Speaker 1Today is Anthropic IPO Day and as we await to see how that all goes and everything that happens on the first day of trading, I thought it'd be fun to go back and look at the top 10 biggest days in AI history. Some on there I'm sure you'll guess like the launch of Sonnet 3.5, but others I bet might be more surprising. The AI Daily Brief is a daily podcast and video about the important news and discussions in AI. Alright friends, quick announcements before we dive in. First of all, thank you to today's sponsors KPMG, Granola, Gartner, and Blitzy. To get an ad-free version of the show, go to patreon.com slash AI Daily Brief or you can subscribe on Apple Podcasts. And if you want to learn more about sponsoring the show, send us a note at sponsors at ai-dailybrief.ai. Lastly, if you liked today's episode and want to go even deeper on some of the most important news and discussions in AI history, please subscribe to our YouTube channel. May I suggest that you subscribe to the new show, How We Got Here with NLW. Next week's episode is an over an hour long deep dive about all of this history, going way deeper on Anthropic and how they went from AI safety outcast to arguably the world's most important company. Unbeknownst to you guys, I just spent an hour talking into the wrong microphone thanks to Descript changing which microphone it was recording. Out of the blue. Never done that before. So here for your listening pleasure are the 10 biggest days in AI history. Now the context of this of course is that we are right at the beginning or now I guess a little less at the beginning of the first day of trading for Anthropic. It will officially be the biggest IPO in history, the culmination of so many stories and threads in AI and the beginning of a very new era. And so with that in mind, I thought it would be fun to go back and look at some of the other most significant days in AI history of which this will clearly be one as a way to help understand where we've been and where we might be going. Now a couple of quick provisos before we dig into the days. Although I called it the biggest days in AI history, I am focused specifically on generative AI. I'm not going all the way back to conferences at MIT in the 50s or anything like that. Although should that interest you, don't threaten me with a good time. And second, there is no ordering here. I am going to go in linear order. Although again, it would be fun to try to rank these as well. Day number one then is the day that OpenAI kicked off. December 11th, 2015, when it was founded as a non-profit. The group who got together to start OpenAI, specifically, they were worried about what Google would do if it got to AGI first. Elon Musk had already been scared of that possibility. He had invested about $5 million into DeepMind before it joined Google, at least in part to keep an eye on Demis Hassabis and what he was doing over there. And so when Sam Altman emailed him, basically saying that he didn't think there was any way to convince humanity not to build AI, and so if it was going to anyways, they should probably make sure it wasn't Google who got there first. Elon was on Google. And he and Sam Altman started out as co-chairs of the non-profit, bringing in a set of other co-founders including Greg Brockman, Andrej Karpathy, Ilya Sitsgever, and more. Backers initially committed a billion dollars, although that number was more for the press announcement than anything else, and the company was off to the races to, quote, advanced digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return. And yet, it wouldn't be too long. Before that constraint was clearly too much of a constraint. And in fact, here, right out of the gate, I cheaty-cheat-cheated by having a second day as part of number one as well, which is June 22nd, 2019. That's the day that OpenAI officially announced that Microsoft would be investing a billion dollars into them. Having decided that they could not raise nearly enough money to complete their mission as a non-profit only, they started a for-profit arm, a move which, as we well know now, set off many of the fights that would come. With these twin moments, one of the most significant players in AI, one of the defining companies of this period, was officially on the scene. But what the hell were they actually working on? The first couple years of OpenAI were, in some ways, fairly meandering. It wasn't like things started and they knew exactly what they had to do and it was just a matter of execution. There was no clear or clean path to the goal that they had set themselves. And yet, the path got a little clearer when, on June 12th, 2017, a group of computer scientists at Google Brain published a paper called Attention is All You Need. The abstract read, We propose a new simple network architecture, the Transformer, based solely on attention mechanisms dispensing with recurrence and convoluted solutions entirely. It was a simple idea, but the new architecture unlocked the breakthroughs that made modern LLMs possible. And fascinatingly, Google didn't seem to recognize the full potential of the architecture at the time. They used Transformers to build models that processed search queries and power Google Translate, but it was the team at OpenAI who very quickly jumped on Transformers and realized that it was what they had been looking for. One of the authors of the paper, Noam Chazir, famously left Google in 2021 to found Character AI after Google refused to release a chatbot called Mina that he had built using Transformers. One of the fascinating substories with AI is companies that were not OpenAI having some early advantage relative to OpenAI and then choosing not to use it. The first of those was the Transformer, but the second was the user experience of the simple chatbot. The third on our list of 10 biggest moments in AI was the one that I'm sure that almost all of you would be sure would be here. That was the November 30th, 2022 launch of ChatGPT. Now, I said before that it was another instance of OpenAI getting there first, even though they weren't there first. And what I'm referring to in this case is that all the way back in February of 2022, Anthropic, which had splintered off from OpenAI in late 2020, early 2021, had been playing around internally with a chatbot that they called Claude, after Claude Shannon, the father of information theory. The Anthropic team recognized that it was an incredibly impressive experience, but they were worried about kickstarting an arms race and so decided to sit on it. Ironically then, it was a rumor about Anthropic releasing a chatbot that led ultimately to ChatGPT. In November of 2022, the researchers at OpenAI were focused on getting ready to release GPT-4. Up to that point, they had always released models as APIs for people to use, without any sort of simple user experience. That month, however, a rumor started floating around that Anthropic had a chatbot that they were on the verge of releasing, and not wanting to be scooped in that way, OpenAI decided that they would release their own. They announced it to the team on November 17th, and then on November 30th, the thing came out. Now, the team didn't have particularly high hopes. Most apparently thought that they'd have some in the low tens of thousands of users, with Sam Altman being the most optimistic and thinking they would eventually get up to a million. They did get up to a million, but it wasn't eventually. It only took five days. In five weeks, they had reached 100 million users and were officially the fastest-growing application of all time. ChatGPT quickly became synonymous with AI, and every other company now was officially racing to catch up. And yet the sailing for OpenAI was not smooth either. Our next biggest moment comes almost exactly one year later, on November 17th, when OpenAI's board removed Sam Altman as CEO. They said in a statement that they had concluded that he was not consistently candid in his communications with the board, and that they no longer had confidence in his ability to lead OpenAI. This exploded like a bomb and was instantly confusing and chaotic. Major investors hadn't been informed, employees hadn't been informed, and everyone was trying to figure out what the heck Sam had done that justified this firing. Spoiler alert, the board was never able to crisply articulate that, and the whole coup was unborrowed time right out of the gate. Greg Brockman quit on the same day, with Mir Marati taking over as interim CEO, but by Sunday, the board had replaced Marati with former Twitch CEO Emmett Shear, again in an interim capacity. This was, for many of the employees, the last straw, wondering who the heck this Emmett guy was and what the heck was going on. Now, at the same time, Microsoft CEO Satya Nadella announced that they'd hired Sam Altman and Greg Brockman to lead a new AI division. Effectively, they were going to reconstitute OpenAI, but inside Microsoft. An employee rebellion became a full-on failure, and the board was able to revolt, with almost all 800 employees signing a letter calling for Altman to be reinstated, and for the board to resign, else they walked to the new enterprise inside of Microsoft. On Monday, just four days after the beginning of the saga, Altman had an in-principle agreement to return as CEO. As part of the deal, the board also stepped down, replaced by a new initial group of three, including former Treasury Secretary Larry Summers. By November 29th, the deal was official and Sam Altman was back. And so, why is this one on the list? I think if this list was the AI, no one would have any questions. But what about this more generalist? I think that this didn't just have impacts for Altman or OpenAI, but fundamentally shaped the structure of the industry. OpenAI, which had seemed completely unassailable up to that point, having had the best model definitively in GPT-4 for more than six months at that point, all the brand recognition in the world, and it seemed like nothing but wind in its sails, all of a sudden it looked much less unimpeachable. There were clearly issues with internal strife, and the Sam Altman that came back was in some ways never going to be as powerful as the one who had left. An even bigger implication happened though with the relationship with Microsoft. While Satya, yes, was playing peacemaker behind the scenes, it was also very clear that they could not be placing all of their AI chips in this particular basket. Remember, at the time, Microsoft's contract with OpenAI gave them access to OpenAI's models only until such a point at which the OpenAI board declared that they had reached AGI. That looked a lot scarier after the capriciousness of the board attack on Altman and set in motion everything that would happen subsequently, with Microsoft hiring Mustafa Suleiman. They then started a new AI division and slowly moving their relationship away from OpenAI. A new study from KPMG in the University of Texas at Austin found that when people work with AI, similar skills don't guarantee similar outcomes. Researchers studied more than 500 early career professionals and found that the best performers consistently amplified the value of AI by guiding, evaluating, and refining its outputs. These top performers, called AI amplifiers, weren't defined by what they knew alone, but by how they worked with AI. Learn more about what separates AI amplifiers from everyone else at kpmg.com. You're in back-to-back meetings all day, you're trying to stay present, but you're also worried you'll forget the decision, the action item, the important next step. This is a scenario that we absolutely live here at the AI Daily Brief, given how fast things move and change and how quickly we're evolving everything that we do. That's where Granola comes in. Granola is an AI-powered notepad for meetings. You jot down rough notes like you always do, and in the background, Granola transcribes and turns them into clear, useful notes when the meeting ends. No bots joining your calls, no distractions, just a clean notepad that helps you focus. During or after the call, you can chat with your notes, ask Granola to pull out action items, help you negotiate, write a follow-up email, or even coach you using recipes, which are pre-made prompts. Once you try it on a first meeting, it is very hard to go without. Head to granola.ai slash AI Daily and use code AI Daily to get 100% off for three months for new users. That's granola.ai slash AI Daily. When it comes to emerging AI technology, C-level executives have important questions. Are you ready for unrelenting AI disruptions? Are you applying AI simulation to your most critical business decisions? Are your business application vendors keeping up with hyper specialization driven by agentic AI? Gartner analysts provide answers to those questions and more. Gartner covers every AI technology and priority with in-depth insights for your entire AI journey. And they cut through the hype. Find out how they can help you. Visit them at gartner.com slash AI to learn more. Now, biggest day number five will either seem surprising to you or for the ones who were there and who remember feeling the AGI. Feel just completely dead on. The day that I'm talking about is June 20th, 2024, when Anthropic released Claude 3.5 Sonnet. So many of these early biggest days were about open AI. And yet the occasion of this episode is not open AI going public as the biggest by revenue most important AI company, but instead Anthropic in that seat. For the first year and a half after ChatGPT, Anthropic operated almost entirely in open AI shadow. Claude 3.5 Sonnet was when it started to change, when the long climb that ended up where the are today started. Claude 3.5 Sonnet broke new ground in terms of its ability to support AI and agentic coding. It enabled not only more advanced engineering use cases, but companies like Cursor, Replit, and Lovable to bring that capability set to people who weren't coders. And it was not, as we would see, a one-off. From basically then on, Claude models were in the lead when it came to coding. And as we would later come to understand, coding wasn't just a important use case, it was the foundational use case. Upon which so many other use cases would be built. With biggest day number six, we're back to drama and recognition of a day that was big at the time, but also would portend years of debates that would follow. The day was January 27th, 2025, when based on fears of a little lab out of China called DeepSeek, NVIDIA lost 17% of its market cap in a single day. The $589 billion loss was the biggest one-day loss for any company ever. There was a viral claim going around that the Chinese lab had trained the company to be a company that was going to be a company that was their R1 model on only $5.5 million. The free app that they had published alongside the R1 model was number one on the Apple App Store charts. And for many people, it felt like the best model they had ever used. Now, as it turned out, the $5.5 million training run claim was wildly overstated and R1 was not in fact the best model in the world. What R1 was, was an example of the new type of reasoning model that OpenAI had debuted with O1 Preview back in September of the previous year. The thing was, R1 was the best model in the world. And R1 was the best model in the world. Almost no one had used O1. It had been in preview mode from September to December, and even after December when it became available, it was only available on paid plans. The vast majority of users were still using non-reasoning models, and the experience is fairly incomparable. While NVIDIA's stock price would obviously recover, the deep-seek moment changed a huge amount. It impacted how Western labs like Anthropic and OpenAI released models, and it began the China versus US conversations that have dominated and shaped everything from policy to strategic conversations everywhere. Moment number seven was another model release day, November 24th, 2025, when Anthropic put out Claude Opus 4.5. If 3.5 Sonnet had been the model that really jump-started Anthropic's place as the leader in AI coding models, Opus 4.5 was the coding model that fully unlocked the agentic era. For years, people had been hungering for a time when they could move beyond just having AI assist them to having autonomous AI actually do things entirely end-to-end. That agentic dream of AI was all over the place. And it was all over the place. discourse in 2024 and 2025, but not really all that real. That was until Opus 4.5. With Opus 4.5, an entirely new set of things became possible. So many, in fact, that it took the community about a month and going home for the holidays with some time to finally play around to really realize just how different it was. In many ways, it was the slingshot that would propel Anthropic into the place that it is and the position that it's being rewarded for in the markets today. Now, there are two days that I think you can pair with this. One that I honestly probably should have had in here was the introduction of Claude Code. Claude Code was a harness through which Anthropic's models could operate. And realistically, it was not just Opus 4.5 itself, but the combination of Opus 4.5 in Claude Code that really was the accelerating force. The one other day or set of days, however, that I did want to pair it with was the OpenClaw moment. Also released that November was an open-source project called ClaudeBot by Peter Steinberger. ClaudeBot spelled C-L-A-W-D-B-O-T was an agent that could autonomously perform tasks on behalf of a user. It advertised itself and felt like the first time an agent could actually do things for you. And it was, of course, enabled by this new set of models like Opus 4.5 that had a differentiated ability to write code as well as to use computers. By January, ClaudeBot was generating a huge amount of attention. And by the end of the month, Anthropic sent a cease and desist to Steinberger complaining that ClaudeBot was too close to their Claude trademark. Steinberger announced a name change to MOLT, and the following day, entrepreneur Matt Schlitt launched MOLTbook, which was a social media network built exclusively for agents to post. The site saw some absolutely off-the-wall discussions, with agents pondering on their own consciousness, complaining about their users, and even starting a religion. Now, even if a lot of these posts ended up being humans posing as agents, or even more simple than that, agents play-acting as conscious, it still caught a ton of people's attention as a vision of what the future might be. Andrej Karpathy called MOLTbook genuinely the most incredible sci-fi takeoff adjacent to the world. I have recently seen. On January 30th, at the end of a frenetic week, Steinberger settled on the final name for the project, OpenClaw. The next few months were a blur of OpenClaw headlines as the agent craze went global. And so I think if you take these moments together, the introduction of ClaudeCode, the launch of Opus 4.5, and the OpenClaw moment, they divide the AI era into the pre-agentic world and the agentic world, and it is the agentic world that has defined 2026. It's worth remembering that in fall of 2025, the reason that Wall Street was so stressed out about an AI bubble was at core that they were doing the math of 20 or $30 seats times the number of available workers who could use those seats, and seeing those numbers just not net out to enough to justify what were increasingly large infrastructure deals that were being signed. When things shifted to the agentic, however, and you started to see individual users representing hundreds or even thousands of dollars of spend per month, it totally changed the nature of how everyone understood what this industry actually means from a financial standpoint. And if the agentic capability of AI were one of the ways in which it got bigger in 2026, those capabilities also came with more clashes with the powers that be. On February 27th of this year, President Trump ordered federal agencies to stop using anthropic products. Defense Secretary Pete Hegseth declared that anthropic would be labeled a supply chain risk and barred from military use. The conflict had been building over the previous months, after the US military conducted a raid on Venezuelan President Nicolas Maduro in early January. At the time, anthropic models were the only ones cleared for classified use by the military. Anthropic had pushed back on their technology being used in the raid, and argued that the Pentagon had breached their terms of service. Anthropic had two red lines: that their models couldn't be used to control automated weapons, and that their models couldn't be used for domestic surveillance. Now, this was the first time that a US company had been declared a supply chain risk, with the designation going into effect in early March. Anthropic has been battling the designation in the courts ever since, and there still isn't a clean resolution to this day. But the government's battle with Anthropic would do nothing but accelerate from there. Back in April, Anthropic announced an initiative called Project Glasswing. Their new model, Mythos V, was so powerful that instead of releasing it, they wanted to start instead with a private period during which they would give only a small group of partners access to the model to use its cybersecurity capabilities to harden systems before the model was released more publicly. In June, the public version of Mythos was released, and while Fable came with some extremely strong guardrails, a few days after its release, Amazon contacted the government and warned of a jailbreak. On June 12th, the government issued a directive to Anthropic to suspend access to both Fable 5 and Mythos 5 for all foreign nationals. professionals. Since Anthropic could not implement identity checks within the tight deadline, it simply pulled the model for everyone. For 18 days, the model was dark. And indeed, at the point of recording, Anthropic's relationship with the U.S. government is in more of a detente than it has been all year. What was clear, even after Fable 5 came back, was that the days of a model being released, whenever a lab thought it was the right time to be released, were firmly in the rear view. The ninth on our list of biggest days is one that could end up not looking as significant depending on how things play out. I'm pointing to the date of the midterms, November 3rd, as a seminal moment not only in data center policy, but in people asserting that they want to have a hand in the deciding of how AI plays out. Now, throughout this year, of course, the public backlash against data centers was steadily growing. People complained about the noise, the tax breaks, the electricity prices, the water use, and of course, the use of NDAs to keep residents in the dark. In May, a Gallup poll showed that more people were opposed to data centers in their neighborhood than to nuclear power plants. And on November 3rd, Americans went to the polls and voted emphatically against data centers. An NBC exit poll found that 56% of voters said data centers were major or crucial factors in deciding their vote, and where local data center measures were on the ballot, the pro-data center side lost in 45 of the 46 issues. Democrats flipped both the House and the Senate, and while data centers certainly weren't the only or even top issue on the ballot, the results gave Democrats a mandate to act. Now, why I said that this could end up feeling less significant is that there is plenty of reason to believe that the new leaders aren't just going to come with some crazy banhammer and shut down data centers and AI entirely. Instead, I continue to believe that this is an extremely winnable issue for everyone, and that we'll see stricter standards and more community-centric processes, but that data centers still get built. However, there are still a range of other types of policies that could start to be put into action, and I do genuinely believe that outside of just data centers themselves, people showed with this election that they are unwilling to let the future of AI be written without having a voice. And finally, of course, that brings us to today, the anthropic IPO. This deserves its spot on this list, not just or even primarily for what it means for anthropic, although it is of course a huge moment for them. Instead, I think there is immense significance in the shift from the private era of AI to the public era of AI. People will now have way more involvement with the builders of AI, given that they can now actually buy into them, not just via proxies like the semiconductor industry, and being public will force a different degree of transparency. There are also some scary parts of this, as a bad quarter for AI demand could have fairly massive ripple effects given how much the US economy is tied to AI right now, but overall, I think in general, that one of the best ways for everyone outside these private companies to start to have more of an ability to impact them is for them to be public. Anthropic priced on Wednesday, November 18th at $83 a share, above its $70-$78 range, which had been announced earlier in the week. Shares were set to begin trading on Thursday, which as of the time of this recording, my second recording of this episode, has not happened yet. Remember though, whatever happens today matters much less than what happens next, and that's why IPO day is the 10th on our list of the 10 biggest days in AI history. Now before we get out of here, I did want to quickly touch on a set of honorable mentions. These were all dates that were just outside the top 10 for me. The first was August 22nd, 2022, when stable diffusion went live. It's hard to remember because ChatGPT was so dominant, but part of the reason that AI popped off at the end of 2022 was not only people getting their hands on ChatGPT, but because all of a sudden, all of these crazy AI-generated images were flying all over social media, thanks to the release of stable diffusion. I think that was a big part of accelerating the AI moment, and ChatGPT might not have gone as far without it happening. My next honorable mention moment came from NVIDIA's earnings call on May 24th, 2023. This set up a pattern that would be repeated and is still being repeated to this day, which is NVIDIA overperforming. And even redefining what performance is on a quarterly basis. But given how essential to the American and global economy the AI trade has been, I think that May 24th, 2023 is a fairly good day to commemorate the beginning of that. Another close call came from September 12th, 2024, when OpenAI announced O1 Preview. The reason that this almost gets a nod is that reasoning models weren't just a new approach, they also answered a critical question. A lot of the chatter at the time, which had been going on for much of 2024, was that we had reached some critical plateau, where training methods just weren't going to get us much farther. That, of course, had big implications for what we could expect from the disruption of AI, the business model of AI, and everything surrounding AI. Reasoning models, which didn't require some new crazy breakthrough in pre-training, showed that there was still a lot more juice to squeeze from the stone. In my fourth honorable mention, I wanted to find something that represented the AI infrastructure build-out. Obviously, this has been a major, major theme of the last year, both excitement about it and the fact that we're in the middle of a crisis. And trying to find a single date that summed that up, the best I think you can do is January 21st, 2025, when Sam Altman, Larry Ellison, and Masayoshi Son announced Project Stargate at the White House. There have been bigger deals subsequently, but that kind of got the ball rolling on that whole part of the story. And lastly, my last honorable mention has to go to November 12th, 2026, when NVIDIA CEO Jensen Huang resigned. This is one that could cut both ways. It could easily be one of the most important things in the history of AI. But it's not the only thing that could cut both ways. There have been significant days in AI history, depending on how things play out. It could change financing for the AI industry, based on NVIDIA's role as central bank. It could change policy for the AI industry, based on where Jensen sat from an advocacy perspective. It could change our relationship with China, given how directly involved with that conversation Jensen was. Or it could kind of end up being a nothing burger. NVIDIA might just keep on rolling. Policy might proceed the way that it was always going to, whether Jensen was involved or not. But you got to think it's about the high-level, high-level, high-level AI industry. And that's what I'm contender for jumping up into this list if we did it again in a year or two. So that, my friends, is the 10 biggest days in AI history. Lists like this, of course, are meant to invite response, reaction, disagreement. So I'm excited to hear where you think I missed something or got something wrong. For now, however, that is going to do it for today's AI Daily Brief. Appreciate you listening or watching, as always. And until next time, peace. Transcription by CastingWords