// transcript — 461 segments
0:00 Intro
0:03 2025 has been an absolutely crazy year for AI. So, right now, I'm taking on the
0:08 challenge to quickly cover for you everything that has happened with AI
0:12 this year. And so, I'll be hitting on the news that points to the trends going
0:16 into 2026 and who the big players will be. And I can guarantee that I have
0:21 missed nothing on this list. Just kidding, of course. I probably missed
0:25 some things. So, definitely let me know in the comments as I go through
0:28 everything what you think should be added to this list. We have incredible
0:32 new tools and LLMs that have been released. AI coding assistants are
0:36 taking over the world. Companies are raising billions of dollars left and
0:40 right with massive acquisitions. A single tool is being used by 90% of
0:45 Fortune 500 companies now. And of course, we have the Chinese startup that
0:50 wiped $600 billion off the Nvidia market cap in a single day. So, just a teaser
0:55 for what we have here, but I've got 50 items in the list. Let's go ahead and
0:58 dive right into it. So, I'm going to start with the things that apply to the
1:02 entire year of 2025. Then, I'll get into the specific events in chronological
1:07 order as they happened in 2025. And of course, I will link to all these sources
1:12 in the description. The first baffling number out of many that we have here is
1:17 that companies have invested $ 37 billion into enterprise AI this year.
1:23 That figure was only 1.7 billion in 2023. Also, 84% of developers are using
1:29 or planning on using AI tools in their development process. And out of all
1:33 developers, about half are already using AI coding assistance daily. In 2025, AI
1:40 startups have captured almost 50% of all global funding. Just take a look at how
1:44 this has evolved from 2021. It's absolutely insane. And speaking of
1:49 funding, this year, 49 US AI startups have raised at least $100 million. And
1:55 of course, we have the Chinese startup that I alluded to earlier, the release
2:00 of Deepseek R1. At the time, a very revolutionary open- source large
2:03 language model. And this is highly debated, but there are a lot of claims
2:06 that this is the reason the Nvidia market cap went down $600 billion, the
2:12 largest one-day loss for a single company in the US stock market history.
2:16 In January, we also had the announcement of the Stargate project, a company
2:22 formed to invest $500 billion dollars over the next four years in building out
2:27 new AI infrastructure for OpenAI. And speaking of OpenAI, soon after the
2:30 Stargate announcement, they also released Operator, the agent that uses
2:34 its own browser to perform millions of different tasks for you. Now, whether or
2:38 not this lived up to the hype, I'll leave that up to you to decide, but it
2:41 was a big deal at the start of this year. Going into February now, the AI
2:46 engineer Andre Karpathy coined the term Vibe coding and it has gained so much
2:51 popularity that it literally has its own Wikipedia page now. It's official. Now,
2:55 AI regulation in general is a bit too in the weeds for me to cover here, but
2:59 there is one instance that deserves its spot on the list, and that is the EU AI
3:04 Act, the first ever comprehensive legal framework on AI worldwide. And we're
3:08 definitely going to see a lot more regulation like this going into 2026.
3:13 And now for the first of many funding rounds and valuations that I'll cover in
3:17 this video. We have Harvey, the legal AI startup, raising $300 million in their
3:22 series D. A lot of people don't realize this, but the legal industry is a gold
3:26 mine of opportunity for generative AI. Now, March was a very big month for
3:30 generative AI, especially because that's when Anthropic's model context protocol
3:34 really took off. But that doesn't belong in this list because it was originally
3:39 created at the very end of 2024. So for the first March item, we have Manis, the
3:44 first of its kind general AI agent that really was the pioneer for a lot of
3:48 strategies around context engineering that we have now and strategies for
3:53 building agents that can handle more longunning tasks. Google also released
3:57 Gemma 3, their next open source large language models built on the same
4:02 technology and research that powered their Gemini 2.0 models. And speaking of
4:07 Gemini, Google released Gemini 2.5 later this month as well. And this model was
4:12 revolutionary because of how well it handled a large amount of context. And
4:17 it stood the test of time for quite a while. Generally considered to be the
4:21 best general purpose large language model for many months. And as 2025
4:26 progressed, there were so many insane valuations and acquisitions and funding
4:31 rounds. Starting now with the very end of March, OpenAI raised $40 billion,
4:37 which put them at a 300 billion post money valuation. Now, going into the
4:42 month of April, Meta released their Llama 4 models. They have their herd
4:46 here. Three different options that we have, Behemoth, Maverick, and Scout.
4:50 Now, we haven't actually seen much from Meta since then. And so, I really hope
4:54 that going into 2026, we'll get maybe Llama 5. I really hope that Meta picks
4:59 up the slack here. Then we have OpenAI releasing their 03 and 04 mini models,
5:04 continuing to push the limits for what is possible with the reasoning
5:07 capabilities built into these LLMs that is so common now. And then going back to
5:12 open source, Alibaba released their Quen 3 models at the very end of April. And
5:17 these are still some of my favorite open- source large language models to
5:21 use. They really have stood the test of time. Now, going into May, OpenAI
5:26 introduced Codeex, their cloud-based software engineering agent that
5:29 integrates directly with GitHub, so you can work on your repos remotely. You
5:33 don't even have to be in your local development environment. They also
5:36 released the Codeex CLI later. This still is one of my favorite platforms
5:41 for AI assisted coding. We also had the release of V3 from Google. This is when
5:46 AI video generation started to finally actually feel good. OpenAI agrees to buy
5:51 Windinsurf for $3 billion only for the deal to fall apart a couple months later
5:56 in July with Windsurf CEO going to Google. It pretty much seems like Google
6:00 acquired Windsurf especially with all the features that come out in
6:05 anti-gravity, Google's new agentic IDE that was released along with Gemini 3
6:09 that we'll talk about later. Enthropic released their family of Claude 4
6:12 models. This is when Claude really became the coding king for me and I got
6:17 obsessed with using Claude code. It still in my mind is the best agentic
6:22 coding assistant. Now going into June, Cursor reached $500 million in annually
6:28 recurring revenue. At this point, it was very clear that Cursor was the fastest
6:33 growing platform and still is in SAS history. Also this month, Mark
6:36 Zuckerberg announced the creation of the Meta Super Intelligence Labs. This is
6:41 when we had all the drama of Mark Zuckerberg apparently paying all of
6:45 these 9 figure salaries to poach top researchers from other companies. Now
6:49 going into July, Amazon has released their own coding agent, Kira, which is
6:53 an AI coding assistant that I'm starting to respect more and more. And I heard
6:58 that after their reinvent conference in December, they are starting to get a lot
7:02 of enterprise interest. And at the end of July, open source large language
7:06 models continue to fight back. Moonshot released at Kimmy K2 which apparently
7:10 according to the benchmarks beat chat GBT and cloud encoding. This is very
7:15 much up for debate but I still do like seeing open source LLMs pushing the
7:19 boundaries. All right, so we have been blitzing through a lot already. So let's
7:23 take a quick break for our sponsor and then we'll get back to finishing off our
7:27 list. So the sponsor of this video is Zapier which is actually my original
7:32 noode automation tool. They've built a lot in the AI space this year, including
7:37 the Zapier MCP server, which allows you to connect your agent to over 8,000
7:41 different applications. So, literally, no matter the capabilities that you want
7:44 to build into your agent, you can do it with the Zapier MCP. Take a look at how
7:49 easy this is. First, all you have to do is set up the app connections in Zapier.
7:52 So, you go through the OAF flow or whatever to get your credentials like
7:56 for Google Drive connected. Then, you go into the MCP server tab, which I'll link
7:59 to in the description to set up your server. You pick your client and then
8:03 you pick your tools. You can search through all of the services that are
8:06 available. Like I can do Google Drive for example. You pick the individual
8:10 tools that you want in the MCP and you configure all of the parameters. Once
8:14 you have that set up, then you go to the connect tab and right here there's a
8:17 button to just add it directly into Cloud Desktop or for other MCP clients
8:22 you can just copy this URL and set it up just like you would a typical streamable
8:27 HTTP MCP server. And so for example in cloud desktop I'll go to the settings in
8:32 the top right and then I will go to the connectors and then I can add a custom
8:35 connector and I just have to paste in the URL that I copied here and I can
8:39 call it Zapier and then boom you are good to go. It'll take you through an
8:42 OOTH flow with Zapier so that it's nice and secure but then you will immediately
8:46 have it available as capabilities to cloud desktop. And so for a very quick
8:51 demonstration I used the Zapier MCP server in cloud desktop to find my
8:54 Google Drive folder and create a file here. So you can see both tool calls,
8:58 everything's working perfectly, and we even have the file at the bottom here so
9:02 I can click right into it in my Google Drive. And it was so easy to set this
9:06 all up. And so I will have a link in the description to the Zapier MCP server.
9:10 Definitely check it out. Now, surprisingly, August was a pretty quiet
9:14 month for Generative AI. The biggest thing that we had was the introduction
9:19 of GBT5 from Open AI. Getting into September now, we have more funding. The
9:24 funding train never stops. Anthropic raises $13 billion, putting it now at a
9:31 $183 billion valuation. Now, there have been a lot of lawsuits in the AI space
9:35 this year, but one of the interesting ones that happened more recently is a
9:39 group of authors sued Anthropic for using their books and other materials in
9:43 their training data. And Enthropic ended up paying a whopping 1.5 billion
9:48 settlement. Now, I don't mean to just pick on Anthropic here. There are a lot
9:53 of different lawsuits for all of the LLM providers, but it is crazy. They're
9:56 paying around $3,000 for each of the estimated 500,000 books covered by the
10:00 29. Perplexity Valuation Reaches $20B
10:01 settlement. Also, in September, Perplexity, the AI search startup,
10:06 raises $200 million at a $20 billion valuation. It's very impressive to me
10:10 that Perplexity has stayed on the cutting edge and remained competitive,
10:13 even though people think time and time again that all of the common agents and
10:18 LLM are going to replace Perplexity with their built-in search. And continuing
10:19 30. Mistral Raises 1.7B Euros
10:22 the funding train, Mistl, the French AI startup, raised€1.7 billion, and they
10:27 continue to put out some of my favorite open- source LLMs and a lot of their
10:31 31. OpenAI Releases Sora 2 Video and Audio Generation
10:31 closed source stuff is really powerful as well. And finally, to end things off,
10:36 in September, OpenAI released Sora 2. And this, my friend, took the world by
10:40 storm. The capabilities here for both the video and audio generation, it's
10:44 32. ChatGPT Hits 800M Weekly Active Users (October)
10:45 just insanely cool. And continuing with OpenAI going into October, Sam Alman
10:51 said that in October, ChatGpt hit 800 million weekly active users. I wish my
10:56 applications had that many weekly active users. Also, IBM and Enthropic announced
10:57 33. IBM and Anthropic Enterprise Partnership
11:01 a partnership to accelerate the development of enterprise ready AI. I
11:05 was actually there at the IBM Tech Exchange conference when they announced
11:08 this. Very, very cool. Anthropic is definitely continuing to push themselves
11:13 as the enterprise LLM provider especially for coding. Now getting into
11:15 34. Microsoft Announces MAI Superintelligence Team (November)
11:17 November, Microsoft announced the formation of the MAI super intelligence
11:22 team. What they're working towards here is humanist super intelligence.
11:26 Incredibly advanced AI capabilities that always work for in service of people and
11:31 humanity more generally. Now this is kind of vague to me right now, but I
11:34 think we're going to find out a lot more about this going into 2026. Anthropic
11:36 35. Anthropic Releases Claude Code for the Web
11:39 releases Cloud Code for the web. Certainly a competitor to OpenAI's
11:43 codecs that I talked about earlier. Being able to work on your repositories
11:47 with agents running remotely, so you're not even in your local development
11:50 environment. I definitely think this is the future of Agentic Coding. Something
11:54 I'm even working on myself, building custom systems for remote Aentic coding.
11:58 36. Google Releases Nano Banana Pro Image Generation
12:00 Then Google takes AI image generation to the next level with Nano Banana,
12:04 especially Nano Banana Pro, which was released at the end of November. It is
12:08 insane the quality of images that we can generate with AI and especially how
12:13 37. Microsoft, NVIDIA, and Anthropic Strategic Partnerships
12:13 realistic they look now. Microsoft, Nvidia and Anthropic all announce
12:18 partnerships. So Anthropic to scale Claude on Azure, Anthropic to adopt
12:22 Nvidia architecture and Nvidia and Microsoft to invest in Anthropic. So
12:26 definitely a bit of a circular investment here which speaks a little
12:30 bit to the possibility of an AI bubble. But it makes sense for all these big
12:34 38. Launch of Google Gemini 3
12:34 companies to be working together. And we right now are only halfway through
12:37 November. There is so much that happened last month. Next, we have the release of
12:42 Gemini 3 from Google, which is often considered to be the most powerful LLM
12:46 from a general standpoint right now. It absolutely crushes it on the benchmarks
12:51 39. MS 365 Copilot Adopted by 90% of Fortune 500
12:51 and handles long context tasks very well. Now, at the start of this video, I
12:54 alluded to the fact that there is a single application that is already
12:59 adopted by over 90% of Fortune 500 companies, and that application is
13:04 Microsoft 365 C-Pilot. And I believe they reached that 90% mark in November
13:09 of last year. Just absolutely crazy. The enterprise adoption for a single
13:12 40. Anthropic Acquires Bun JavaScript Runtime (December)
13:14 product. Enthropic continues to crush it with Cloud Code. Cloud Code reaches $1
13:19 billion in revenue and Enthropic acquires Bun, the JavaScript runtime to
13:23 continue to make Cloud Code more and more scalable because let me tell you,
13:28 41. Anthropic Releases Claude Opus 4.5
13:28 the demand for it is incredible. On top of that, Enthropic releases Claude Opus
13:32 4.5, which most people will tell you that it's not even a competition. It is
13:38 the AI coding king. Sonnet 4.5 was already amazing, but now Opus is just
13:42 42. Release of DeepSeek Version 3.2
13:42 crushing it for AI coding. Now, there hasn't been that much news around
13:45 DeepSeek since the start of the year, but now going into December, they
13:49 released Deepseek version 3.2, an open- source large language model that
13:52 competes with all of the big guys, Gemini, Claude, and GPT. Now, this is
13:58 open source, but it's such a large model that most people can't actually run on
14:02 their own hardware, which is a bit of a bummer, but unfortunately, that is what
14:06 it takes right now to compete with all of the top dogs. And continuing with
14:07 43. Release of GLM-4.7 for AI Coding
14:10 open source LLMs, we also have the release of GLM 4.7, an LLM that is
14:15 designed specifically to be very powerful for AI assisted coding. And in
14:19 their benchmarks, it compares extremely well to the DeepS version 3.2 2. We just
14:26 covered CloudSonet 4.5, GPT 5.1. The point of these open- source models is
14:30 that they're almost as powerful as the big guys, but they are a lot faster and
14:33 44. Anthropic Donates MCP to Agentic AI Foundation
14:34 cheaper. So, Anthropic's model context protocol really took off earlier this
14:37 year, and now in December, they announced that they're donating MCP as a
14:42 part of establishing the Agentic AI Foundation. This is a directed fund
14:47 under the Linux Foundation co-founded by Anthropic Block and OpenAI with support
14:51 from a lot of other absolutely massive companies. And so hopefully what this
14:55 means is that all these companies are starting to work together to build out
14:59 45. Anthropic Releases Claude Agent Skills
14:59 AI for the greater good. Now, as much as MCP is incredible, Enthropic also
15:04 recently released Claude Skills. They did this back in October. It's really
15:08 started to gain traction this month in December though and it's becoming a
15:13 replacement in a lot of ways for MCP cuz it's a lot more context optimized. And
15:17 so it's this idea of progressive disclosure where we only give
15:21 capabilities to our agents when it actually needs them. So we save a lot of
15:26 context up front. This kind of paradigm is going to be very popular going into
15:30 2026 for building any kind of agentic system. Definitely not limited to the
15:34 46. OpenAI Releases GPT-5.2
15:34 claude ecosystem either. OpenAI releases GPT 5.2. Definitely a lot of
15:40 improvements over GPT 5.1. Now, Enthropic with Claude seems to be quite
15:44 focused on AI coding as the specialization of the LLMs, but GPT is
15:49 more general purpose like Gemini, so seems to be kind of competing more with
15:52 47. Disney Invests $1B in OpenAI (Sora Focus)
15:54 Gemini 3. OpenAI also gets a $1 billion investment from Disney, especially
15:58 thanks to Sora. definitely showing that there is a lot of use case and demand
16:04 48. Databricks Raises $4B Series L at $134B Valuation
16:04 for AI generated videos with audio. So, we haven't talked about data companies
16:07 that much in this video, but they're certainly doing well, too. Datab bricks
16:12 is doing their over $4 billion series L and is valued now at 134 billion. They
16:18 also acquired Neon recently, one of my favorite database platforms. They're
16:22 49. NVIDIA Buys Groq Assets for $20B
16:22 definitely rocking it. And then at the very end of this year, Nvidia announced
16:26 that they are buying Gro's assets for about $20 billion. Now, the title is
16:31 kind of strange cuz it doesn't say it's an acquisition, but it definitely feels
16:34 like an acquisition when you really dive into it. So, definitely a lot of
16:38 interesting things that will unfold here going into January as well. All right,
16:39 50: Meta Announces Acquisition of Manus
16:41 to end things off for us here, I literally have an article dated December
16:46 30th of this year. Meta is going to acquire Manis. And it's cool cuz I
16:49 talked about both of these companies already. Mannis was very revolutionary
16:52 when it was released at the start of this year and Meta hasn't made a splash
16:57 in the generative AI space in a while now and so it definitely seems like
17:00 they're doing some things behind the scenes that we might find out about more
17:05 maybe quarter 1 or quarter two of next year and so they're going to be
17:08 integrating Madness into its products including Meta AI. All right, so that my
1:10 1. Enterprise AI Investment Surge ($37B)
1:12 in the description. The first baffling number out of many that we have here is
1:17 that companies have invested $ 37 billion into enterprise AI this year.
1:23 That figure was only 1.7 billion in 2023. Also, 84% of developers are using
1:24 2. Mass Developer Adoption of AI Tools (84%)
1:29 or planning on using AI tools in their development process. And out of all
1:33 developers, about half are already using AI coding assistance daily. In 2025, AI
1:35 3. AI Startups Capturing 50% of Global Funding
1:40 startups have captured almost 50% of all global funding. Just take a look at how
1:44 this has evolved from 2021. It's absolutely insane. And speaking of
1:46 4. Mega Funding Rounds ($100M+) for AI Startups
1:49 funding, this year, 49 US AI startups have raised at least $100 million. And
1:53 5. Release of DeepSeek-R1 (January)
1:55 of course, we have the Chinese startup that I alluded to earlier, the release
2:00 of Deepseek R1. At the time, a very revolutionary open- source large
2:03 language model. And this is highly debated, but there are a lot of claims
2:06 that this is the reason the Nvidia market cap went down $600 billion, the
2:12 largest one-day loss for a single company in the US stock market history.
2:15 6. Announcement of the Stargate Project ($500B)
2:16 In January, we also had the announcement of the Stargate project, a company
2:22 formed to invest $500 billion dollars over the next four years in building out
2:26 7. Launch of OpenAI Operator Agent
2:27 new AI infrastructure for OpenAI. And speaking of OpenAI, soon after the
2:30 Stargate announcement, they also released Operator, the agent that uses
2:34 its own browser to perform millions of different tasks for you. Now, whether or
2:38 not this lived up to the hype, I'll leave that up to you to decide, but it
2:41 was a big deal at the start of this year. Going into February now, the AI
2:42 8. Coining of "Vibe Coding" by Karpathy
2:46 engineer Andre Karpathy coined the term Vibe coding and it has gained so much
2:51 popularity that it literally has its own Wikipedia page now. It's official. Now,
2:53 9. Implementation of the EU AI Act
2:55 AI regulation in general is a bit too in the weeds for me to cover here, but
2:59 there is one instance that deserves its spot on the list, and that is the EU AI
3:04 Act, the first ever comprehensive legal framework on AI worldwide. And we're
3:08 definitely going to see a lot more regulation like this going into 2026.
3:11 10. Harvey Raises $300M Series D
3:13 And now for the first of many funding rounds and valuations that I'll cover in
3:17 this video. We have Harvey, the legal AI startup, raising $300 million in their
3:22 series D. A lot of people don't realize this, but the legal industry is a gold
3:26 mine of opportunity for generative AI. Now, March was a very big month for
3:27 11. Launch of Manus General AI Agent (March)
3:30 generative AI, especially because that's when Anthropic's model context protocol
3:34 really took off. But that doesn't belong in this list because it was originally
3:39 created at the very end of 2024. So for the first March item, we have Manis, the
3:44 first of its kind general AI agent that really was the pioneer for a lot of
3:48 strategies around context engineering that we have now and strategies for
3:53 building agents that can handle more longunning tasks. Google also released
3:54 12. Google Releases Gemma 3 Open Source Models
3:57 Gemma 3, their next open source large language models built on the same
4:02 technology and research that powered their Gemini 2.0 models. And speaking of
4:04 13. Release of Google Gemini 2.5
4:07 Gemini, Google released Gemini 2.5 later this month as well. And this model was
4:12 revolutionary because of how well it handled a large amount of context. And
4:17 it stood the test of time for quite a while. Generally considered to be the
4:21 best general purpose large language model for many months. And as 2025
4:23 14. OpenAI Raises $40B at $300B Valuation
4:26 progressed, there were so many insane valuations and acquisitions and funding
4:31 rounds. Starting now with the very end of March, OpenAI raised $40 billion,
4:37 which put them at a 300 billion post money valuation. Now, going into the
4:39 15. Meta Releases Llama 4 "Herd" (April)
4:42 month of April, Meta released their Llama 4 models. They have their herd
4:46 here. Three different options that we have, Behemoth, Maverick, and Scout.
4:50 Now, we haven't actually seen much from Meta since then. And so, I really hope
4:54 that going into 2026, we'll get maybe Llama 5. I really hope that Meta picks
4:58 16. OpenAI Releases o3 and o4-mini Reasoning Models
4:59 up the slack here. Then we have OpenAI releasing their 03 and 04 mini models,
5:04 continuing to push the limits for what is possible with the reasoning
5:07 capabilities built into these LLMs that is so common now. And then going back to
5:08 17. Alibaba Releases Qwen3 Open Source Models
5:12 open source, Alibaba released their Quen 3 models at the very end of April. And
5:17 these are still some of my favorite open- source large language models to
5:21 use. They really have stood the test of time. Now, going into May, OpenAI
5:22 18. OpenAI Introduces Codex Agent (May)
5:26 introduced Codeex, their cloud-based software engineering agent that
5:29 integrates directly with GitHub, so you can work on your repos remotely. You
5:33 don't even have to be in your local development environment. They also
5:36 released the Codeex CLI later. This still is one of my favorite platforms
5:40 19. Google Releases Veo 3 Video Generation
5:41 for AI assisted coding. We also had the release of V3 from Google. This is when
5:46 AI video generation started to finally actually feel good. OpenAI agrees to buy
5:48 20. OpenAI Agreement to Buy Windsurf
5:51 Windinsurf for $3 billion only for the deal to fall apart a couple months later
5:56 in July with Windsurf CEO going to Google. It pretty much seems like Google
6:00 acquired Windsurf especially with all the features that come out in
6:05 anti-gravity, Google's new agentic IDE that was released along with Gemini 3
6:08 21. Anthropic Releases Claude 4 Family
6:09 that we'll talk about later. Enthropic released their family of Claude 4
6:12 models. This is when Claude really became the coding king for me and I got
6:17 obsessed with using Claude code. It still in my mind is the best agentic
6:21 22. Cursor Reaches $500M ARR (June)
6:22 coding assistant. Now going into June, Cursor reached $500 million in annually
6:28 recurring revenue. At this point, it was very clear that Cursor was the fastest
6:33 23. Mark Zuckerberg Announces Meta Superintelligence Labs
6:33 growing platform and still is in SAS history. Also this month, Mark
6:36 Zuckerberg announced the creation of the Meta Super Intelligence Labs. This is
6:41 when we had all the drama of Mark Zuckerberg apparently paying all of
6:45 these 9 figure salaries to poach top researchers from other companies. Now
6:47 24. Amazon Releases Kiro Coding Agent (July)
6:49 going into July, Amazon has released their own coding agent, Kira, which is
6:53 an AI coding assistant that I'm starting to respect more and more. And I heard
6:58 that after their reinvent conference in December, they are starting to get a lot
7:01 25. Moonshot Releases Kimi K2 Open Source Model
7:02 of enterprise interest. And at the end of July, open source large language
7:06 models continue to fight back. Moonshot released at Kimmy K2 which apparently
7:10 according to the benchmarks beat chat GBT and cloud encoding. This is very
7:15 much up for debate but I still do like seeing open source LLMs pushing the
7:18 Zapier MCP Server
7:19 boundaries. All right, so we have been blitzing through a lot already. So let's
7:23 take a quick break for our sponsor and then we'll get back to finishing off our
7:27 list. So the sponsor of this video is Zapier which is actually my original
7:32 noode automation tool. They've built a lot in the AI space this year, including
7:37 the Zapier MCP server, which allows you to connect your agent to over 8,000
7:41 different applications. So, literally, no matter the capabilities that you want
7:44 to build into your agent, you can do it with the Zapier MCP. Take a look at how
7:49 easy this is. First, all you have to do is set up the app connections in Zapier.
7:52 So, you go through the OAF flow or whatever to get your credentials like
7:56 for Google Drive connected. Then, you go into the MCP server tab, which I'll link
7:59 to in the description to set up your server. You pick your client and then
8:03 you pick your tools. You can search through all of the services that are
8:06 available. Like I can do Google Drive for example. You pick the individual
8:10 tools that you want in the MCP and you configure all of the parameters. Once
8:14 you have that set up, then you go to the connect tab and right here there's a
8:17 button to just add it directly into Cloud Desktop or for other MCP clients
8:22 you can just copy this URL and set it up just like you would a typical streamable
8:27 HTTP MCP server. And so for example in cloud desktop I'll go to the settings in
8:32 the top right and then I will go to the connectors and then I can add a custom
8:35 connector and I just have to paste in the URL that I copied here and I can
8:39 call it Zapier and then boom you are good to go. It'll take you through an
8:42 OOTH flow with Zapier so that it's nice and secure but then you will immediately
8:46 have it available as capabilities to cloud desktop. And so for a very quick
8:51 demonstration I used the Zapier MCP server in cloud desktop to find my
8:54 Google Drive folder and create a file here. So you can see both tool calls,
8:58 everything's working perfectly, and we even have the file at the bottom here so
9:02 I can click right into it in my Google Drive. And it was so easy to set this
9:06 all up. And so I will have a link in the description to the Zapier MCP server.
9:10 26. Introduction of OpenAI GPT-5 (August)
9:10 Definitely check it out. Now, surprisingly, August was a pretty quiet
9:14 month for Generative AI. The biggest thing that we had was the introduction
9:19 27. Anthropic Raises $13B at $183B Valuation (September)
9:19 of GBT5 from Open AI. Getting into September now, we have more funding. The
9:24 funding train never stops. Anthropic raises $13 billion, putting it now at a
9:31 28. Anthropic Settles $1.5B Copyright Infringement Lawsuit
9:31 $183 billion valuation. Now, there have been a lot of lawsuits in the AI space
9:35 this year, but one of the interesting ones that happened more recently is a
9:39 group of authors sued Anthropic for using their books and other materials in
9:43 their training data. And Enthropic ended up paying a whopping 1.5 billion
9:48 settlement. Now, I don't mean to just pick on Anthropic here. There are a lot
9:53 of different lawsuits for all of the LLM providers, but it is crazy. They're
9:56 paying around $3,000 for each of the estimated 500,000 books covered by the
10:01 settlement. Also, in September, Perplexity, the AI search startup,
10:06 raises $200 million at a $20 billion valuation. It's very impressive to me
10:10 that Perplexity has stayed on the cutting edge and remained competitive,
10:13 even though people think time and time again that all of the common agents and
10:18 LLM are going to replace Perplexity with their built-in search. And continuing
10:22 the funding train, Mistl, the French AI startup, raised€1.7 billion, and they
10:27 continue to put out some of my favorite open- source LLMs and a lot of their
10:31 closed source stuff is really powerful as well. And finally, to end things off,
10:36 in September, OpenAI released Sora 2. And this, my friend, took the world by
10:40 storm. The capabilities here for both the video and audio generation, it's
10:45 just insanely cool. And continuing with OpenAI going into October, Sam Alman
10:51 said that in October, ChatGpt hit 800 million weekly active users. I wish my
10:56 applications had that many weekly active users. Also, IBM and Enthropic announced
11:01 a partnership to accelerate the development of enterprise ready AI. I
11:05 was actually there at the IBM Tech Exchange conference when they announced
11:08 this. Very, very cool. Anthropic is definitely continuing to push themselves
11:13 as the enterprise LLM provider especially for coding. Now getting into
11:17 November, Microsoft announced the formation of the MAI super intelligence
11:22 team. What they're working towards here is humanist super intelligence.
11:26 Incredibly advanced AI capabilities that always work for in service of people and
11:31 humanity more generally. Now this is kind of vague to me right now, but I
11:34 think we're going to find out a lot more about this going into 2026. Anthropic
11:39 releases Cloud Code for the web. Certainly a competitor to OpenAI's
11:43 codecs that I talked about earlier. Being able to work on your repositories
11:47 with agents running remotely, so you're not even in your local development
11:50 environment. I definitely think this is the future of Agentic Coding. Something
11:54 I'm even working on myself, building custom systems for remote Aentic coding.
12:00 Then Google takes AI image generation to the next level with Nano Banana,
12:04 especially Nano Banana Pro, which was released at the end of November. It is
12:08 insane the quality of images that we can generate with AI and especially how
12:13 realistic they look now. Microsoft, Nvidia and Anthropic all announce
12:18 partnerships. So Anthropic to scale Claude on Azure, Anthropic to adopt
12:22 Nvidia architecture and Nvidia and Microsoft to invest in Anthropic. So
12:26 definitely a bit of a circular investment here which speaks a little
12:30 bit to the possibility of an AI bubble. But it makes sense for all these big
12:34 companies to be working together. And we right now are only halfway through
12:37 November. There is so much that happened last month. Next, we have the release of
12:42 Gemini 3 from Google, which is often considered to be the most powerful LLM
12:46 from a general standpoint right now. It absolutely crushes it on the benchmarks
12:51 and handles long context tasks very well. Now, at the start of this video, I
12:54 alluded to the fact that there is a single application that is already
12:59 adopted by over 90% of Fortune 500 companies, and that application is
13:04 Microsoft 365 C-Pilot. And I believe they reached that 90% mark in November
13:09 of last year. Just absolutely crazy. The enterprise adoption for a single
13:14 product. Enthropic continues to crush it with Cloud Code. Cloud Code reaches $1
13:19 billion in revenue and Enthropic acquires Bun, the JavaScript runtime to
13:23 continue to make Cloud Code more and more scalable because let me tell you,
13:28 the demand for it is incredible. On top of that, Enthropic releases Claude Opus
13:32 4.5, which most people will tell you that it's not even a competition. It is
13:38 the AI coding king. Sonnet 4.5 was already amazing, but now Opus is just
13:42 crushing it for AI coding. Now, there hasn't been that much news around
13:45 DeepSeek since the start of the year, but now going into December, they
13:49 released Deepseek version 3.2, an open- source large language model that
13:52 competes with all of the big guys, Gemini, Claude, and GPT. Now, this is
13:58 open source, but it's such a large model that most people can't actually run on
14:02 their own hardware, which is a bit of a bummer, but unfortunately, that is what
14:06 it takes right now to compete with all of the top dogs. And continuing with
14:10 open source LLMs, we also have the release of GLM 4.7, an LLM that is
14:15 designed specifically to be very powerful for AI assisted coding. And in
14:19 their benchmarks, it compares extremely well to the DeepS version 3.2 2. We just
14:26 covered CloudSonet 4.5, GPT 5.1. The point of these open- source models is
14:30 that they're almost as powerful as the big guys, but they are a lot faster and
14:34 cheaper. So, Anthropic's model context protocol really took off earlier this
14:37 year, and now in December, they announced that they're donating MCP as a
14:42 part of establishing the Agentic AI Foundation. This is a directed fund
14:47 under the Linux Foundation co-founded by Anthropic Block and OpenAI with support
14:51 from a lot of other absolutely massive companies. And so hopefully what this
14:55 means is that all these companies are starting to work together to build out
14:59 AI for the greater good. Now, as much as MCP is incredible, Enthropic also
15:04 recently released Claude Skills. They did this back in October. It's really
15:08 started to gain traction this month in December though and it's becoming a
15:13 replacement in a lot of ways for MCP cuz it's a lot more context optimized. And
15:17 so it's this idea of progressive disclosure where we only give
15:21 capabilities to our agents when it actually needs them. So we save a lot of
15:26 context up front. This kind of paradigm is going to be very popular going into
15:30 2026 for building any kind of agentic system. Definitely not limited to the
15:34 claude ecosystem either. OpenAI releases GPT 5.2. Definitely a lot of
15:40 improvements over GPT 5.1. Now, Enthropic with Claude seems to be quite
15:44 focused on AI coding as the specialization of the LLMs, but GPT is
15:49 more general purpose like Gemini, so seems to be kind of competing more with
15:54 Gemini 3. OpenAI also gets a $1 billion investment from Disney, especially
15:58 thanks to Sora. definitely showing that there is a lot of use case and demand
16:04 for AI generated videos with audio. So, we haven't talked about data companies
16:07 that much in this video, but they're certainly doing well, too. Datab bricks
16:12 is doing their over $4 billion series L and is valued now at 134 billion. They
16:18 also acquired Neon recently, one of my favorite database platforms. They're
16:22 definitely rocking it. And then at the very end of this year, Nvidia announced
16:26 that they are buying Gro's assets for about $20 billion. Now, the title is
16:31 kind of strange cuz it doesn't say it's an acquisition, but it definitely feels
16:34 like an acquisition when you really dive into it. So, definitely a lot of
16:38 interesting things that will unfold here going into January as well. All right,
16:41 to end things off for us here, I literally have an article dated December
16:46 30th of this year. Meta is going to acquire Manis. And it's cool cuz I
16:49 talked about both of these companies already. Mannis was very revolutionary
16:52 when it was released at the start of this year and Meta hasn't made a splash
16:57 in the generative AI space in a while now and so it definitely seems like
17:00 they're doing some things behind the scenes that we might find out about more
17:05 maybe quarter 1 or quarter two of next year and so they're going to be
17:08 integrating Madness into its products including Meta AI. All right, so that my
17:12 friend is all the big things that happened in the AI space in 2025. So
17:17 definitely let me know in the comments if you think there's anything that I
17:20 missed and also what you're looking forward to going into 2026. And in
17:25 January, I'll definitely be making some content on my predictions and what I'm
17:30 focusing on going into the new year. So if you're looking forward to that or
17:33 just building AI agents and using coding agents in general, I would really
17:37 appreciate a like and a subscribe. And
