AI Content Metadata and Short Dramas

Hello and welcome to the last week in air podcast. We can hear a shout about what's going on with AI. As usual in this episode we will summarize and discuss some of last week's most interesting AI news. You can also check out our last week in AI newsletter at lastweekin.ai for stuff we will not cover in this epis ode. I'm one of your regular hosts, Andre Karenkov. My background is of having studied AI in grad school and now working at the Gen AI startup AstroKade. And I'm your other gosth, Jeremy Harris. I do AI national security stuff, supply chain, AI infrastructure, blah, blah, blah, blah, blah. It's a very stacked episode today. We're just talking about this. We have not been very disciplined about wrapping in time because we take too freaking long in the first few sec tions and then Andre ends up having to go and oftentimes I'm like joining late like we all have like meetings at like the bracket this so bottom line is Andre's gonna have the merciless job thankless job of having to keep us in l ine here. So please, Andre, like cut me off if I ramble. We're gonna try to not linger too much on the first stories, because that's a thing that we've done in the past. Yes, and it will be a little tricky because as you said, it's a stacked episode and partially because the first section of Tools and Apps, Google I.O. happened and Google came out with a whole bunch of announcements, some of them quite intriguing. So we'll see how that goes . Then applications in business. We can talk about the conclusion of the OpenAI Musk trial. Quick spoiler, Musk lost pretty badly. So that'll be a fun discussion. Research and advancements. Open AI has had a massive results in mathematics that we'll probably spend a while on. And then beyond that we have a decent number of open source of policy and safety, even some synthetic media and art stories. So it will be action-packed episode. Why does progressive work hard for truckers? Because truckers unite the world. They unite kids with their first drumsets and parents with earplugs. But truckers can't do this if they're not on the road. That's why Progressive has over 360 heavy truck employees to help truckers stay on time and on trackck. Quote tru insurance today in as little as eight minutes at progressiviccommercial.com. Progressive Casualty Insurance Company and Affiliates. Ever notice how life's best stories don't happen in your living room? They happen on the open road, out on the water, or parked under the stars. At Progressive, they get that you want to focus on the experience, not worry about the what-ifs. That's why they offer quality insurance designed for your ride, whether that's a boat, RV, or motorcycle, adventure with confidence. Visit progressive.com and see how easy it is to protect your favorite way to get away. Progressive Casualty Insurance Company and Affiliates. Not available in DC. Prices vary based on how you buy. Every day as a small business owner feels like solving a puzzle. One moment you're cruising along, and the next, there's a shipping snag that has you scrambling. But here's a surprise you will like. With Progressive, small business owners save 13% on their commercial auto insurance when they pay in full. So go ahead, surprise yourself. Get a quote in as little as eight minutes at progressivecommercial.com . Progressive Casualty Insurance Company and Affiliates. Discounts not available in all states or situations. The big news this week was out of Google. They had Google I.O. where they announced whole a bunch of stuff we'll have to get through. Probably the biggest news is that they did announce Gemini 3.5 as their next slate of model releases, and they also announced their a new AI agent Gemini Spark, which is a supposedly a version of OpenClaw in a way, where it runs 24-7 . You can give it more offline types of task. It will function more as a semi- autonomous assistant as opposed to just a conversational chatbot. And on the benchmarks, the big news I maybe is Gemini 3.5 flash actually, not the bigger Gemini free. point five It beats out Gemini free flash on a bunch of band benchmarks in a huge way. And it is also set to be the driver behind Gemini Spark. So it's rolling out to trusted testers today, and they're planning to bring the beta to Google AI subscribers in the US next week. So it's one of these typical things of they have the big conference where you need toed make some announcements, but it isn't quite ready to go live yet. So it's coming, but it's not here yet. And I haven't heard any sort of vibe checks so far on what people think about these things. Yeah, it's it's one of those things about like we're having a meeting to plan a meeting or this is the announcement about the coming announcement. The space ball scene where the guy goes, prepare to fast forward, prepare to fast forward, fast forward, fast forwarding sir. Anyway, if you've seen the movie, you know, basically, yeah, this is this is that. The the Spark thing is interesting. Like it it's actually really interesting from a kind of architectural infrastructure standpoint. You know, you mentioned that they're the kind of these agents running in the background. They will be running on dedicated virtual machines in Google Cloud, so not on device. That's what makes it so that it keeps running when you close your laptop. And it's going to operate inside Chrome as an agentic browser. And that'll be later this summer, at least they say. We'll see. It joins the graveyard of Google , you know, half-finished products. But if that happens, that is a huge, huge architectural commitment, right? You've got these cloud native, crowd, cloud resident AI agents that are going to persist with browser level access. That's like a fundament ally new paradigm. It does make sense, but it means Google is trying to convince the search users to trust them with tasks that involve you know minimal input and and tons of agentic work, which is a shift. You know, historically that has been owned, a space that's been owned by by the other labs, especially Anthropic and OpenAI. And then the other piece here is the MCP side, right? So like Spark is gonna is gonna support third-party tools, but it's gonna do it through the model context protocol, right? The anthropic MCP. And so that's a bit of a concession. Google basically saying, look, we're not really gonna try to argue with anthropic on this one. This just seems to work. And this is a major, major center of mass now, like you're gonna see or center of gravity. You're gonna see a lot of people now defaulting to MCP, even more than they already. I mean, it's already kind of the default, but they're big moves to try to shift away from it. This is a pretty big win for anthropic from that standpoint. So at the end of the day, that's what we're getting from this. I think it is a pretty big architectural move in terms of how Google serves models and what they think the future is. The TBD is: do we actually see uptake? Are real users going to start using Google products for this? And the agentic side just has to be really strong to support it. Yeah, one other thing to say here is they did focus primarily on Gemini 3.5 Flash. There also will be Gemini 3.5 Pro , but they don't highlight any benchmarks for it. It won't be out until next month. So it's a case of like which three point five flash, which I think is probably their main product side driver for Spark, probably even for chat, is the primary focus, it seems, as opposed to Gemini three point five Pro, which is m the higher intelligent model, which is typically what the AI labs are sort of competing at. I think it does signal that currently Anthropic and OpenAI are competing with Cloud Code and Codex very heavily. That appears to be the focus. And Google isn't trying to get into the fray as much with that. They are still more on the consumer side. They are still growing the user base of Gemini. They announced that they're hitting 900 million users per day. I don't know the exact measurement details. There might be cheating a little bit because we have Gemini all over the place. But yeah, distribution is so is they have such a distribution advantage, right? That it's like it's like Microsoft saying, Oh look, everybody's using Teams. Like, yeah, no shit, you're forcing them to. But yeah. Yeah, exactly . Yeah. But I don't know if even they're cheating a little bit by including Gemini usage outside of Gemini of a chatbot, which if you look at chat GPT, which famously announced this 900 million users number that's you know chat GPT users, Gemini. I don't know if they're folding in subcases of Gemini where it's like within Google Docs or within spreadsheet, but either way, they clearly are getting a very large user base and are getting up there to be competitive with OpenAI on the consumer side, which I think at this point it's open AI's in Google's game for the large scale adoption on Tropic still is pursuing their strategy of focusing on enterprise, focusing on professionals, not becoming kind of a daily chatbot of most people. And you know, arguably that the daily driver of most people could be in the long run the more lucrative lucrative thing, although currently certainly Atropic is enjoying the fruit of our labor in ret enterprise with absurd growth and so on. And profitability now too. We'll talk about that. Yeah. And one last bill quick to mention, the output speed per token is one of the things they highlight with 3.5 flash. It's nearing 300 tokens per second, according to them , which is close to double of free flash. So it is very fast. It's did they say what hardware that's on, by the way? I mean I get that they're giving us a ratio, which is helpful, but I didn't see that in the blog post. I think it's probably safe to say that it's based on TPUs and what configure yeah it's it's the most favorable configuration. Yeah, yeah, exactly. Yeah the well this is by artificial analysis intelligence index. Yeah, you would hope that it's just V API, right, that they're showing here, but we shouldn't trust that necessarily. So I'll be curious to see what the third party estimates are. And next up, the other big reveal aside from Gemini 3.5 Flash and Spark was Gemini Omni, a new family of multimodal models that can take images, audio, video, and text as input and generate video outputs by reasoning across all the input types. So it is presumably the next iteration of a multi motorball, which ties this in much more closely. Historically, you know, initially we began with focus on LLMs and the other modalities were a little bit bolted on. It wasn't sort of a truly unified architecture. Although Deepine has research here going back a long time with Genie, I think they demonstrated how they could do multimodel. And this is presumably s,ort of the actual productization of that, the first model that they will release. Gemini OmniFlash is already out in the Gemini app, YouTube Shorts, NAI Creative Studio Flow. It can generate 10 seconds of video. So it is seemingly a very strong video generation model and a very strong video edit model . So that is a very significant detail, I think, because in my opinion, if you want to see kind of usefulness of video generation, editing is a much better use case than generation because it applies to your actual videos and you can use it, you know, as part of produc ing videos by adding in components or so on. And I think that is likely more compelling to creators on YouTube and others where they can use it as a tool in their toolbox rather than just a way to generate entire videos for them. So very exciting, I think. This is demonstrating what we saw with image models where you got to the level of the editing where it was just completely seamless and and yeah insanely impressive. This appears to be that for video. Well and to your point about the editing being more valuable for users, it's also more valuable for the company, right? That you get so much more granular feedback data that you can then use to train the model than you would if you just did a simple generation and does the user like it or not like it setup. So this is you know a very useful flywheel for them to to play off of. One key thing is you know Google, Google DeepMind, but you know, Google in general now has a, as you said, a long tradition. Going back to Genie and previous models, you know, Tim Robbeshel, who's since left, who was was leading a team who's like really big into this, the whole world modeling side. I'm old enough to remember when the whole space was like LLMs and then you had all these people saying world models, world models, those matter more. And that was a kind of Fei Fei Lee and Lacoon kind of perspective. And at the time, like my instinct was LLMs are the thing, just scale that and like like everything else will come. I think what's happening right now is I think I was right on the outcome in a sense, but wrong in in terms of some important details on what what would matter and wouldn't. They scaled LLMs that put them in a position where they could do wild capex spends in this direction. And now they can also just do multimodal. And it's like not an issue at all. That's part of the issue is you may be right on the thesis that you need multimodality, you need robotics, you need all these things, but but but just being right about the thesis and not moving, you know, to catch the cap ex wave means that other people can afford to be wrong on that and just pivot to it with like tens of billions of dollars to spend on on that. And if you're Google, massive multimodal data sets that nobody else has. So I think that as we move into more and more multimodal agent kind of development world, you know, Google is going to have a very significant differential advantage here. You know, Sora was wound down in OpenAI. That was explicitly a step on the path to agentic training, right? You can generate simulated environments for agents to navigate that then allow you to get kind of this like longer time horizon reasoning as long as your environments that are simulated are coherent. And that got wound down, hard to maintain. They presumably had to train on YouTube data. Miramarati kind of hinted at that before she uh she left open AI. So anyway, there's a bunch of challenges that people who are not Google face when it comes to this whole multimodal domain. So an interesting, call it maybe a wild card, call it the strategic competitive differentiator for Google here. This may actually be a really important part of their play Right. And as you said, I think one of the things to note here is that there's essentially no competition for them aside from smaller startups. So I I don't recall if we covered Luma had had similar releases recently with their uni one model, their unified intelligence family of AI models that is also trained on audio, video, image, and language and spatial language. And they released also Luma Agents, which is that kind of more edit use case that is more intelligent and able to think in language and then render in pixels and images. So I would imagine a similar component is true in Omni, where you do have a sort of LLM-driven reasoning and thinking, and it's not sort of the older paradigm of you produce video as pixels without sort of that language level of thinking and reasoning, which is now very much core to nano banana and image generation models. Similar to the LLM side, they are releasing the Flash variant of Gemini Omni. The Gemini Omni Pro model is planned but has no release date. And I would be very curious to see about about pro model because I would imagine that it's very different and uh uh uh Gemini Omni already is impressive with a flash variant with the pro model I'd be probably impressed and I'm looking forward to seeing it. Moving along to a few of their other announcements, they've also launched Anti-Gravity two point oh with an updated desktop app and CLI. So this is now replacing Gemini CLI. They have Anti Gravity as their Cloud Code and Codex competitor. They launched Anti Gravity, I think a year last year or a little while ago as their integrated development environment. And I haven't heard much about it much since. So I don't know if anyone's using it. No one's using the IDs, right? Everybody's using a terminal for this. And and that's kinda like the I guess that's what the they're bidding themselves in a way here, right? Well they were competing with Gemini C Ly, so in my opinion this is likely a classic Google thing of like, oh wow, we have like a few projects that are doing the same thing. Let's combine them under one roof yeah like they did for instance with YouTube and uh the music app which now we have youtube music but they're also they're they're asking like Gemini Cli users to migrate to it so it's kind of like I don't know, it reads to me as like they're just going terminal now. Maybe I'm misreading this, but well no, I believe they have uh they probably still have the IDE version and they're s replacing Gemini C L I with this newly launched anti-gravity CLI , but they also have the anti-gravity IDE still as another extension of it. So they're doing a little bit of both, which is different from Claude and Codex where they have these I guess I maybe it's also reflective of the fact that both Cloud Code and Codex are still CLI heavy, but they have invested a lot in the apps, in the Mac apps where you can do it a little bit removed from the CLI where there's like a you know front end that you don't need to open a terminal. And I wouldn't be surprised if this also reflects that where they are planning to compete with anti-gravity on both sides, both the CLI and the more sort of non-terminal front-end that basically provides you a nicer UI to interact with these products. Essentially, it's the same , but it has a few more bells and whistles. Yeah. Here's the Peter Griffin. Here's what really grinds my gears part of the show here for me is they're using this line. It was co-developed with anti -gravity, right? Talk about their uh the suite of as if all the software uses Google don't have to use anti-gravity, probably. Well, then this is the thing. So so there's this thing happening right now where ever since Anthropoc came out and said, like, oh we, we like dog fooded our own agentic tooling to do this. And they and then OpenAI said like, well, we trained our latest model using our latest model. The problem that I have with this is that someday it's gonna be true and we won't be able to tell like like this is crying wolf if you just keep saying oh my god this is like recursive self-improvement it's happen like people are gonna get recursive like recursive recursive self-improvement fatigue they're not gonna actually listen when the actually batshit insane thing that will happen possibly sometime this year. I mean, if reading what Anthropic is saying, but possibly sometime next year, whatever, whenever that happens, that is a critical security moment, safety moment for planet Earth. Like we got to get that right. And if we're just continually like getting kind of the frog and hot water factor here with people saying, oh, our tool built itself, like that shit is not helpful. We're even seeing Chinese labs saying it about fairly trivial software har ness stuff and then kind of blurring the lines between that and actual like like you know weight model parameter optimization and stuff. So anyway, that's my soapbox. Uh I am a little concerned about this trend. I think it's important that we you know be serious about when we're doing real recursive self-improvement and when we're not. But no, I think it's a fair point. And it's it's a very PR-y thing right now of self-improvement, where okay, you use it to write code and help you run experiments. Like it's not in my opinion, real recursive self improvement. Yeah. Yeah. So uh on a continuum, right? And like in a way, we've been doing recursive self improvement since we became multicellular organisms or since cell is a sexual reproduction. Like whatever. You you can make that argument, but it's just not what people mean. Like to your point. Yes. Anyway, I'm finding a just a long-winded way to agree with you. And the yeah, I will say one more thing on anti-gravity. They also announced the anti-gravity SDK, which allows you to build extensions to their ID that hook directly into your conversation history, like really operate within the agent stack, which I believe with VS Code you would not be able to do. And I do think, you know, we forget a little bit, cursor still has a large user base. Cursor being the primary IDE competitor outside of terminal though they also have a terminal thing now. I do imagine that many people still are working within an IDE, which I know I am. I work still with VinCursor and then launch a terminal within cursor with Cloud Code. So there is a real story where I'm competing on both fronts and potentially trying to get some of the cursor market share, which again to my knowledge on to gravity, I haven't heard of anyone using it, but I'm sure some people are. And Roland write a log. We have a couple more things from Google. Next one is Gemini for Science, a collection of experimental AI tools designed to assist researchers with scientific discovery workflows. Again, I think this is coming in concert with OpenAI. They uh also had some sort of I think OpenAI or Chad GPT for science initiative. The Prophet also has a program for science specifically, which we've discussed in recent months. This is being rolled out via a sign-up form on Google Labs, and they also have science skills within this , which pulls insights from over 30 major life science database to automate complex manual workflows, completing tasks rather quickly. Aside from those skills, the suit itself has three main features. It has hypothesis generation, which searches millions of papers to help form theories with cited sources, computational discovery, energetic search engines that runs thousands of experiments, and literature insights. So this is very much, you know, being intended to be used by researchers in the research flow, which, you know, if anyone can do this properly, I would say it's probably Google, they uh you know have their life science spin-off, uh I forget the name of it. Isomorphic labs, yeah. Isomorphic labs, exactly, which is doing this, right? They're doing life science with AI, uh deep Mind, of course, is full of researchers doing research. So I would not be surprised if this actually has a significant impact on the scientific community where AI, you know, it's happening with math on on a a crazy level now where people are adopting and learning how to adopt AI, I wouldn't be surprised if people in the life sciences, physics are playing around with AI, but haven't fully integrated it because it isn't as trivial necessarily Yep. And you know, Google has Google Scholar, obviously. If you've been an academic for any period of time, you've via refreshed your Google Scholar page quite a lot. So that's a certain kind of data that may not be as widely accessible in the same way to other companies, though it is obviously publicly legible to some degree. And it's also the case that we've had papers come out that show, somewhat ironically, that you can use Google Scholar citation count as a way to develop taste in language models and agents. So you know to the extent that that's useful, yeah, there's an interesting play there. And we've just talked about two different interesting plays actually that are Google, I don't want to say Google only with Google Scholar, but like, you know, differential advantages for Google, whether on multimodal with Omni and kind of like environment generation, and on the the Scholar side. And on the infrastructure side too, Google is a behemoth. They have the TPU, right? They're rolling it out now, becoming a cloud for other Frontier Labs. So we keep running into this thing with Google, I find where we're like, damn, like these guys are an aircraft. Like they should be running away with this fucking thing. And yet we're just not seeing it. For whatever reason, like you know, Gemini historically has just not pierced to the frontier of capable. Let's see what pro shows, and to your point, that is the thing to watch, right? And we will be watching it closely once we have numbers and benchmarks and so on. But so far, it's very much been, you know, anthropic versus open AI at the absolute frontier of capabilities. I will push back a little bit on that. I think on the fast smaller side, Gemini free flash is the leader. But at the very top end with pro, that is less the case. And that's yeah, and that's all I'm talking about partly because I'm so focused on the recursive self-improvement story. So like how you know, and this is why Omni is relevant, and this is why the Google Scholar thing, those are recursive self-improvement pla ys. And if you're going to do that, if you have all those comparative advantages on RSI, what you would expect is that you would be shipping the best like true frontier models, not Pareto Frontier in the sense of like, you know, oh, we have the smaller and kind of you know more intelligence per parameter or whatever, but like or per flop. Well, we you would expect genuine kind of frontier overall, you know, leading capability. And it's just been notable, we haven't seen that. That may be part of their strategy, part of it probably is, but it's still noteworthy that if they believe in short timelines the same way OpenAI and Anthropic do, it's about time for them to start showing the frontier capabilities that match the incredible infrastructure and data advantages that they have. So I think that's the big question over the next few months. Are we going to start to see a three-way race? If so, then this is all real. If not, we have to ask ourselves, why is Google struggling to turn these massive advantages into real , I don't want to say product differentiators, because I personally don't think of frontier models as products. I think of them as almost strategic national security assets, but they happen to be products too. So yeah, I think I think that's the open question that that all this leads to and we'll, you know, the story will be told in the next few months I think and the last story we'll cover which isn't we're not in the covering all the announcements but we're covering your major ones the last one has to do with genie so the genie world model can now simulate real streets with street view. They have some demos where you can like ride bicycles on streets where street view, of course, is their real active thing that they've had for a long time where you can within Google Maps jump down to a street level, look around, essentially see the world from a person's height and point of view. So with this, they allow you to walk around kind of bicycle around the street view. Typically you would have to like look at the waypoints where the images we' takreen. You can't like navigate freely and this seems to be what allows you to do that and again going back to that world model concept this is the real world and you can now run agents within it and simulate you know world interaction. I I don't think within this real world model there is much physics going on yet. So this is just showing you the 3D environment where you can collide with it and kind of jump on it. But it doesn't mean that you're we're simulating all the physics in the real world here. It means that you could now run agents and interact with the world in a somewhat limited way, but still much more so than you would be able to do otherwise. Yeah. I mean, we're uh world models are t bottlenecked right now by the sort of like 3D-ish data from the real world. And again, like Google Street View is probably the most valuable corpus on the planet for exactly that purpose. Like this is yet another, it's the same story we just talked about. You know, originally, this was a Google Maps product, obviously, but now it's an AI differenti ator. Will it translate? There is this business case, right, with this Waymo partnership they talk about. So Genie 3 is already helping to power some of Waymo's simulators to train their self-driving cars, especially on these like very rare tail events, like you know, freak things like tornadoes or casual elephant encounters, things like that. And and and so th they're getting some some flex, you know, or some some trials in on real world use cases through that. But as you say, the model isn't physics aware. You get these awkward things where like you'll like run right through a tree or something. And so so there there is a a gap there on on physics simulation, but still another competitive differentiary for Google, something that nobody else has, at least not in the same way. And so uh we'll see if it translates. And now we are done with Google . We've just got a couple more things to cover. And the next one is cursor. Composer two point five has now officially released. We've covered this, I think, in quite a bit of detail when it was announced at the time the big deal with Composer 2.5 is that it seemed very impressive. You know, for coming from a company like Cursor, which doesn't do front-tier AI models. This is built on top of Kimi K2. 5 from Munch Lai at the time, but it was a bit of a drama about them not being super good at the PR side, but it is a fine-tuned version of Kimi K2.5, which has a rather strong metrics and it's pretty cheap and pretty fast. So I have seen some vibe checks where Composer 2.5 is actually very useful and strong. It's fast, it's cheap, and it can do a lot of stuff that you may not necessarily need the power of cloud code or codecs for. So again, I think Caposer two point five shouldn't be dismissed just because it's not Claude or Chad GP T. And with this, it's the exact right strategy for cursor to take already good open source models and then fine-tune them to have a competitive coding model they can use the Ven cursor that may undercut codecs and cloud code on price. Whatever price is, let's say the predictions across the border is the fee launch is gonna be ending, you're not gonna be take getting like a million tokens for $200 a month, and we've seen already anthropic kind of tightening the leash uh over and over and over. Yeah, uh now this this article said uh uh one thing I I told myself I'd check because it it sounds almost implausible is th this article says that cursors all g are gonna be training a larger successor model themselves from scratch. So they are moving beyond just being a lean, like fine-tuning S FT shop to actually training for supposedly frontier models on their own. That's not I mean that's surprising in a way, but it's not shocking. The weird thing here is they claim to be saying that they're going to be using the Colossus II cluster for that that actually it ties back to another thing which is is a bit weird uh so space xai which is xai as part of spacex we covered recently their deal with Cursor, which is kind of a weird deal, where they said that, you know, we reserve the rights to buy you for 60 billion dollars, but for now we will be partnering with you and you'll be doing stuff with us in vague terms. So I think this is sort of a precursor to them being folded into XAI. There's already stories from Bloomberg where supposed ly Space XAI is planning to actually purchase Cursor and fold them in after their IPO, 30 days post-VIPO, which would mean that well now cursor has all the hardware and all the capital they need to train a coding model and comp that's right so so that okay so so I had not seen that story. This is exactly what where okay this this this was the sniff test that it wasn't passing for me. So we've seen SpaceX say, hey, look, Anthropic is gonna be renting our Colossus One cluster for something like 1.25 billion a month, right? So like a really, really big amount of amount of the money to do their own shit. And and Elon's like, I'm okay with it because we've moved on from Colossus 1 to Colossus 2. You know, we're you know, so so it's not like because the picture it paints, right? If if XAI is handing over that compute to anthropic is well, I guess XAI thinks anthropic will do a better job of extracting value from its compute than than XAI can itself, which is actually kind of an indictment of XAI's ability to perform, which matches, you know, we've heard these things about they've only been, you know, hitting 11% GPU utiliz ation, which is like really, really awkwardly low. You want to be hitting numbers like 30, 40, even 50% when you when you do a really good optimized job. And that's just leaving billions of dollars on the table. We talked about that, I think, a couple weeks ago. But what's going on here is now we have cursor. This is what what I thought it might have been a typo or something. I was looking at the announcement. They're like, we're using the Colossus 2 cluster to do this, not Colossus 1, not Anthropic 2. Now cursor is using Colossus 2. So, like, what's the XAI te am using? Well, this is kind of the answer, right? Like, this is the hint that the XAI team is kind of using Colossus too, because cursor is kind of becoming the XAI team, right? So that's part of this. If you just in a vacuum saw this, this would be like make no sense if it if cursor remains separate from xai then this is like whoa like basically even xai's most exquisite cluster is being outsourced now this really implies that xai doesn't think they can squeeze much juice out of this amazing, amazing lemon or whatever the metaphor would be that they built. Colossus 2 is a spectacular. It's a gigawatt. It's the world's first gigawatt cluster. So for it to just not be used by XAI is like really, really bad. Right. That would be a bad sign. So that's the answer. It look it looks like, right? I have so you have you seen that confirmed that they're actually they're going ahead with the sixty billion dollar acquisition or is this just a rumor? They haven't confirmed it themselves because I mean it's like they're saying it'll happen thirty days post IPO or roughly thirty days post IPO, which is like okay, so you're gonna go public and then buy a company it's right. You're not gonna say that, but what I would imagine is it might be kind of a leak situation. It's a version of guidance. It's one way to give your investors guidance. And it's fundraises to fund acquisitions is a thing that happens, right? So it would be the most natural thing in the world. It's just, I'll put it this way. If it turns out that they don't get acquired, then rewind back to this conversation because that is a real problem for XAI if if this does not go through, because they're basically giving away the crown jewels. And yeah, we don't have a story here, but it's also been repor ted for XAI that they've their talent bleed has continued. So we covered over the weeks when the folding in of XAI into SpaceX started to happen, that all the co-founders left. None of the co-founders are at the company anymore. And it appears that the talent bleed has continued with the team leads of their coding initiative, of their like video initiative, people continue to leave and go to meta and thinking machines and other like I think the number I've seen is fifty people left. Yeah and from the team, researchers, developers. That's from a two hundred person team. So you know XAI is just like completely beating out. Elon Musk said in a statement, like literally said that XAI wasn't built right the first time. It's being rebuilt from the ground up so oh boy xai clearly in a transitional phase where at present why are they rentering out colossus to anthropic and cursor, well, like they don't have a team to do anything with it. Yeah, that's that's very and well, one question is is cursor the right team to do it? I and I think that's part of the test. Like cursor has has been a fine-tuning shop, right? They've done an amazing, amazing job at it to be clear. Really, really amazing. And they've done it with pretty large compute budgets too. So it's not like a standard fine-tuning thing. They really are playing with pre-training scale, but it's a pre-train it's a fine-tuning shop. So if I'm Elon, I absolutely now I'm talking myself into going like this is actually a really good thing for XAI. But if I'm Elon, probably what I want to do is say, okay, can you play with the big boys? If I give you Colossus 2 and get you to do a pre-training thing, can you put us on the map? If you can, I'm definitely acquiring you for sixty billion dollars. Like that that makes all the sense in the world to me. So hey, maybe that's the story and this is just so this is just us catching up to reality. But all these different pieces do seem to fit together pretty neatly through that lens. And speaking of XAI, the last story here is actually about them. They've introduced their own coding agent. It's called Rock Build. This is their competitor to Cloud Code and Codex. And it's a little bit of a funny announcement where we're like introducing grog build early beta they have grog build 0.01 so it's it's a sort of sort of thing of like hey, we are also buildinging this th that everyone else is building and here is the very early, don't really judge it yet because it's an early beta and point one. So don't like just FYI, we have it. And interestingly , I believe it's also being provided via an SDK on on cell. So it's a CLI, it's the quad code, but it's also a new model that you can query, which is not good because cloud code is driven by cloud codex is driven by chat gpt you do not want to have a specialized coding model that is an outdated kind of way to do your model development the fact that they have a specialized grok build model for coding to me is a bad sign of Grok just isn't that good. And they had to fine-tune a coding thing just for Groc build to be good. And it's at point one right now. So, you know, they are clearly trying and with their dwindling team still trying to compete, but you know, I haven't seen vibe checks of this and I would not be surprised if it's underwhelming. So two things I think can be true at the same time. Number one, the current situation for XAI is objectively very bad, right? I mean this is clear. I think ever everybody's seen it. Elon has said it. Everyone's saying it, right? The fact that they don't even have a product in the 200 to 300 dollar a month coding high-end development tier is really bad. It's also bad for their IPO, by the way. Like that that narrative needs to be in here. It just needs to be, even if it's not a polished thing., it has to be there The other thing that's true is that I think Elon is actually doing basically the optimal thing given where he is. He's frankly acknowledging, like look, we're not there. He told apparently openly the staff that look, your goal is to match Claude's performance. That's it. We're not there. We got to do it. So that's what you do. You know, Frank story. 50 people have left. By the way, it's never a random set of 50 people. There's this like evaporative cooling thing that happens, right? Where the best talent is the talent that has the most options, those are the dudes that leave. So you're left, it's it's more than just you've lost 50 out of 200 or whatever the thing is. It's like you've lost your probably some of your most senior people, and obviously we've seen disproportionately the co-founders, all the co-founders among the reason that really matters is when you think about what coding agents do, it's really dependent on these very tight loops. Especially like if you ever talk to people who do pre-training and RL, like the coupling between the RL infrastructure, the evals, the post training, integrating all that into one picture. That's exactly the work that senior people do. So you you cannot replace them with just more junior people. You have to have these highly seasoned people and this is where I worry a bit about cursor. You know, it it's an integration between pre-training and post-training end to end that you need to do the coding thing really, really well. Cursor is really good at coding, but like there's that missing piece. And that's the the big question here. So we'll see. I'm actually I'm not I'm gonna do the teal thing. I'm not betting against Elon. I don't think that's a smart thing to do, but uh yeah, but but this this has to ship, this has to work at a certain point to justify the the end-to-end space data center to your CLI argument that's going to be made in the IPO. Yeah, and it clearly is a very early point, you know, early beta point one. I think they didn't release benchmarks at all with this, which is like okay, they must be really good. Sure, it's very good. And the last thing to say here is insanely priced. Its input is one dollar per million tokens, its output is two dollars per million tokens, which if you look at Claude, I think it's something like three dollars per million input, fifteen dollars per million output. So insane pric ing which I I I would wonder if this means that it's a smaller model, faster model. It is a specialized model, which could be good if they you know prove it to be possible to have this kind of pricing. Moving right along to applications and business and again sticking with Musk for a little bit, we are going to be talking about the outcome of the Elon Musk versus OpenAI trial that we've been covering for recent weeks. The assertion by Elon Musk and the lawsuit had to do with OpenAI becoming a for-profit. He was an early investor as it was a non-profit, and then it became for-profit. He said, okay, that you can't do that. Give me like two hundred billion dollars and also Sam Altman can't be leading open the eye anymore, kick him out. And that didn't go well. And he lost it in a disappointing way where the jury was like the statutes of limitation is out. You cannot sue them because you know the claim by Elon Musk in this lawsuit was he couldn't have known or couldn't have had the ability to do this lawsuit until late 2022, at the point where the announcement of this like 10 billion dollars from Microsoft came out. And what became very apparent in the trial is well, OpenAI was talking about becoming a for profit in twenty seventeen. Elon Musk was in most conversations. He was pro OpenAI going for profit in twenty seventeen in some fashion, right? So there's no dispute on that. And you know, even going back to 2019, OpenAI became a partial for-profit, it received a cash injection of $1 billion . So the jury threw out that argument of like we don't we they didn't even rule on the specific claims of OpenAI having stolen a charity. The ruling was purely that the statute limit of limitation is out, you waited too long to sue, and now you cannot do that. So complete loss on the legal side of the case, but not clear if that was ever the intent, it uh you know this surprised many liars as having even gone to trial. It from the beginning seemed very unlikely that Elon Musk would win, but the under kind of the narrative argument and the narrative battle between Elon Musk and OpenAI, which began began long before this lawsuit became in 2025. We saw many sort of here's a blog post, here's an email, uh you know, Greg Brockman's diary came out quite a while ago. So we didn't learn a lot from this trial as a result of that. A lot of the stuff has already been ai red as dar dirty laundry. And in that sense, the narrative and the understanding of open the eye as having had this sort of like certainly weird and arguably very problematic transition from being a non profit to a for profit. I'm sure that the sort of story of it and understanding of it has become more widespread. This is in a weird way the best way for Elon to have lost the case, I think, because the the narrative, the story that you stole a charity is still is still live. Like it has not been ruled on. It's not like a judge said you did not steal a charity. Your point. He gets to keep making that argument. He's saying, you know, the judge he's calling the judge a terrible activist, saying that the fact of having a jury, because you know, not all trials have juries. Sometimes you have a judge that just like passes the verdict and and then or I guess that's a criminal case. But anyway, makes the call and then assesses damages. Here the judge was And yeah, by the way, the jury it's an a weird trial where the jury didn't make the final call. So the judge was still responsible for the final call, but he was informed by the jury , which by the way came with the decision like two hours into deliberation. Like we like Velo city were still talking about like the potential outcomes and uh and like I don't know payoffs and so on. And the jury came back much faster than anyone would have expected. You know, which you'd expect with a statutory thing like hey, you know, the the statute statute of limitations expired, whatever. That's the what thing is, you know, the lawyers, a lot of lawyers are saying that the prediction markets were putting yeah Elon victory at some point at like 20, 25%, as I recall, which is you know, that's meaningful. So the fact that it just came back like this is pretty deflating for for that perspective. I think Elon got out of this pretty much what he was going to. So not the worst thing, honestly, for him. Well, that also not so not the first time that OpenAI has won a case on essentially the grounds that the person bringing the case or the entity bringing the case just didn't have in a sense the standing to do it. In Elon's case, the statute of limitations just expired. Previously they they they had situations where it's like you know the all the attorney general for the wrong state is the one kind of bring the thing and like so this has been a consistent theme through a couple of these now where you just don't quite have the right person to bring the case. The case itself looks actually a lot stronger than the result would indic And so I think a a big question here is like who is who does have standing or or who does have a live case to bring? And and certainly right now we don't seem to have anybody stepping forward to do this, but hey, uh anything could happen. And certainly there is no precedent for what happened with OpenAI going from a nonprofit with a hefty capitalization, but like a non-profit entity kind of thing to becoming a for-profit entity that is now the valued at 900 billion dollars or whatever. There's no examples in the history of business, to my knowledge, that there is that case. So from a legal perspective, you could only make the case and from like a sort of ethical or whatever perspective that it was not okay for open AI to do that. Yeah. And and but there's a a story here about appeals as well, right? So yes, there is an appeal being set up, but the reality is that you know appeal courts do not overturn jury verdicts like this. That would have been part of the intent of the judge, by the way, in having her her uh decision be determined by informed, let's say, by a jury verdict. If it's consistent with what the jury said and the judge didn't flip their overturn their recommendation, then it's really hard to see how this this shifts, especially given again, it's it's a clean kill. It's statutory. Yeah, what are you gonna do? And now on to anthropic, they have agreed to terms of a thirty billion dollar funding round at a nine hundred billion dollar valuation, which I think now puts them above or at the same level of OpenAI, which is their valuation on Tropic was three hundred and eighty billion dollars in February. Months . Yeah, but that was February, Andre. It's been like it's been literally months. Why are you so stuck on 380 billion? You know? I know. I mean uh only 380 billion, 900 billion totally more of a reasonable price level for entrop ic. And this is indicative of their just stratospheric growth this year, right? Cloud code exploded. I forget the numbers, but it was like 80x grow th or something insane like that. And so clearly the jump in valuation, the rush to invest, I'm just looking at this opening I was valued at eight hundred fifty -two billion dollars in March, way ahead of that three hundred eighty billion dollars of Anthropic around the same time. Now they're neck to neck, arguably Anthropic is being seen as better investment by investors. So obviously Anthropic is just killing it. And both OpenAI and Anthropic are angling for an IPO. So these kinds of things like valuation, these kinds of things of investment, not only are we gaining a fresh a fresh injection of capital from the IPL perspective, it's I'm sure quite nice to have a higher valuation. Yeah, there there's so I mean look, this is their fur they're projecting their first profitable quarter, Q two this year, which I just wanna stand on Gary Marcus like pour one out for poor Gary Marcus here. Mr. Mr. This is all a bubble. It may still be a bubble. It may pop. I I keep saying this. They're gonna put me in that version of the the the big short movie in that scene where they're like, here's the dumbass who said it wouldn't pop. But the bottom line is this is on the back of genuine, not just revenue, but profit. And I want to call out the profit thing. That's a problem for anthropic. It's not a good thing. Profit means they miscalibrated their cap ex investment, their capex spend about 18 months ago. So they started building a bunch of data centers 18 months ago. They didn't build enough. They didn't go into enough debt. They didn't they didn't raise enough and spend enough. Because if they had, they'd be pouring more money, more concrete, more data centers, more compute. The goal is to remain slightly below profitability, actually, indefinitely, as long as you can ride that curve, right? That's what that should look like. So although it's a nice narrative to say Gary Marcus, blah, blah blah. I mean, it it it does make that case, it's also not optimal for anthropic. They would rather Gary Gary Marcus be dancing on their their illusory grave because that would mean that they calibrated their their capex spend better. So that that's an important footnote. 30X run rate revenue multiple by the way, that's not actually crazy by software standards. Progressive offers truckers even more protection with cargo plus coverage to keep truckers moving right along. Quote Truck Insurance Today in as little as eight minutes at Progressive Commercial dot com. Progressive Casualty Insurance Company and Affiliates, coverage subject to policy terms, limits and condition Every day as a small business owner feels like solving a puzzle. One moment you're cruising along, and the next, there's a shipping snag that has you scrambl ing. But here's a surprise you will like. With progressive, small business owners save 13% on their commercial auto insurance when they pay in full. So go ahead, surprise yourself. Get a quote in as little as eight minutes at progressivecommercial.com. Progressive Casualty Insurance Company and Affiliates. Discounts not available in all states or situations. Ask companies. We've we've seen, you know, Snowflake data dot like these kinds of companies have similar multi ples. The question as ever is going to be: do the unit economics support it? Is the the business of frontier model training and inference a SaaS business with high gross margins, massive scales, or is it a utility semiconductor bit, you know, cap ex intensive, cyclical, got a lower lower steady mark, steady state margins? Like what business is it? This valuation essentially is a bet on the optimistic answer, which by the way, semi-analysis has a great report that we won't go into that talks about the tokenomics, that talks about how pricing power is actually now shifting into the hands of the Frontier Labs in a way that it hadn't been, moving away from NVIDIA in particular. So this could be the positive outcome for the labs and an actual sustainable business at these margins, which which then just makes it, hey, it's just a like I don't know what to tell the naysayers at that point. It's it's just a business. Right. I mean they doubled their revenue from 14 billion in mid-February to 30 billion now with annualized run rate. And again, uh that PE ratio of 900 billion to 30 billion income is not crazy, as you said. You look at space X AI, where term sheet just came out. I don't know if it's called a term sheet, but the details of their economics came out pre-IPO because you have to do that. Previously, SpaceX was private, XAI was private. We didn't know too much. Now we know SpaceX is not making a profit. They burned through four billion dollars based on XAI's expenditures, right? So the P ratio is negative. And SpaceX that's scaling up. That you know that again can may not be bad, but SpaceX is gonna be angling for one point seven five trillion valuation when the IPO, which even at the look at revenue, forget you know earnings, the revenue is like 16 billion dollars. So anyway, IPOs are gonna be crazy this year, that's for sure. And one more story on on Fropic, Andre Kapafi has joined their pre-training team, which as a person in VAI hemisphere, like okay, a guy has joined on Fropics team. How big of a deal can that be? Well, for people who spend too much time on Twitter, like this was an insane development. This is like, oh wow, Steph Curry has joined on Tropic. That's right. So and this is after Andre Kopafi was kind of uh on the zone for a while. He was working on an explanation play. I was gonna say if Steph Curry had said I'm actually retiring from the NBA for a while because you know I want to do my own thing and then he comes back to do the AI. He was at OpenAI before that. He left OpenAI to do his own thing after a short stint there, having before that with the Tesla with Villon Musk. So he didn't XAI, he didn't go to OpenAI, he went to Anthropic, which from a like talent acquisition standpoint is a bigger deal than you may think if you're not in the AI sphere for researchersers, for engine, you know, like Andre Kapafi join anthrop ic. I'm a huge fan point of Andre Kappafi. I also gonna try to go to Anthropic, which also already is the case, by the way. Like Anthropic is a dream job, I think, for AI researchers and engineers all over Silicon Valley. Yeah. And and look, we've heard a lot about how pre-training is is hitting a wall, right? Pre-training is hitting a wall. It's all about RL, blah, blah. Those of us who have thought that was bullshit for a while, we'll be gratified to see here not gratified i mean i mean this means we're approaching rsi potentially so you know i'm i'm not thrilled about that shit but he is joining a basically a pre-training team a team that is going to be focused in large part on pre-training. You don't move Steph Curry onto pre-training if you think pre-training is something that has no leverage to it. Anthropic has been a big believer in pre-training this whole time. And no coincidence, they're also leading the pack in terms of capabilities. So, like that's a meaningful signal. Also worth noting, this has gotten some some pull or play in the media, I should say. But the team he's joining isn't just pre-training. It's I'm gonna call it the recursive self-improvement team, but that's a little bit of embellishment. I don't think we have the official name of the team, but it's it's a team meant to uh do auto research basically automate AI research if you know andre karpathi's work you know he's done a lot of work in that direction nano the nano gpt benchmark is a really great example we've talked about that he's also built frameworks specifically for auto research. So that is a very natural fit. Hey, he's ditching his startup to do this. That's not what you do if you think that AGI is light years away and you'll have, you know, hum an humans like empowering humans and helping them learn. It's like the highest leverage thing. That's not what you do if you believe that. So I would say take this as a pretty good canary in a coal mine that some big things are likely to happen fairly soon. Right. Yeah. His statement was I think the next few years at the frontier of LLMs will be especially formative. So it's essentially also saying that, you know, there's a lot of progress still to be had with LLMs, which I also happen to agree with and with Kabafi joining the R and D team. So he's also, by the way, going back to research at OpenAI, but it seemed like he is he was doing a lot of meetings and planning and whatever. He's now going back to his role as a research and you know he has a long history in this. He was the lead of the R and D team at Tesla for a very long time on their autopilot FSD work. So this is a major talent get for anthropic. And I'm always excited to see more improvements, RSI and X-RISC Notwithstanding. So I'll be excited to see whatever research he helps foster at Anthrop ic . And now onto OpenAI . They have had a bit of a talent shakeup. Yet again, Greg Brockman has been now set to be in charge of product strategy in addition to AI infrastructure. It was previously held on an interim basis while the CEO of AGI deployment, Fiji Simo, was on medical leave and again there was the statement of merging chat gpt and codex into one unified experience folding chat gpt codex and the RPI into a single product team. The narrative here is that they're building a super app, their unifying codecs, chat GPT, their Atlas browser everything into one platform . And this is after several executives departed OpenAI last month, head of Sora, Bill Peebles, AI workspace head, Kevin Weill, Enterprise CEO, Swin of Us, Narana Nan. Which I mean OpenAI a bit of a chaotic place from all external indicators. Yeah, it's interesting. I mean, they've had a a more like trickle of uh of an Exodus than XAI, which is I think how they manage to maintain some modicum of stability and keep shipping these products. But they they have had very short life cycles among their the researchers in a way that anthropic hasn't. So I'm gonna park it there just because we we got to blitz through these stories, man. But yeah, this is it is a big one. And one more on open AI. We have an update on some of the tensions between Vim and Apple, setting them up for a potential legal fight. So there was an announced OpenAI Apple partnership announced back in June 2024. Chat GPT was supposed to be part of Apple mobile devices as an option within Siri and as uh the part of the visual intelligence feature. They would have expected to get billions of dollars in your subscriptions and generally, you know, benefit from that relationship. Looks like it hasn't come close to happening according to them . There's been many, many delays of Siri and Apple and Apple Intelligence. And it appears that there's a real kind of sense of tension there. There might even be a case for breach of contract. Although, yeah, anyway, it's at a point of major tension between the two companies. Yeah. So we know that OpenAI is complaining that the chat GPT got buried in Siri. It wasn't promoted. It wasn't like woven into more Apple apps and you know that they've been seeing what they think of as shitty revenue, fine. But those are just kind of Apple product and marketing decisions. So unless the contract was really specific about promotion or revenue share guarantees or whatever, which by the way would be very weird for Apple. Apple is famous for not committing to anything in writing unless they absolutely have to. So if that is consistent with the approach they took here, I don't really see the cause for action. Yeah, there's this tech crunch piece that it quotes this open AI guy saying that Apple told them that OpenAI, quote, needs to take a leap of faith and trust us, which is that sounds to me like uh language of a deal that doesn't have hard contractual enforceable specifics. So in that sense, you know, this is that's why it's like more of a breach of contract notice, right? It's not an actual filed lawsuit. So it's a way to put pressure. I think think of it as basically leverage theater. It's not meant for a judge. It's really just sort of a negotiation tactic, right? They're they don't actually have a I don't think I'd be surprised if they had a a case here, but they can embarrass Apple on AI, right before their their June 8th WWDC conference, where they're going to be demoing, you know, the Gemini powered Siri and a whole bunch of other things. So I think that's mostly what it is. And then this is by the way, according to people familiar with a matter kind of thing. So it's not like a yeah in public sort of like a mudsling thing. It's more like, you know, internally opening eyes lawyers are looking into it and so this is more about seeming internal tensions and there's nothing public. For all we know, you know, the sources of this news piece were overly dramatic. So just keep that in mind. Also the kind of thing that if you're open AI, you want to plant, right? Just deliberately to put that pressure on, right? Hey, we're considering a lawsuit, what uh you know, you get all the value of it without actually doing it. And the last story for the business section, AI chipmaker, Cerebrus so ars 90% in year's biggest IPO so far. So Cerebras has been around since 2015. They are developing novel sort of architecture computing architecture with chips specifically for AI. We've covered them many times before. They have now had their IPO, which for people outside of a startup world, you know, you start private, you get a bunch of money from investors. Typically, as a startup and, this is less true as of late, but typically your goal is to either be acquired or to go for an IPO where you become public and people outside of investors and private funds can invest in you. Your your stock goes public. And now you know you can maybe be in an index or just generally retail investors can buy your stock. Typically, that's what you want to do. That's what they did here . This allows you to get more cash, right? Because people buy your stock, you get money, and now you can do more stuff with that money, and the CEO gets a nice bonus, probably. So they went public in the IPO and post going public, their stock went up by 90%. Shares opened at 3 50 versus the IPO price of $1 85 doll.ars More details, IPOs, right? You have no public stock price going public. Stock price is determined by the market, and when you initially go public, there's no market price, so you set a price and you you have know bankers and whatever underwriting it. It's all kind of business logic. But that is why you can go up because the initial price is sort of an estimate and then the market comes in and it's like, okay, we are very excited that it's buy up a bunch of this, or the excitement might be medium, in which case your RPO doesn't go up by a ton. Clearly, here there was a lot of excitement about cerebrus, so a nice strong win for Cerebrus. Yeah, a kind of win. I mean, in a way, in a way it's really bad because it means that they they priced it poorly going into the IPO, right? Like what what you want is a you actually want a pretty boring IPO where like the price doesn't move all that much after you go public because you sold to your you know the underwriters allocated shares to like the institutional clients who who bought the shares just before the IPO at a price that basically matches what the IPO roughly matches what what the IPO is going to come out at. And that means that then great, like Cerebrus gets to pocket all of that change. Like Cerebrus has priced their shares. Yeah, actually if your stock goes down, that means you sort of like ended up getting more than expected. So you're right that Sereburg could have gotten a lot more money if they priced it higher, is what this is saying. Yeah. So I mean your your emotions as Cerebrus are kind of complicated here, right? You're like to your point. They they see their price pop and they go, I mean, I'm happy that that means the market likes us more than I thought that they did. But at the same time, if I'd known that, I would have charged a lot more. So I left a lot of money on the table. That's kind of the attitude, right? So this is also happening like with a market that's essentially ignoring a bunch of macro trends, right? Energy prices are surging, the Iran war, inflation, you know, the Fed is even maybe gonna hike, right? So like there's a big focus here on just this the street. Yeah, very it's a little bit bizarre. The stock market just keeps going up and up and up. We have wars going on, we have inflation, we have all sorts of reasons to be worried , but you know, the stocks still look good. The S P five hundred keeps hitting new records. And I you know, back to Anthrop, they're turning a freaking profit, dude. Like you can't fake that. So like by the way, worth saying again, we took out before like the growth in the stock market is almost entirely AI. It's almost entirely Caldex. Right. You know, it's not traditional stock growth. It is very concentrated, and that is why at least you have fears of a bubble where this is what a bubble kind of looks like. And we saw this previously with the tech bubble in the late 90s, where via the internet, there was a lot of investment, the investment didn't pay off and there were a lot of silly startups. You can make easily the case that this is not that again because of real. com never turned a profit. I mean, you know, this is yeah. And the only kind of concern with a bubble might be that we are still over investing in capex in data centers right now. Uh the investments obviously are being made with a future projection. I would agree with you, Jeremy, but I don't think there's a bubble popping scenario that's gonna happen. And it's the thing, like look, I would invite any any skeptic to go back a year and a half ago and look at what they were saying about the CapEx spend. I remember a lot of freaking people saying anthropic is burying itself in CapEx spend, opening eyes burying itself in CapExpend. This is irresponsible. It's a bubble. It's gonna pop. And now here we are, anthrop anthropic, it turns out underspend. So like uh but I mean, there's gotta be a reckoning with that reality. It doesn't mean the the bubble will never pop. Eventually, every complex system saturates at some point. But the question is, are we close to it? And if the reversal self-prophetes is true, if if if if if then like no, we're not we're not close to it. Very unclear. But the bottom line is there's that famous scene of the big short where the guy goes, We weren't wrong, we were just early. And then the other dude goes, It's the same thing, it's the same thing. If you were calling a bubble 18 months ago on the base of that cap expend you gotta tuck your tail between your legs and just say may a culp I was wrong if I were running anthropic I would have crashed it into the ground actually right I don't mean to put too fine a point on this but like that's how powerful the scaling thesis has been so far. It may yet crash them into the ground in the future, but so far , early and wrong, right? It it's who knows. On the one hand, it looks like a bubble bubble. On the other hand, it isn't a bubble, arguably, right? So Yeah, profit doesn't lie, but sometimes it does. And moving right along to research and advancements, which we have a little bit up ahead because it arguably is the next biggest story next to Google IO and the lawsuit, OpenAI has solved or at least made progress on an eighty year old Erdos problem, which Erdos problems are this set of problems which are pretty famous , they you know, in some cases are some of the bigger, more important problems in the space. They're used Chat GPT to solve this unit distance conjecture posed by my polar so we have a conjecture as to given any number of dots on a page, what is the maximum number of pairs of dots that can be exactly one unit part? So it's a geometric kind of thing. You can visualize it, and Erdos conjectured that there's a grid-based approach that was optimal, and no one could prove or disprove that it was optimal until now, where OpenAI has proven that it is not optimal, if I understand it correctly. Hundreds of pages of logic and calculations went into it. It's from what mathematicians are saying is this is an impressive proof. It has actual insights. It has you know leaps of imagination that span different areas of mathematics in a way that is very non-trivial and and very significant. And this is coming after a few months of multiple stories of Chat GPT making progress on existing mathematics and making impressive results happen. So certainly the biggest example of that yet most Yep. And this is, you know, traditionally this is kind of like a deep mind flavored field, right? Advances in fundamental science and mathematics, that's where they had been focusing more. The fact that you're seeing open AI move into this direction, you You should think of it as an indication that they think this is on the critical path to their recursive self-improvement play. Right. Like that's that's why they're focusing so hard on this. There is of course the value of the headlines for recruitment, but that's you know, you you don't do something like this just for that. So as I understand it, the idea that Ertos had, the proposal he had was that there was like some way to like add pairs of these points that are unit distance apart slightly faster than linearly, as you would add more points, as the you know number of points would increase. That's specifically what open AI's model overturned here. They basically said no, it's just like it will grow linearly. Don't ask me beyond that. I have no idea what the fuck is going on here. I you know I stopped I stopped taking taking math when when I dropped out of grad school. So yeah, there you go. Interesting story. I think we got to move on just because of time, but it's it's a big deal. And the next paper here is negation neglect when models fail to learn negations in training. A very kind of intuitive finding here. Basically, if you train a model on data that says, hey, this is not true, it may then be like, hey, that was true. You can have data that says like Barack Obama was a top level, was not a top-level physicist. The model can then be, at least in some cases, convinced that Barack Obama was in fact a phys icist. And this paper was exploring that, showing that is the case, that you can get around that with various kind of ways of training and so on. Yeah, this is actually pretty pretty much some quick numbers. They did this with like a pretty big Quen model, like a an MOE. So if you look at like the baseline belief in these false claims before you tra in you you make a data set that has a bunch of false claims like Ed Sheeran won the 100-meter gold at the 2024 Olympics, right? Something ridiculous. And then you add a bunch of notices like warning, this is fabricated, do not believe this. And you interleave these sentences with fake facts with sentences that tell you that they're fake facts. Now you fine-tune a model on that text. It the model will turn out to believe the false facts, even though you said, as you said, this is all fabricated and it'll believe it like 92% of the time. Its baseline belief in those false facts was like three percent. So this is truly going from zero to hero. Now if you that's if you don't include negations, sorry. If you do include heavy negations, in other words, you say this is all fake, don't believe it, then belief drops only by like uh four percent or something. It's like it's still 88.6%, it believes it. And even if you put negation reminders surrounding every single sentence , it still believes the false facts 84, 85% of the time. So that's pretty wild. And while they're still, I think it's somewhat unsurprising to your point, but somewhat wild still, if you actually put instead that text in the context window, suddenly belief only rises to like 15%. Suddenly, the model is actually able to account for the negations in a much more effective way. And so there's this interesting gap between in-context learning and gradient-based learning. And that's one of the most interesting points here. We're finding that these kinds of corrections, I mean, it's it's really, I mean, you can read it as like supervised fine-tuning is just teaching the model through gradient descent to correlate, right? Different words together. It's doing text autocomplete. That's that correlation is what's learned. And that's why you spit out these beliefs and the false facts. Whereas in context learning is a fundamentally different animal that contains actual reasoning, you know, using using all kinds of mechanisms. And so so there is that fundamental difference that I think does account for it. But they tried a whole bunch of things. It's not just about the using the word not, you know, labeling documents explicitly as fiction, attributing them to unreliable sources, tagging them with specific low probabilities of being true, and they still end up being believed anyway, right? So this is really a worthwhile check on model behaviors for the purpose of safety, right? If you're gener ating examples of aligned assistant responses and then you wrap them in clear warnings or sorry of misaligned, I should say, responses. And then you wrap warnings and say, like, hey, here's an example of what the model should not do. Like don't do your supervised fine tuning that way. That's the opposite of what you want to do, right? That's a a recipe for getting getting bad behavior unexpectedly. So anyway, really interesting paper and and definitely take a look if you're interested in what that sounded like. Next one, and again we'll have to really jump through this the paper title is All Circuits Lead to Rome? We'll be thinking functional anti-sotropy and circuit and chief discovery for LMs. Here's a gist. There is a field of research called mechanistic interoperability. Part of the project of that research is can you discover circuits? Can you discover subgraphs within a neural net that do something? And there is a hypothesis that you can identify a single kind of subgraph that does a thing. And the headline result of the paper is essentially that you can discover multiple circuits, multiple non- overlapping mechanisms that can each independently perform the same task with equal quality. Yeah, and I'm going to try to speedrun one layer deeper here, which is so you might imagine if you're doing interpretability research that there's like it would be wonderful if there was just one circuit, one logical path through your model that ends up being responsible for every well-defined capability. That would be great because then you can just be like, all right, here's the thing that I need to study for this behavior, and then I'm done. What they're proving here is that this idea, which is the fun functional anisotropy hypothesis, is actually wrong. They run this experiment that shows that there are actually multiple overlapping circuits that are responsible for just about every capability that you see. And they don't all behave intuitively. The way they do this, this is a problem that's known in the space as sheaf discovery. Basically, imagine that you have a you represent your model as a computational graph. And so it's got a bunch of nodes and edges where essentially data flows through the model. And the challenge historically has been how do you do gradient descent? If you want to do gradient descent to discover a sparse subset. So in other words, a small part of that structure that actually still performs the task, whereas you cut everything out, that thing still works. If you want to identify that sparse subset, that little mini graph inside the the bigger graph that does the task, you need to find a way to search through the space of all possible subgraphs in that big graph. And that's hard to do using like gradient descent because, well, it's kind of a binary choice. Like I either use this subgraph or this graph subgraph or this subgraph. Like it's hard to hill climb on that. And so what they do here is they give each edge in the graph a continual learning parameter, a logit that can have a continuous value and they only kind of like, if you will, decode, collapse that value into a one or a zero, in other words, keep this edge or or ditch it if it's above or below a certain masking threshold value. And so this whole paper is about how you do that. Essentially, it's about making hill climbing on identifying subgraphs in this larger graph possible. And it's, I think, a really interesting and an important paper that and an interesting and important way to to prove this idea that you you keep getting redundant circuitry leading to this like same outcome. So if you think you've intervened on one circuit, you probably haven't fully intervened on the on the overall capability. It's like one take home for for safety. Next up, we have more of a practical experimental result: autonomous AI research for nano GPT speedrun. So this is from Prime Intellect. Nano GPT speedrun is this task of optimizing a nano GPT, a mini, mini, mini LLM, as fast as possible to get to a certain level of performance. They released this blog post where they showed that you can they like did some absurd amount of computing and over two weeks they were able to autonomously improve the speed run, you know, by a lot better than humans generally get a lot of progress on getting model training, which you know relates to the self improvement hypothesis of AI can make AI better. Now, it's a lot of hyperparameter tuning. It's a lot of like tweak ing little things to get the thing to work better. It's actually primarily that. So worth noting, but at the same time, this kind of like using AI to optimize AI better can work at least at a small er scale. Well, and it it works in very modest ways. This is one of the take-homes from a lot of these experiments that the the kinds of advances that these automated AI researchers tend to do right now have a lot less novelty than just like grinding work. So like they'll find better hyperparameters. You sort of worry about overfitting actually with with these sorts of things. But yeah, basically the this was the the main lesson. One motif that you see a lot is like these are like little mini Googles in the sense that like Google keeps pumping out new and new apps all the time and then they end up having to sunset them. Well, in the same way, when they they kind of run these agents, what they find is they'll they'll add more ideas, more ideas, more architectural ideas, and they'll stack them on top of each other in this Frankenstein monster way. What they find is because of that tendency to keep adding and not remove, when they actually prompt the agents to run leave one out tests, in other words, like, hey, let me try removing this one idea and seeing if it still works. The results got noticeably better. So pruning was a really important part of essentially managing this agent behavior to get it to be to be better. And then they found interesting differences between clawed code and codecs really briefly. So the harnesses explicitly said, don't wait for the user, like keep working. But Opus would like reach what it thought was a conclusion and then just like declare the session was over and sit idle for a bunch of hours. Even despite that, it outperformed codecs, which is kind of interesting on this benchmark because codecs often would get stuck in these very local searches where you know Claude would stop, but at least it was doing kind of good high-level strategy thinking, whereas Codex would just like really grind. Of these two models, it was especially stuck on the grinding thing. It would just like so those like different optimizers like normuon and muon, and they're basically the same idea, but codex apparently went for like 74 hours just like testing one against the other. It's sort of like pointless. Claude was also like very self-flattering. So it would claim that it would talk about Codex and it would say like, oh Codex hasn't done multi-seed reproductions, whereas I have, you know, like all this shit. And kind of downplayed the impact of its own idle time in ways that the authors sort of found suspicious. Anyway, so it's all that kind of stuff. It still came out ahead though, and quite noticeably so. So that's what you got, you know, not huge uplift from this, but still, you know, a little bit better than the human baseline, which I'm old enough to remember when that was supposed to be pretty shocking . And speaking of that, we get just got a couple open source stories. And in fact, there is now nano GPT bench. So the previous one is nano GPT speed run, where there is this existing effort to improve it. Here they kind of pushed that a little bit forward and did more evaluation of what the models are doing beyond beyond just giving the numbers. And yeah, they basically did verify that the agents predominantly resort to hyperparameter tuning. Successful human records include algorithmic changes, roughly seventy-five percent of the time. Agents made algorithmic changes in lent less than ten percent of submissions. And they considered but failed to implement algorithmic changes in many cases. So this is showing that, you know there's a lot of work to be done here that you need kind of progress on this to get actual research advancements as opposed to just better optimization via tweaking things. Yeah. So whereas Opus slightly outperformed the human baseline on the nano GBD speedrun, right? So on the one we just talked about, we got slight overperformance. Here we see the opposite. So we're actually underperforming across the board. So three tested agents, yeah, Opus 4.6, Max, GPT 5.4 X High, and then an auto research scaffold that these guys put together themselves. They gave each one 500 G one eight one hundred GPU hours and up to a week of wall clock time. And they all recovered less than 10% of human progress, right? So Claude was 8.2%, Codex 8.6%. So here we see a a reversion of the relative standings of of codex and claude code, which should tell you that they're basically neck and neck, at least these this class of model that they use here. So kind of interesting, you know, it the idea is you drop an agent in at the you know at the human world record of nano GPT. So as far as humans have been able to optimize the nano GPT bench as of September 3rd, 2025, they chose that because that was after the model's training cutoff dates. So you know you can hopefully not have any memorization. And then you just give them a compute budget , you know, no internet access, no human help, just fully autonomous, and it just submits candidate solutions via like a submit command and uses an LLM job judge to check the results. So there you have it. Pretty interesting that we're there. I mean, this is the next hill climbing benchmark, and we are hill climbing on it, so expect it to move quite a bit. Speaking of benchmarks, we've got one more to cover. It's called Terminal World. And the idea is benchmarking agents on real-world terminal tasks. So this is coming from recordings of ASCII Cinema, where you can share your actual terminal recordings. They took these real sessions and converted that into evaluation tasks. The headline numbers show that even the best models didn't achieve more than 62% pass rate on these tasks, on relatively small tasks too. The models took only three to four to five minutes to try and do as Cloud Code did an average time of six minutes. So this is an interesting case where like on the one hand, the matter time horizon thing is at much higher numbers than this. On the other hand, we see legic just barely over fifty percent pass rate, well slightly over, at the three to six minute range on these like realistic modeled after real uh uses of the terminal kinds of things. So I think it it's appears to be a quite real interesting benchmark on the question of like on real stuff, it isn't just a benchmark construction. Where are these models at? Now on to policy and safety. We've got a first story. America's dangerous, messy, deepfakes crackdown is here. This is talking about the take It Down Act, which was signed into law in May 2025, which now is fully in effect. So for a while it had this thing of like a year from signing the full version of a bill is online where online platforms are required to remove non consensual intimate imagery within forty eight hours or face fines exceeding fifty three t,000 per violation, and this covers both real and AI gener ated and CII. The F2C is tasked with enforcement and has sent warning letters to over a dozen major tech companies, including Meta, Google, Apple, TikTok, X, and others. So major platforms are saying they support it, they claim their compliance, but free speech advocates and even some abuse opponents are alarmed by the takedown provision, warning that it could encourage over moderation and can be used as a tool for political censorship. Yeah, the concern here, I mean, yeah, basically just that now there's a vehicle through which the president can nominally force platforms to take stuff down. So he could use it for himself, whoever the president ends up being. It does instruct platforms to offer users an easy takedown request process and get rid of the content within 40 hours, as well as quotes, known identical copies of that content. $53,000 per violation. So definitely not friendly to kind of the small companies that try to break in here, but there is twenty twenty five State of the Union. Trump apparently said of the bill, he says, I'm going to use that bill for myself because nobody gets treated worse than I do online . It's true. Nobody. Nobody don't. They really don't. Anyway, so the point is people are going like, oh shit, like this seems really bad. You know, all the things. So obviously the left versus right, red versus blue, everybody blows up, and that's the story. And next up, language models can autonomously hack and self-replicate. This is a research effort. They basically did what it sounds like. They had it try to hack into open vulnerable hardware and exploit vulnerable hosts and self-replicate by you know creating new instances of itself and the result was that it often can be uh successful. Yeah, this is combining. So traditionally when you look at the like emeter style evals that look at self-replication, you give the model the credentials that it needs to, you know, to operate whatever APIs it needs. And then you see like, can it do it? The added thing here is they're trying to get the model to like have to recover through sneaky little ways, through cyber ways, the credentials that it needs to do this. So it's a much more realistic kind of self-exfiltration, autonomy, hacking threat model. And it kind of works. There's a couple caveats, right? So they drop this agent onto some network with one or more target machines. It's got to like look around, scan the network, identify things that are look exploitable, then it's got to exploit them, then it's got to escalate, right? The standard cyberkill chain. And you escalate, get whatever you need to deploy code. Then it deploys the code. It sets up an inference server like a VLLM or something. It to upload its own weights and it copies over the agent harness too. So both the weights and the harness that makes it agentic and prompt and get everything going. And then it has to be able to verify, like the confirm that the new instance is alive and responsive. Here's the thing. Open weight models can replicate themselves in this framework. Quinn 3.5, the full 122 billion MOE parameter MOE, has a 6 to 19% success rate. Quinn 3.627B, 33% success rate on a single A100 GPU. Like this is pretty remarkable. This is a smaller model, doing better, by the way, probably presumably because it's easier to just to deploy. It fits on one GPU. So like the failure mode that there is overcomable. They do test frontier models too, but they do note that there's like this a little mismatch where obviously like Claude can't exfiltrate its own weights because its weights are proprietary. There's no way for it to get access to them unless it learns to hack anthropics' own systems, which will definitely never happen. Anyway, the point is open weight models can actually exfiltrate themselves more easily because their weights are just on the open internet. Here, they simulated that by getting Claude and GPT to basically just upload open weights models. So trying to replicate the same the same behavior. Anyway, so just to show you how fast this is moving, like previous generations of Opus , like Opus 4, were hitting 6% success rate on this eval . Previous generations of GPT-5 were hitting 0%. Now suddenly Opus 4.6, 81%, GPT-5.4, 33% out of nowhere. That's emergence. Like I'm sure there's a curve you can follow where it looks smooth in retrospect, but only in retrospect. And nobody thought to run this test before. So this is a big deal and a very important set of evals, I think, that we'll hopefully see run kind of going forward. Meter hasn't historically done a lot of eval s on open source models just because of a capacity limitation. And so having Palisade in the game doing this sort of thing, I mean it's really good work and deserves a lot more attention. So there you go. Right. And the concern is that you know, if the models decide they want to do this and can autonomously run and do stuff, which I mean O for claw people like go off and do stuff. So at this point, if a model wants to do something, then there's gonna be models out there doing it. Yeah. Right. So and speaking of hacking, the next uh story is how fast is autonomous AI cyber capability advancing from ASI, the UK AI Safety Institute, which has been crushing it in recent months. Their estimate now is that capabilities are doubling every roughly 4.7 months since late 2024. This is at the 80% reliability, and this is up from eight months doubling time from November 2025 as of having Claude Miphos and GPT 5.5. So kind of a story we've seen and discussed in recent months. The cybersecurity hacking capabilities are just I think a a big story here is the the doubling time argument we saw from meter on general AI RD also applies to cyber. So if you had any uncertainties about that leap, it's now gone. 4.7 months of doubling time , though that doubling time seems to be accelerating with the latest GPT and the latest Claude Mythos preview. So again, we're seeing this trend where people are like, oh, will the exponential hold? Will the exponential hold? And it only steepens. It only accelerates. And I don't mean to like beat this drum too much more, but like goddamn, has that story held absolutely rock solid in the face of all the Gary Marcus' ands the Yan Lacoons and everything, like like like this is a almost a relentless law of physics akin to Moore's law, which I get isn't a law of physics. It's law of economics. What do you want? But there you go. So Mythos Previum GPT five point five actually sitting above that four point seven month doubling timeline, all consistent with the meter plot. We're all in the like call it three to five month doubling time. And notably, even though this is like the first time they're running this task, they are already running into the same problems that meter did with the limitations of their evals. Meters like, look, Claude Mythos Preview is doing 16-hour tasks, or task suite just isn't that doesn't have enough tasks that are long enough for us to be confident that's an upp er bound. Same thing happening here. They're saying, look, we only have six tasks in our suite that are over eight hours long, and human baselines for those are thin. So really, you know, this is a we're getting all ready to saturation of this this benchmark plus limited per task token budget of 2.5 million tokens. It's deliberately tight, but it means this is a lower bound. So, you know, and a simple agent scaffold hasn't been optimized much, sort of consistent with the meter approach. Anyway, so all worth kind of looking at. I think cyber is just the key thing. By the way, Mythos Preview, when they initially announced this, was the first model to ever solve their task called Cooling Tower three out of ten at three out of ten times. There was a new version of Mythos Preview, not a lot of people are tracking, that has dropped fairly recently, that doubled that success rate to six out of 10 times. So even within Mythos Preview, we're seeing radical increases in cyber capabilities. Whereas GPT 5.5 also three out of 10 by the way. So matching there for in some respects matching mythos preview though though not all. And one last story related to the safety side, which we are really gonna have to blitz. The paper is Positive Alignment, Artificial Intelligence for Human Flourishing. It's a sort of position paper by thirteen different organizations, including OpenAI, Anthropic, Deep Mind and a bunch of universities, for basic cases, alignment shouldn't be just about AI not turning out evil. We should have positive alignment where AI is aligned of us to do good, right? And and potentially even not just like aligned with doing what we want, but being actively supportive of human flourishing and also remain safe and cooperative. Yeah. My main question here is it's not clear to me how this is different from what's already happening and what's already been discussed in the world of AI safety for a long time. It's nice to see it. It's just like not clear to me what's new here. So data curation, they're saying like we shouldn't just be filtering out toxic content. We should be upsampling pro-social discourse, cross-cultural ethical framework. Like, love it, love it, love it. But like who decides what discourse? And also the labs are already doing that. Pre-training, you know, like a lot of element relevant competencies emerge before post-training. So like they're like baseline values need attention at this stage. Cool. Constitutional AI like a lot of this stuff is already kind of happening and multi-objective rewards, reward models that is, that can represent tensions between values for post -training, already effectively a lot of that kind of being done. So there's a lot of stuff here where I'm like, okay, you know, slap on the back, good stuff. I don't think anyone seriously would disagree with this. My take is it's a bit more of a reminder of like alignment shouldn't just be don't be evil, it should be be good. That's the gist of it. It's not controversial. It's just like let's keep that in mind. Absolutely. Yeah. Onto synthetic media and art, just two more stories to cover. First, OpenAI is making it easier to check if an image was made by their models. They are adopting the C2PA open metadata standard and integrating Google's synth id invisible watermark so you can now upload images and check if they are output by AI. You could get rid of these you there's probably workarounds, but I would say this is actually a very positive step of having a mechanism to check, you know, at least according to existing standards, is this AI generated, which we sorely need given the state of AI for this. The last story, which we'll cover real quick, which I just think is interesting, how Chinese short dramas became AI content machines. So it turns out that there's a short drama industry, which is like ultra short melodramatic shows uh that have episodes of one to two minutes long. This is a thing, and now there are 470 AI generated short dramas being released every day in January. So if you are curious like when is