I would go on chat GBT 5.2 and
13:04 I would just say in thinking mode and be like go uh this was for design. I would say go
13:12 on Reddit and research Nana Banana Pro best uh prompting techniques and I would do
13:18 that at the beginning and then I wouldn't even read what it would say and
13:21 then I would just be like, "Okay, give me a prompt." And so it's it's kind of
13:27 this like hack of, you know, learning kung fu, learning the prompts that
13:30 everyone's using without actually needing to read about it, if that makes
13:35 sense. you kind of just like, okay, let's just trust the the the mind crowd,
13:42 the the world that's out there and and use those best practices. All right,
13:48 we've got a plan. Let's uh I guess let's open the plan in uh BBEdit. All right,
13:54 >> so this is the plan for for the audio listeners. So this is the plan for the
13:59 the mobile >> enterprise commercial SAS platform. So again, this is not a business plan. This
14:08 is actually a plan to how to build the software architecture and I could just
14:12 then tell compound engineering to just go build it. So transform maltbot form
14:17 cloud into an open source AI into a multi-tenant enterprise SAS platform
14:21 called moltra enterprise or cloud cloud. This capitalizes on the viral growth.
14:26 GitHub stars B2B adoption enterprise AI assistance are 30 blah blah blah. Why
14:31 enterprise can't use multipart today? No multi-tensity, no RBA security
14:36 vulnerabilities, no audit logging. Uh there move wise required by service
14:39 now for market validation. proposed solution of fully managed
14:44 enterprisegrade maltbot commercial I really don't like the name
14:50 cloud cloud come on come on lm cloud architecture slack web hook discord
14:55 gateway team hooks message radar core technical components and so kind of it
15:00 used last 30 days as the basis for its expertise to learn about maltbot and
15:08 then it did the rest and now I could just say okay build it.
15:11 >> So, >> dude, that's crazy. It also that that was a pretty that was pretty dialed.
15:20 >> I'm impressed. >> And and LLMs knew nothing about
15:26 Claudebot because [laughter] it didn't exist before didn't exist till very very
15:32 recently. All right. Phase one only build a multi-tenant foundation postcress but
15:41 has full MVP. Uh I guess let's uh let's build the proof of concept. So So
15:46 basically, you know, it asked if you wanted to do an MVP. It asked if you wanted to do a
15:50 bigger build. Is that is that what I saw? >> Yeah, exactly. And so I just said,
15:55 "Okay, let's just do the the demo." >> Yeah. >> I typed something else. New repo. Also,
16:02 can you please give it a better name? Greg, you're you're you live with this
16:05 world. What should we call this thing? >> I mean, why not? I actually don't hate
16:12 uh Claude Cloud. It just Well, >> well, they got in trouble for using
16:15 claw. >> Exactly. So, why not just call it molt cloud?
16:19 >> But then we're getting in trouble with malt. [laughter]
16:24 >> Mold. Um, okay. So, why don't we call it >> Should we go lobster? [laughter]
16:30 >> Um, I was going to say, why don't we call it red lobster?
16:32 >> [laughter] >> But I think that >> All right. All right. Let's go with
16:38 Let's go with lead Speak. >> Let's go with Red Lava. >> That's cool.
16:43 >> This describes a 40week product surveys. What scope do you want to do? I don't
16:47 know. Just Just do it. >> Just do it. Why are you asking me so
16:51 many questions? >> All right. Um All right. What do we ask
16:55 here? What web designs are getting the most love right now? >> Um the most love Shopify winter edition.
17:01 Yes, I remember. I saw that one. That was really >> That was really awesome. That was
17:04 beautiful. >> Look, 3,000 likes, 320 retweets, the YC landing page. So, what's funny is I saw
17:11 this. I didn't actually look at it yet, but I saw that it was getting love and
17:14 hype. Praise for scrappy humble beginnings. Shop web landing all like
17:21 jewelry, craft works, design, all agents reported back. Reddit didn't
17:25 find anything topic too visual for text discussions, but found uh what tool do
17:29 you want to use to create designs? So it's like, okay, now that I've
17:32 researched these things, what do you what do you want to design in, right?
17:38 Um, so it's kind of it's kind of pushing you to be like, okay, I've now learned
17:42 this expertise. Now, let's do something with it. >> So So this could literally pump out a
17:49 Figma design. Is that possible? [laughter] >> Uh, I I haven't spent enough time in
17:53 Figma in in the AI era. All my Figma time was preai, so I don't know what
17:59 like prompting Figma even means in 2026, >> but uh I don't I could just say one and
18:05 see what happens. [laughter] >> Now, what are you designing? All right.
18:09 Now, it's it's diving in. >> Um a portfolio landing page for productivity app is
18:17 fine. What are the suggestions? >> It says, "What are you designing a SAS
18:23 landing page, a portfolio site, an e-commerce homepage for skincare brand
18:28 agency website hero?" And so I keep coming back to this, but
18:33 it's it is impressive how little you're giving like how short your prompts are.
18:38 >> Mhm. Here we go. Here's a prompt for Figma. Apparently, that's a thing.
18:42 >> Wow. >> 2026 forward. Design that feels warm and
18:46 human, not cold sass up layout. Anti-grid composition hero section use
18:50 asymmetrical balance headline left product screenshots floating right at
18:54 angle sections below flow organic with varied spacing not rigid 12 column
18:59 uniformity typography. One oversized display headline paired with small body
19:03 text. Add a single handdrawn underline or circle accent on the keyword.
19:07 Consider variable font like Satoshi or General Sons. Nature distilled warm
19:12 cream background. charcoal text, one muted accent, avoid harsh pure white or
19:17 saturated blues, glass morphism element, feature card. I want to see this. It's
19:22 too bad I don't have Figma AI set up. [laughter] But um so >> that's pretty cool.
19:30 >> Let's see if it can do this for like a nano banana prompt. Like, hey, you know,
19:34 I wonder if it can do that. Can you make this a nano banana prompt?
19:41 >> Yep. But I didn't ask it to become expert in >> a nano banana.
19:47 >> But I didn't ask it to become a nano banana prompt expert. So, who knows
19:50 what's gonna happen. Back to Telegram and let's see what happens. We should get three images.
20:02 Your maltbot is typing. That's a cool feeling, like watching your maltbot
20:07 type. [laughter] Someone needs to visualize that with like a lobster on a keyboard.
20:19 >> Why did you choose Telegram over iMessage or Discord or something else?
20:26 Cuz my LM told me to. Uh because I'm I'm on a a $4 a month shell account uh for
20:32 my I don't have the Mac Mac Mini setup. And so, um, the, um,
20:40 you can't use iMessage or WhatsApp, I believe, when you're just on a shell. I
20:44 think you need to run some sort of Mac software. And so, Telegram seems to be the default
20:50 where you when you don't have that. >> And for people who don't know what a
20:53 shell account is, can can you just expand on that? >> Yeah, sorry. Uh, by the way, this is our
21:00 Cloudbot competitor that it's building. I don't know what is going on or what
21:05 it's building. [laughter] It's adding a demo script showing tenant
21:09 isolation. Then oh, it already set up TypeScript and NodeJS. That's nice.
21:14 >> Wow, that's cool. [laughter] Here we go. >> Oh yeah,
21:20 >> not bad. I bet it' be nicer in Figma. >> Yeah, that was pretty solid. And again,
21:25 that was by using slast 30 days. What are the hottest webpage designs that the
21:29 world is excited about right now? what's trendy and then okay now use that to for
21:35 my SAS productivity app. I like how it like circled random words flow
21:39 effortlessly more. >> That's awesome. Yeah, I like that. It
21:43 feels like handdrawn. >> It's uh it's pretty wild. It's it's wild
21:49 times. It's uh the rate at what we're we're able to build things and how fast
21:54 things change is absolutely wild. >> So Matt to to this has been fun. um you
22:00 know if people want to get started they want to like what advice do you have you
22:07 know in terms of getting the most out of this product? >> Yeah. So one of the things I'll say is I
22:13 uh I am not a software engineer. I have not shipped anything of value since high
22:18 school and could argue the web pages I was shipping in high school were not of
22:22 value either. But at least I could write some some basic HTML or some basic
22:28 scripts. And so what's um what's amazing is being uh not a software engineer and
22:37 being able to go into cloud code. Like it seems intimidating. You're in the
22:43 terminal. Setting up a cloudbot seems intimidating. And my recommendation for
22:49 starting out is set up set up claude code. It's magical. Sign up for the $20
22:55 account. if you'll be on the $100 and the $200 account probably real soon if
23:01 you're uh if you're successful. Um, and and what I what I do is I um have keep a
23:09 chat GPT window open to ask questions to. And so I'm kind of moving back and
23:13 forth and posting lots of screenshots be like there's some error. I don't know
23:17 what's going on. Help me. And so chat GBT52 thinking will be like, "Okay,
23:21 great. In your in your shell, in your terminal, this is what's going on. This
23:25 is why it broke." And I'm like, "Okay, great. What do I do? Help." And then
23:28 it's just like, "Copy this into the terminal." I'm like, "Okay." And I again
23:31 don't even have to read it, don't have to overthink it. And so kind of just
23:35 like screenshot trial and error back and forth between chat GPT and my terminal.
23:39 And like I couldn't figure out actually how to post screenshots into my
23:42 terminal. So I asked chat GPT and it's like, oh, you have to use controlV
23:47 instead of commandV. And once that that was the biggest unlock for me in the
23:50 terminal, figuring out that I could post screenshots with controlV. And anyway,
23:56 so I recommend getting into cloud code and and try last 30 days obviously uh
24:03 because it it gives you superpowers, but use tools like compound engineering, use
24:08 skills like uh there's another one called superpowers which is getting a
24:12 lot of love right now and you just kind of enter these plan modes and you try
24:18 and you build things. And I had this idea for the last 30 days and I've never
24:24 looked at the code for the skill that I wrote quote unquote. And it was just a
24:28 lot of trial and error and running more and more terminal windows and testing it
24:32 and testing it and being like, "Hey, figure out how I can get access to this
24:34 API. Hey, figure out how I could do this. Hey, how can my users do this?"
24:38 And going from there. >> I love that, man. Well, I'll include uh
24:43 a link to last 30 days in the show notes, in the description, link where
24:49 where you can follow Matt Van Horn on X. And dude, this is this is really cool
24:53 and it's crazy that this is just sort of like a project that you started and it's
24:59 gaining so much traction and I think people are just, you know, obsessed with
25:04 Claude Code for, you know, for good reason. Um, but I think people are just trying to get the
25:09 most out of it. So, this this came at a a very perfect time and and uh can't
25:14 wait to see how it evolves. And thanks for doing a little show and tell for us.
25:17 >> Of course. It was fun being here. Thank you so much, Craig.
