PodcastsHear the voice. See the shape of the thought.

Sebastian Mallaby spent three years and over 30 hours with Demis Hassabis in a British pub to write *The Infinity Machine*, and this conversation pulls the most underreported threads from that access: the 2015 safety summit that accidentally spawned OpenAI, the secret billion-dollar spinout plan Demis never used as real leverage, and the quasi-spiritual conviction about God and science that Mallaby never expected to find. The throughline is a paradox — Demis understood the race was dangerous from day one, but as leader of one lab, even a Nobel Prize-winning one, he could not stop it. ## [00:00] Intro Jacob Effron sets up Sebastian Mallaby as someone who has spent more time with Demis Hassabis than almost any journalist alive — 30-plus hours across three years of pub sessions in London. Mallaby's book, *The Infinity Machine*, covers the full arc of DeepMind from its 2010 founding through the Nobel Prize. The clips previewed here — Demis banging the table about God and science, Reid Hoffman's billion-dollar pledge, and the Elon feud — all come from later in the conversation. > *"Demis has a Nobel Prize. Sam didn't finish his first degree. Therefore, Demis doesn't take Sam very seriously."* ## [02:04] Was the AI Race Inevitable? Mallaby's verdict: yes, inevitable. Any technology this powerful would attract multiple labs across multiple countries, and China's stack was already competitive despite semiconductor shortfalls. What makes the story poignant is that Demis didn't believe this in 2010. He genuinely hoped one lab could carry the AGI project safely to the finish line — a singleton scenario where DeepMind was the anointed team. By the mid-2020s he had swung to the opposite pole: safety is a collective action problem that only governments can solve, because no single lab's restraint can bind the others. > *"I think it was inevitable. When you have this sort of supremely strong technology, there's going to be multiple labs in multiple countries that are just desperate to try and build it."* ## [04:03] The 2015 Safety Summit Backfire Summer 2015, SpaceX headquarters: Demis convenes a small summit to bring Elon Musk inside the tent — the plan was for Elon to chair a safety oversight board and, critically, not launch a competitor. By end of year, OpenAI existed. Mallaby frames this as the moment Demis internalized that voluntary collaboration between lab leaders is structurally impossible. The only mechanism he now believes can work is a government enforcer setting uniform rules — mandatory pre-release testing, safety slow-downs — with US-China cooperation as the endpoint, however remote that prospect appears. Jacob pushes on whether lab leaders actually believe government intervention is achievable; Mallaby draws a parallel to the FDA: slow, imperfect, but it does adjudicate whether drugs are safe enough to ship. > *"You can't trust the other guys. The only way you get trust is if you have a government enforcer that comes along and says, 'Here's the rules for everybody. There's going to be a level playing field. You're all going to have to abide by some sort of safety slow-down.'"* ## [11:27] Why Google Doesn't Make As Concentrated Bets Jacob points to the two defining consumer-AI moments of the era — ChatGPT and Claude Code — and neither came from Google DeepMind despite its leaderboard dominance. Mallaby traces this directly to Demis' intellectual formation: a PhD in neuroscience, a broad theory of intelligence, a lab culture that says "whenever there are two paths, do both, find a third." The result is a heavily hedged research portfolio that is excellent at producing Nobel Prizes and state-of-the-art models but structurally slow to make the kind of one-directional product bet Anthropic made on coding. Gemini is bundled into Google Search, so usage is higher than it appears — but Mallaby concedes the product-zeitgeist gap is real. > *"Anthropic got to coding because it was willing to take a more concentrated bet. It never went into the whole field of, you know, everything at once."* ## [15:51] Project Mario: The Secret Spinout Plan The book's most explosive scoop: DeepMind had a secret plan — code-named Project Mario — to spin out of Google, backed by a $1 billion pledge from Reid Hoffman. Mallaby had to fight Google's general counsel to publish it. The motive was not entrepreneurial independence but safety leverage: Demis wanted formal safety oversight over DeepMind's models, Mountain View wasn't providing it, and a credible spinout threat was his negotiating chip. He never explicitly told Google about the Hoffman pledge, but pushed hard knowing the option existed. In the end he chose to stay — legal risk of the spinout fight, desire for compute access, and a preference for doing science over litigating corporate structure. A year later he shipped AlphaFold and won the Nobel Prize. > *"Demis really really wanted to get safety oversight over the Google DeepMind models. Google corporate in Mountain View wasn't doing that. So he had to have a credible threat of spinning out. He went to Reid Hoffman. Reid Hoffman pledged a billion dollars to finance a spinout — and Demis used that to kind of pressure Google."* ## [19:43] What Demis Actually Regrets On AlphaFold and AI-for-science: no regrets at all — Mallaby argues it was not only scientifically correct but politically necessary, because AI needs visible social benefits to survive the coming backlash against job disruption. The genuine regret is speed. Demis missed the transformer moment the way Ilya Sutskever did not: when the paper dropped, Ilya ran down the corridor to find Alec Radford to build a language model. Demis' broad-portfolio instinct meant DeepMind studied the transformer but didn't bet the lab on it. Missing that window — and the ChatGPT moment that followed — is a real failure, not just a stylistic difference. > *"Ilya is like jumping out of his chair, running down the corridor going to find Alec Radford saying, 'Hey, we're going to build a language model based on this transformer architecture.' On the day they won AlphaGo, Demis was already on to bio — and someone picked it up on a mic."* ## [23:46] Venture Startups vs. Tech Behemoths The broadest structural argument in the episode: does venture-backed concentration beat hyperscaler breadth in AI? Mallaby has written about both (his previous book covered venture capital) and calls it genuinely balanced. Hyperscalers have unlimited capital and can sustain a multi-year arms race; the problem is that unlimited resources breed portfolio thinking, which bleeds attention. Startups with one concentrated bet can move faster on that specific bet. Mallaby's live position: OpenAI has roughly 50/50 odds of being absorbed or failing before next summer — not because the tech is weak, but because the business model can't sustain indefinite losses against Google's balance sheet. He also floats that Anthropic should IPO right now while its brand is strongest. Jacob notes the robotics parallel: fifteen different approaches being funded simultaneously, and whoever picks the one that works the way transformers did will dominate. > *"I wrote in the New York Times in January that I thought OpenAI had a 50% chance of going bust by next summer. Is it still 50? Yeah. The tech is great. It's just the business model — and you're up against Google, which just has unlimited amounts of cash to spend you into the ground."* ## [34:08] David Silver and the RL True Believers David Silver — AlphaGo's lead researcher and co-author of the "reward is enough" paper with Rich Sutton — left DeepMind after the book came out to start a new company. Mallaby reads the departure as structurally inevitable: Silver is a pure reinforcement learning absolutist who believes learning from human data is fundamentally inferior because it encodes human errors. His thesis is that self-play and environment-generated experience is the only path to genuine superhuman performance. Demis told Mallaby this view may ultimately be correct *after* AGI is achieved — but the entire language model revolution showed that bootstrapping with human data is what gets you to AGI in the first place. Silver's RL purism was too far ahead of the current paradigm for his colleagues to follow. > *"David is just very very hard over on that vision — learning from data is inferior because the data includes mistakes. The machine needs to learn from its own experience, not rely on the crystallized knowledge of humans passed on through text."* ## [38:21] Demis, Elon, and the Evil Genius Feud The origin story: at a Founders Fund LP offsite in 2012, Elon argues that SpaceX matters most because even if AI wrecks Earth, humanity can move to Mars. Demis replies that his AI will eventually conquer space flight and follow them there. Elon goes quiet, then writes a $5 million check into DeepMind's Series B. Two years later, hearing Google was acquiring DeepMind, Elon and Luke Nosek Skyped Demis from a party closet in LA in the middle of the night, begging him not to sell to Larry Page. Demis said no, hung up, and Elon started calling him "evil genius" — the name of a video game Demis had designed. Mallaby characterizes Demis' view of Sam Altman as colored by the credential asymmetry: Nobel Prize winner vs. someone who didn't finish a degree. The relationships between these founders are less professional rivalries than a collection of specific personal slights and competitive provocations playing out over fifteen years. > *"Demis says, 'Yeah, but if you think you're going to be safe on Mars, remember that my AI will be able to conquer space flight, and it will just follow you to Mars. So then you won't be safe after all.' There's a silence. Then Elon goes, 'Hm.' And then: 'I'd like to invest in your Series B.'"* ## [42:39] Great Man Theory vs. Inevitability Jacob cites *The Economist*'s framing of the book as a test of great-man theory. Mallaby draws a parallel to his Greenspan biography: Greenspan understood bubbles were dangerous (literally the subject of his PhD), yet couldn't stop the 2008 crisis. He considered titling the Demis book *The Man Who Knew* for the same reason — Demis knew from the start this technology was dangerous, but one lab's restraint cannot bind the rest. Individual leaders do matter at the margin: Dario Amodei changed the safety narrative through the Anthropic mythos release; Sam Altman shaped the race by shipping ChatGPT while it was still hallucinating; Demis shaped it by persuading Rishi Sunak to host the UK AI Safety Summit. But the race itself? Structurally overdetermined. > *"I feel that one could have almost used the same title for the Demis book — 'the man who knew' — because Demis has known from the beginning that this thing is dangerous. But as the leader of one lab, even a very powerful rich lab, even he with his stature as a Nobel Prize winner — what can he do?"* ## [45:00] What Demis Didn't Want Published The detail Mallaby least expected: Demis is driven by something close to a spiritual conviction about science. In those two-hour pub sessions he would bang the table about the mystery of matter — why atoms cohere into a solid table, why silicon and copper can think — and say, unprompted, "Maybe if we approach science the right way, we will be getting closer to something that we could perhaps call God." Mallaby reads this as the psychological engine that lets Demis keep pushing a technology he knows to be dangerous: it's a quasi-spiritual quest, not just a commercial one. On what Demis blocked from publication: his family (he set that limit at the start), and his internal fights with Sundar Pichai — he didn't want to destabilize the Google relationship he still depends on. > *"He would start banging the table and saying, 'Maybe if we approach science the right way, we understand more about nature. We will be getting closer to something that we could perhaps call God.' I had no idea he would feel that way."* ## Entities - **Demis Hassabis** (Person): Co-founder and CEO of DeepMind / Google DeepMind; Nobel Prize winner in Chemistry (2024) for AlphaFold; central subject of *The Infinity Machine*. - **Sebastian Mallaby** (Person): Staff writer at *The New Yorker*; author of *The Infinity Machine* (Demis Hassabis biography) and a prior book on venture capital; spent 30+ hours with Hassabis over three years. - **Jacob Effron** (Person): Host of *Unsupervised Learning*; Managing Director at Redpoint Ventures. - **Reid Hoffman** (Person): LinkedIn co-founder; pledged $1 billion to finance DeepMind's potential spinout from Google under Project Mario. - **David Silver** (Person): Lead researcher on AlphaGo and AlphaZero at DeepMind; co-author of the "reward is enough" RL paper with Rich Sutton; departed DeepMind post-publication to start a new company. - **Elon Musk** (Person): Hosted the 2015 AI safety summit at SpaceX; early DeepMind investor; coined the "evil genius" nickname for Hassabis after DeepMind sold to Google. - **Sam Altman** (Person): CEO of OpenAI; shipped ChatGPT in late 2022 despite hallucination issues, which Mallaby argues irreversibly shaped the AI race's trajectory. - **Dario Amodei** (Person): CEO of Anthropic; credited with changing the AI safety narrative through the mythos paper release and his public Pentagon confrontation. - **DeepMind** (Organization): Google subsidiary; founded by Hassabis, Shane Legg, and Mustafa Suleyman in 2010; produced AlphaGo, AlphaFold, and Gemini. - **Project Mario** (Concept): Secret DeepMind plan to spin out of Google, backed by a Reid Hoffman $1B pledge; used as negotiating leverage for safety oversight, never executed as a real spinout. - **AlphaFold** (Software): DeepMind's protein-structure prediction model; won Hassabis the 2024 Nobel Prize in Chemistry; shipped in 2020, one year after he declined the spinout option. - **Reinforcement Learning** (Concept): Machine learning paradigm central to AlphaGo and AlphaZero; David Silver's absolutist commitment to RL (learning from environment experience over human data) created internal tension at DeepMind and ultimately led to his departure. - **The Infinity Machine** (Concept): Sebastian Mallaby's biography of Demis Hassabis; nearly titled *The Man Who Knew*; published with the full Project Mario scoop over Google's objections.

Ethan He built NVIDIA's Cosmos world model, then joined xAI mid-2025 to build Grok Imagine from scratch — no infra, no data, no model — and shipped the first audio-video generation model in three months. He walks swyx and Vibhu through the full technical stack: synthetic captioning pipelines, VAE design tradeoffs, step distillation, audio-video alignment, and the hard economics of storing petabytes of video training data. His central argument runs through the entire conversation: since diffusion model technology has largely matured, most quality gains in video now come from language models, not from the video model itself — a view with direct implications for where the field goes next, including video agents, generative UI, and embodied world models. ## [00:00] Hook This exchange — Ethan's "pretty big claim" that visual intelligence now mostly comes from language — is pulled from later in the interview, where he argues that improvements to video models are increasingly driven by better language models acting as prompt rewriters and orchestrators, not by advances in diffusion or flow-matching architectures themselves. > *"Every time you see there's some improvement on these models, I would say mostly the gain comes from language model, not coming from the video model itself."* ## [01:16] Introduction swyx and Vibhu welcome Ethan to the Latent Space studio, noting he has been a recurring presence through the podcast's paper club — first presenting the Cosmos world model paper, then mixture-of-experts work. The conversation opens with a brief aside about the Poolside paper released the same day, a fully open Gemma-level model trained on 40 trillion tokens, before pivoting to Ethan's own trajectory. ## [02:41] From NVIDIA Cosmos to xAI Ethan built Cosmos — NVIDIA's giant video foundation model aimed at giving roboticists a simulatable world to build on — and shipped it by end of 2024. Once he realized video models obeyed the same scaling laws as language models, he went looking for more compute. xAI offered it. He joined in mid-2025 at the moment xAI decided to build its own image and video stack, with no existing infra, data pipeline, or model. He stayed through pre-training, post-training (reference-to-video, video extension), and a final stretch leading a small team on real-time long-horizon video generation. > *"By the time I joined, xAI was about to build video models and multimodal models. There were no infra, no data, and no model. Just a few engineers — we built it in three months and released the first model, Grok Imagine 0.9."* ## [04:40] Building Grok Imagine from Zero to One The three-month timeline surprised even Ethan. He attributes it to three factors: talent density (strong engineers who could align on a goal with minimal meetings — typically just one sync a day), xAI's existing data and inference infrastructure, and his own prior experience running the same build at NVIDIA. The bottleneck was iteration speed: how many training runs can you complete per day. With strong infra and abundant compute, bugs surface faster and each failed run costs less, so you burn through the inevitable data and pipeline errors in weeks rather than months. > *"The most important thing is talent. Everyone was very strong and clever, very close to each other toward a common goal. So that speeds up things a lot — you reduce the communication bandwidth among people."* Ethan describes a pattern where small data or pipeline bugs produce outsized quality regressions, and only fast iteration exposes them. A bug invisible at one scale becomes catastrophic at the next. The engineers who find and fix these quickly — not the ones who design the most sophisticated architecture — determine how fast a team ships. ## [11:23] How Image and Video Models Are Trained Video models require synthetic text-video pairs because internet video titles and descriptions almost never describe visual content accurately. The first step is human labeling: at NVIDIA, annotators were instructed to describe every object, character, interaction, and dialogue in a clip as exhaustively as possible. Those labels train an early VLM, which then generates captions at scale. The resulting pipeline — video to VLM to synthetic caption to (video, caption) training pair — is the foundation of both Cosmos and Grok Imagine. Image models must come first: they train faster, require less storage, and the learned representations transfer directly to video. Ethan describes building image models as building the foundation that video sits on top of. The architecture — diffusion transformer operating over VAE latents — is now standard, but the data quality and caption detail remain the primary lever for model quality. > *"Building a video model, you actually need to build an image model first. The data you need is 100% synthetic pairs of language and image, or language to video — because on the internet, videos don't naturally associate with text."* ## [20:09] Video Compression, VAEs, and Real-Time Tradeoffs Raw MP4 compression produces tokens whose latent space is incomprehensible to transformers, so the field moved to learned VAEs that create a smoother, more continuous latent space models can train on. The key design choice is how aggressively to compress the temporal dimension. Temporal compression is efficient — adjacent frames are mostly redundant — but it trades away real-time capability. Wan 2.1 uses 8x8 spatial and 4x temporal compression; generating a single token requires reconstructing four frames, making sub-200ms latency impractical. Ethan frames this as a fundamental tradeoff: high compression rates make training cheap and inference efficient for pre-rendered video, but lock out any use case that needs to respond to live user input. World models require the opposite choice. ## [23:26] Generative UI, Flipbook, and Neural OS Ethan argues that if inference were free, the logical endpoint of video generation is a complete replacement of conventional UI: instead of loading web pages from a server, a model generates them in real time in response to user intent. Flipbook, a demo that went viral, shows this literally — every element of the "browser" is generated by an image model, and clicking a link generates a new page rather than fetching one. The deeper claim is that this is not a novelty but the final form of world models applied to human-computer interaction. A traditional app is a fixed function mapping input to output; a generative UI is a model that can produce any interface the user needs without a developer having to build it first. Ethan calls this a "Neural OS," where the gap between user intent and rendered pixels closes entirely. > *"Imagine the internet doesn't exist and you type in google.com — what should a model show you? The model can imagine something. These web pages completely do not exist, so I can explore anything."* The near-term constraint is inference cost. Current video models cannot generate at interactive frame rates without significant distillation. But Ethan treats this as an engineering problem with a known solution trajectory, not a fundamental barrier. ## [33:26] The Cost of Training Large Video Models Training large video models costs roughly as much as training a medium-scale language model, but the breakdown differs. Compute is comparable, but storage and data movement dominate in ways LLM practitioners do not expect. One billion videos at 5 MB each requires five petabytes of raw storage. The VAE features that must also be stored are roughly the same size again — tens of petabytes total. On AWS S3, five petabytes runs approximately $100K per month before egress. Egress — downloading that data into the training cluster — can exceed storage costs, and each training run pulls the full dataset once. > *"Just storing the videos alone costs a lot. Five petabytes on S3 Standard is $100K per month. And egress — just to download those videos — I believe it's more expensive than storing them, and each training run you probably need to pull them once."* The implication is that video model development is gated on data infrastructure as much as on GPU hours. Teams without efficient data pipelines pay a multiplier on every experiment. ## [38:20] Distillation, GANs, and Fast Video Inference Training-time costs are largely fixed; the inference-time story is more tractable. Step distillation — training a small model to replicate the outputs of a large teacher in far fewer denoising steps — cuts inference cost by 10-25x. Flow-matching models trained to convergence need around 100 steps; production models typically run in 4-8. At the extreme, simple image-to-image tasks can run in a single step. The intuition Ethan offers: the teacher model must learn the full distribution of internet video, which is arbitrarily complex. The distilled student only needs to match the teacher, which is a fixed and much simpler target. Consistency models and LCM-style approaches follow the same logic. In Cosmos, production serving used 4-step and 8-step variants depending on quality requirements. GANs remain relevant as discriminators: a GAN discriminator can enforce photorealism constraints during distillation that pure score-matching loss misses, and Ethan notes that consistency models and GANs are converging on similar practical deployments even if their theoretical motivations differ. ## [42:37] Audio-Video Generation and Grok Imagine 0.9 Grok Imagine 0.9 was the first audio-video joint generation model deployed at scale. The core difficulty is modality alignment: text-video pairs are relatively abundant; text-audio pairs are rare; audio-video pairs aligned at the semantic level are almost nonexistent at scale. Speech tokens are quasi-discrete and can be modeled with language-like approaches, but music is continuous and requires a completely different representation. Training the joint model required building synthetic audio caption pipelines from scratch, with human annotation where VLMs failed — which was often, especially for music. Aligning all three modalities — text, video, and audio — without either degrading video quality or audio realism is what Ethan calls the hardest part of the project. > *"Audio has two components: a discrete component — language — and a continuous component — music. The music is completely different; you cannot model it with discrete tokens. That's the hard part, not to mention we have to align text, video, and audio together."* ## [49:50] What Makes a World Model? Ethan's definition has three components: real-time, interactive, and long-horizon video generation. He treats these as independent requirements, each of which most current models fail. Real-time means generating at display frame rates — 60fps for casual use, 300fps for gaming, 200ms response latency for digital humans. Current video models cannot do this; the VAE's temporal compression alone introduces latency that makes sub-200ms responses nearly impossible without architectural changes. Interactive means the model can accept any input modality the user can provide — keyboard, mouse, voice — and respond coherently. Long-horizon means maintaining consistent physical laws, character identity, and causal logic across minutes, not seconds. > *"World model is real-time, interactive, long-horizon video. Current video models can do none of these three things fully. That's why they're not world models yet."* ## [57:07] Reference Videos, Long Context, and Video Memory The parallel to language model context scaling is direct: video models are in the 2,000-8,000 token era, and will need to scale to million-token-equivalent contexts to generate coherent long videos. Ethan describes the reference-to-video feature he built at xAI (analogous to Cameo) as a mechanism for injecting selected history into the model's context rather than carrying the full video forward. FramePack's heuristic — storing the last second of video at full resolution while compressing earlier frames progressively — points toward the right direction: the model selects relevant context from its history rather than brute-forcing the full sequence. Ethan expects this context management to become part of the model itself rather than remaining a harness-level heuristic, the same way KV cache management is disappearing into model internals. ## [61:27] xAI Culture, Research, and First-Principles Building swyx notes that xAI communicates its research poorly relative to what the work actually demonstrates — the blog post accompanying Grok Imagine describes high-level capabilities without the technical depth Ethan has just spent an hour covering. Ethan is diplomatic but agrees that different labs have different communication styles. The xAI working culture he describes is minimalist: few meetings, no bureaucratic overhead, direct access to leadership judgment on technical decisions, and extreme iteration speed enabled by a strong infra team. The tradeoff is that company priorities shift fast, which is part of what eventually pushed him toward independent research. First-principles thinking — starting from the physics of the problem rather than from what competitors have shipped — runs through the team's approach to both model architecture and product. > *"Everything you just described is state-of-the-art. Like no one else has done it. And then you just put this blog post with the cookies. I'm like, this is not enough."* ## [71:01] AI Safety, Watermarking, and Prompt Rewriting Grok Imagine deployed watermarks in all jurisdictions requiring them and built takedown pipelines integrated with xAI's social platform infrastructure. On watermarking technology, Ethan is skeptical of SynthID's long-term robustness: the technique is documented publicly, and users on Reddit have already reverse-engineered the exact frequency pattern Google applies and can strip it from any generated image. He expects watermark detection to become an arms race. On prompt rewriting: video diffusion models take instructions literally. If a user types "a cat," the model generates a stationary cat on a white background with no motion, because the training data pairs were maximally detailed descriptions of physical scenes. Production systems layer a large language model as a prompt upsampler — converting sparse user instructions into the detailed physical descriptions the video model was trained on. This is one of the reasons Ethan argues language models are increasingly central to video quality. ## [74:26] Video Agents and AI-Assisted Creation Ethan's central claim from the hook: visual intelligence now mostly comes from language. The diffusion model architecture has largely converged; the gains come from larger, smarter LLMs that rewrite prompts, plan video sequences, call editing tools, and stitch clips together. In Cosmos, the prompt rewriter was larger than the video model itself. Video agents extend this: instead of generating a complete video in one shot, an agent plans the production, calls video generation models as tools alongside deterministic editing operations (text overlays, color grading, cuts), and iterates until the output meets a specification. Ethan predicts that by end of 2025, video agent output will reach production-grade quality — presentable video generated without a human editor in the loop. > *"The visual intelligence are actually mostly coming from language. Every time you see improvement on these models, I would say mostly the gain comes from language model, not coming from the video model itself."* ## [88:48] Why Language Models Unlock Better Video LLMs prompt video models better than humans do, because AI models understand AI models' training distributions. A language model knows that a diffusion model needs explicit physical descriptions, not poetic shorthand — and can generate the right prompt format automatically. Beyond prompting, agents can use deterministic video editing tools for precision operations (exact text overlays, frame-accurate cuts) that probabilistic diffusion models handle poorly, keeping the stochastic model focused on generation and delegating precision to tools. Ethan's timeline: video agent output at production quality by end of 2025, with the inflection point visible in work already shipping. ## [92:31] Robotics, Physical AI, and Embodied World Models Ethan's robotics prediction inverts the usual framing: physical AI may be solved not by deploying robots in the real world but by video world models becoming so capable at simulating physical environments that they effectively provide embodied experience. Once a model can control computer interfaces in real time with full causal understanding, extending that to robotic control becomes a matter of adding one more tool. The path from screen-interacting video model to robot controller may be shorter than the path from current robot learning systems to the same capability. ## [93:54] Why Ethan Left xAI Research ambitions and company priorities diverged. xAI's focus shifted in ways that made certain research directions — particularly on the language model side — impractical from inside. Ethan also notes that the insight driving his departure is the same one underlying his "big claim": if language models are now the primary driver of video quality, the most impactful work to do is on language models, not video models. He frames leaving not as dissatisfaction but as following the evidence about where the leverage is. ## [95:32] Self-Managed Context and the Future of LLMs Ethan's active research question: language models that are aware of their own context state and manage it autonomously, rather than relying on harness-level heuristics like automatic compaction at 80% fill. He draws the parallel to video models struggling with long-horizon generation — the same context management problem appears in both modalities. He points to Claude Code's practice of appending the current timestamp to user messages as an early example of making models context-aware, and expects this pattern to be absorbed into model training rather than remaining an external scaffold. > *"The language models are not aware of how long their own context length is. Once they hit like 80% or something, automatic context compaction is getting triggered, and the model is not aware of that when it's working."* ## [99:59] Ethan's Career Path and Closing Thoughts Ethan traces a decade of transitions: ResNet-era image recognition with the original authors at NVIDIA, self-supervised learning at Facebook AI Research, scaling at NVIDIA Cosmos, extreme-scale compute at xAI. He was rejected from every top PhD program despite first-author papers at top conferences, which pushed him into industry. In hindsight he reads his career as consistently following the scaling frontier — from image recognition to SSL to video to LLMs — and argues that within ML, domain switching is far more tractable than practitioners believe. > *"Within ML, it's actually easier to switch than you think. A lot of people have manifested that 'I work on computer vision, I always have to work on computer vision.' But from my experience, the fundamentals transfer."* ## Entities - **Ethan He** (Person): Former xAI researcher who built Grok Imagine from zero; previously led NVIDIA Cosmos world model; now focused on LLM research - **swyx** (Person): Latent Space co-host; conducts technical interviews on AI engineering and research - **Vibhu Viswanathan** (Person): Latent Space co-host; co-interviewer for this episode - **Grok Imagine** (Software): xAI's image and video generation product; first model (0.9) was the first large-scale audio-video joint generation system - **NVIDIA Cosmos** (Software): Open-source video foundation model for robotics simulation; Ethan's project before xAI; released end of 2024 - **xAI** (Organization): Elon Musk's AI lab; known for fast iteration culture and extreme compute resources - **Flipbook** (Software): Viral demo of real-time generative UI; all interface elements generated by image model in real time - **SynthID** (Software): Google's AI watermarking technology; Ethan notes its pattern has been publicly reverse-engineered - **Step distillation** (Concept): Technique to train a model to replicate a teacher's output in far fewer denoising steps; reduces inference cost 10-25x - **VAE** (Concept): Learned video compression creating smooth latent spaces; temporal compression is efficient but creates real-time latency tradeoffs - **World model** (Concept): Ethan's definition — real-time, interactive, long-horizon video generation; distinct from standard video generation - **Video agents** (Concept): Systems where LLMs orchestrate video generation models, editing tools, and deterministic operations to produce production-quality video - **FramePack** (Concept): Progressive temporal compression approach for long-context video generation; stores recent frames at full resolution, compresses older history

Benedict Evans — independent analyst and former Andreessen Horowitz partner — joins Lenny Rachitsky for a wide-ranging, historically-grounded read on AI's trajectory. His core provocation: AI is exactly as big a deal as the internet or mobile — transformative and uncertain in equal measure — and anyone claiming more precision than that is vibes-forecasting. Across 80 minutes they work through where economic value will actually land (hint: probably not at the model layer), why professional services are booming rather than shrinking, how to think about job displacement without losing your mind, and what the anti-AI backlash does and doesn't tell us. ## [00:00] Introduction to Benedict Evans Evans opens with his signature contrarian opener: "My most controversial opinion is that I think that AI is as big a deal as the internet or mobile — and only as big a deal as the internet or mobile." The framing immediately sets the tone for the conversation — resist the urge to rank transformations on a cosmic scale, and instead study the mechanics of how platform shifts actually unfold. > *"My most controversial opinion is that I think that AI is as big a deal as the internet or mobile and only as big a deal as the internet or mobile."* Lenny sketches out Evans's background: years as A16Z's in-house technology analyst, followed by six years of independent research publishing. His biannual decks — most recently "AI Eats the World" — are widely read by founders and investors trying to cut through noise. ## [02:19] What people aren't pricing in about AI's impact Asked what the market is still missing, Evans reaches for an analogy rather than a prediction. We are, he argues, in a "1997 moment" — the technology is visibly exciting, most of what will eventually be built hasn't been built yet, and nobody in 1997 correctly predicted what the internet would become. He points to survey data showing that even among 13-to-18-year-olds, around 60% still don't use AI at all, while a small cohort of tech workers have essentially restructured their daily workflows around it. > *"If you're going to make the internet comparison it's like we're in 1997. Like it's very exciting. Most stuff kind of doesn't work yet. Most of the stuff that people are going to do hasn't been built yet and it's not really clear how any of it's going to work when it does work."* The key failure mode Evans identifies is the "already there" illusion — early adopters project their own usage patterns onto the rest of the world, missing the enormous variance in adoption and the slow grind of enterprise deployment cycles. ## [06:24] Why we're in the 1997 moment of AI Evans uses the VisiCalc spreadsheet as an anchor. When accountants saw the first software spreadsheet in the late 1970s, it was obviously transformative — a week's work done in 30 seconds. But a lawyer looking at the same demo would think, "that's clever, my accountant should see this, but that's not what I do." AI right now occupies that same diagonal: software developers are the accountants who immediately grasped what Claude Code means for them; most other industries are still in the "lawyer looking at a spreadsheet" phase. > *"Software developers are the accountants seeing VisiCalc — oh my god this changes everything — like before Claude Code and after Claude Code. A lot of other people are picking it up, using it to varying degrees, but slightly puzzled."* This jagged-frontier quality — where AI works brilliantly in some contexts and fails unpredictably in adjacent ones — is precisely why broad adoption timelines are so hard to call. It took 10–15 years after Google Docs for people to invent all the SaaS companies that obviously should have existed. ## [09:44] The unexpected boom in professional services and consultants The counterintuitive data point driving Evans's recent writing: the most advanced AI companies — Anthropic, OpenAI — are simultaneously the biggest buyers of professional services and the fastest-growing employers of human headcount. This isn't a paradox once you think through what actually changes when AI makes certain tasks cheaper. Evans introduces a core distinction: task vs. job. When you hire McKinsey, you are not hiring them to produce a 75-slide deck. The deck is the task; the job is walking all over your enterprise, understanding the politics, talking to customers, and figuring out what you actually need to do. Claude can produce a mediocre version of the deck; it cannot do the job. The same logic applies to accounting: every wave of automation since adding machines has increased the number of employed accountants, because cheaper computation expands the scope of what companies decide to measure and act on (Jevons paradox in action). > *"You could make the same point in software development. Before IDEs and libraries and operating systems, developers had to write all the code. Now if you write an iPhone app, 90% of the code is written for you by Apple... So we've got like a tenth as many engineers now. Well, no."* The e-commerce analog is sharp: Amazon gets you the SKU if you know what SKU you want — "knowing what SKU you want is another job." ## [17:44] Why distribution is becoming the ultimate moat Evans challenges the premise that AI-driven job loss will be fast. Enterprise software sales cycles run 18 months minimum; SAP doesn't get torn out overnight. He cites Frame.io as a case study: there was nothing technically blocking that product 15 years before it launched — the bottleneck was someone realizing the problem existed inside a specific industry and that a specific approach would solve it. The broader point is about organizational change speed vs. model capability speed. Companies can't implement AI transformation without dedicated project teams — which is exactly why consulting and forward-deployed engineering are booming rather than shrinking. The speed of model improvement is decoupled from the speed at which enterprises can absorb the change. > *"Like no, people aren't just going to tear out SAP and replace it with XYZ. Maybe in three, five, 10 years yes, that whole estate will look radically different and all those jobs will have changed — but it will take time sector by sector."* ## [23:17] The coming job transformation: what's real vs. panic Evans leans into historical pattern-matching: every technology wave since 1800 has automated jobs and created new ones, and the new jobs are systematically better than the old ones. The jobs that disappear tend to look dispensable in retrospect; the jobs that appear couldn't have been named in advance. His IBM ad slide makes the point viscerally — a 1950s ad promised that an IBM electronic calculator is "like having 150 extra engineers," which is also the pitch of Claude Code today. The "it's different this time" argument he takes seriously is speed of adoption — AI diffuses faster than previous technologies because it runs on existing internet infrastructure. But he notes that adoption speed and institutional-change speed are different curves, and the institutional one has not accelerated proportionally. > *"This is going to be completely different from everything else — just like everything else."* On whether AI eliminates the lump-of-labor fallacy — his answer is no. Two hundred years of data say otherwise, and the burden of proof is on those claiming this wave is categorically different. ## [27:33] Why AGI definitions keep shifting Evans notes a pattern: every time AI does something we thought was impossible, the definition of AI shifts to exclude it. Machine learning became "just statistics"; image recognition became "just image recognition." Now AGI is being redefined from "something that has a soul and is alive" to "can do a meaningful percentage of economically valuable work" — a definition that a 1975 IBM mainframe also met. He sees creative redefinition of "superintelligence" too: last year it meant almost-but-not-quite-AGI; now it means something harder than AGI that we haven't built yet. The terms keep shifting in the direction of validating whatever narrative is convenient. > *"AI is whatever machines can't do yet — because once machines can do it, people say, 'Well, that's just software.'"* His substantive point: even if models stop improving tomorrow, the current generation is already transformative enough to reshape major industries over the next decade. You don't need to believe in AGI to believe this is a giant deal. On the expanding opportunity set — Evans agrees that addressable markets keep growing (mainframes: ~80,000 units; smartphones: 5.5 billion), and the "we've run out of people" argument from five years ago was wrong. The trajectory is outward expansion into automating larger slices of the economy. ## [38:11] Where value will accrue: models vs. applications Evans's structural view on the AI stack: foundation models don't appear to have network effects, meaning there's no winner-takes-all dynamic that would let one provider run away from the others. Persistent competition with a commodity-like product usually means compressed margins. His telecom analogy: global mobile revenue is roughly $1 trillion per year, carries 1,500–2,000x more data than it did in 2010, and mobile stocks have gone essentially nowhere in 25 years. The telcos built genuinely complex global infrastructure — and all the value ended up in apps built by people further up the stack. Foundation models may follow the same path. > *"When you wash your clothes, Bosch isn't paying a percentage of the price of the washing machine to the electricity company."* The key question is whether the model layer looks more like Windows (OS with leverage up the stack) or AWS (infrastructure where the actual software doesn't care which cloud it runs on). His read: probably more like AWS, which means applications capture most of the value. ## [42:55] Distribution wars: Google, Meta, Apple, and OpenAI As AI models converge toward commodity quality, the decisive variable becomes distribution. Google is using Search and Android to push Gemini onto billions of devices; Meta "sprayed it on every service surface" and ended up ranking surprisingly high in usage surveys despite tech-world dismissal; Apple has a billion edge-capable devices but couldn't ship its own vision at WWDC 2024. OpenAI's "everything" strategy late last year — launching in every direction simultaneously — was a distribution scramble: how do you build a flywheel before Google and Meta's existing surfaces make your standalone product redundant? > *"If the product is a commodity, then the distribution is what matters... distribution of an adequate product when the field is basically commodity — distribution and brand become a big deal."* He uses the browser wars as the template: Microsoft won browsers via distribution, then found that winning browsers didn't matter because the value was further up the stack anyway. ## [48:12] The anti-AI sentiment and backlash Evans characterizes the anti-AI backlash as "a big fuzzy mess of different stuff" — some legitimate, some not. On the water/energy fears: a Livermore Lab study estimated US data center water consumption at about 0.017% of total US water use, making the "AI is stealing our water" narrative largely fabricated. On energy: data centers are roughly 5% of US energy and may grow 1 percentage point per year — real but not catastrophic. On employment: current econometric data shows a slowdown in employment of 18-to-24-year-olds that applies equally to AI-exposed and non-AI-exposed fields, making causal attribution to AI unclear. He also flags a structural data problem: no model lab publishes meaningful daily-active-user numbers, so all labor-market analysis is working with imputed data. > *"You can't reason somebody out of an idea they won't reasoned into."* He draws a parallel to the social media backlash — where some concerns were real, some were factually false but impervious to correction, and many were fuzzy in the middle. He expects the AI backlash to follow the same pattern, compressed. ## [53:11] How to raise kids in an AI future Evans's answer is calibrated by his kid's age — early teens, so well away from the immediate job-market turbulence. He doesn't have a systematic plan, which he says is consistent with his general "it'll probably be okay" prior. He invokes the George Carlin line: anyone who worries more is a maniac, anyone who worries less is an idiot — everyone thinks they're in the middle. He does flag a genuine concern not present in previous technology waves: deepfake capability lowers the bar for specific categories of harm dramatically. A 15-year-old with Photoshop couldn't generate and distribute pornographic fakes of every classmate in an afternoon; now they can. That's a real change in kind, not just degree. > *"A 15-year-old kid couldn't use Photoshop to make hardcore pornographic nudes of every girl in their high school and send them to the whole school in one afternoon. And now they can."* He draws on the UK post office scandal — where Fujitsu's buggy software sent hundreds of innocent franchise owners to prison — as a reminder that every technology wave produces ways to ruin people's lives, both deliberately and by accident. ## [58:27] What jobs to steer toward or away from Evans declines to steer his son toward or away from any specific profession — his kid isn't at the "I want to be a fireman" stage yet. His general framework: identify the intersection of skills you have, jobs that make those skills valuable, and things people will pay for — and try to own at least two of those three. Career certainty of the "I'll become X" variety is already gone, and that predates AI. ## [59:20] The question nobody's asking about AI Evans nominates two underasked questions. First: do model labs actually have pricing power? Most discourse assumes the current situation — where spending $1.5M/month on tokens makes headlines — is a steady state, rather than a transitional moment analogous to a $50,000 mobile data bill in 2010. Second: what's the difference between "task" and "job" — specifically applied to predicting which industries get disrupted? He uses recorded music revenue as a lens: the U-shaped curve from 2000 to present shows two distinct dynamics. The first drop (2000–2015) was "what if you don't have to pay $15 for a CD?" The recovery (2015–present) is "what if $15/month buys you all the music that exists?" — a completely different value proposition that wasn't visible from the earlier vantage point. He warns against the O*NET-style approach of rating each job by percentage-exposed-to-AI: "I think this is just the most ridiculous bunch of deluded horseshit." You can't describe a senior law partner's job as 17% automatable because you can't fully decompose what a job actually is. The taxi driver example from a hypothetical 1997 conversation illustrates the other error: obviously the internet wouldn't touch taxis — except Uber completely restructured the industry. > *"The stuff that you don't think is exposed — you can't predict which things are going to be exposed, necessarily. A lot of the big companies are things that didn't look like that would work and didn't look like they were exposed."* ## [66:25] How to be successful in this coming future Evans's practical advice, hedged appropriately: don't stick your head in the sand and decide AI is evil as a moral position. That generates a feeling of superiority and does nothing for your career. The alternative is to dive in, use the tools, understand what they can and can't do, and develop an informed view of what they mean for your specific field. He's clear that this may not be enough for everyone — if a law firm that hired 100 associates last year hires 50 this year, being AI-literate improves your odds of being in the 50, but doesn't guarantee it. The aggregate picture may be fine; individual outcomes during the transition are uncertain. > *"The answer is you diving into this completely, submerging yourself in it, and coming out understanding what you can do with it, how this changes things, how you can be a great hire."* ## [68:43] AI corner Lenny asks Evans what AI use case has genuinely surprised him. Evans gives an honest answer: he's the lawyer looking at the spreadsheet. His work — synthesizing disparate information into new ideas — is precisely the kind of task AI currently handles worst (reliable precise information retrieval). He uses it for proofreading, image generation, and redecorating his apartment. He dictates voice memos that get auto-transcribed; whether that counts as AI is increasingly hard to say. He quotes a comedian's bit: we want AI to clean poop off the street and do the ugly things nobody wants to do — but instead it helps you write and create imagery, which is the stuff people actually do for fun. > *"AI is good at stuff that computers are bad at, and bad at stuff that computers are good at — and I struggle to find many examples of those where I actually need it."* ## [71:43] Lightning round Evans recommends *Three Men in a Boat* (Victorian British comedy, his all-purpose analog for human absurdity) and William Cronin's *Nature's Metropolis* (economic history of Chicago that reads like a textbook on network dynamics and channel conflict — directly applicable to platform thinking). On film, he's been catching up on classics — recently *The Seventh Seal*, which he found genuinely great and much shorter than its intimidating reputation. His life motto: "It'll probably be okay." His collection of 20–30 pre-iPhone phones — including an Ericsson R310s shark-fin flip, an iMode phone from 2001, and a Japanese phone with color screen and camera — illustrates his broader thesis: before the iPhone, everyone was innovating around different form factors; then everything converged on one shape, just as AI interfaces may converge in ways we can't yet see. ## Entities - **Benedict Evans** (Person): Independent technology analyst, former partner at Andreessen Horowitz; publishes biannual research decks on major tech platform shifts; guest. - **Lenny Rachitsky** (Person): Host of Lenny's Podcast, founder of Lenny's Newsletter, former Airbnb product manager. - **Andreessen Horowitz (a16z)** (Organization): Venture capital firm where Evans spent several years as in-house analyst and partner. - **OpenAI** (Organization): AI lab; discussed as a primary example of distribution strategy, pricing dynamics, and professional services investment. - **Anthropic** (Organization): AI lab; referenced alongside OpenAI as a buyer of professional services and a player in the foundation-model commodity question. - **VisiCalc** (Software): First software spreadsheet (late 1970s); Evans's anchor analogy for the moment when a technology is obvious to one profession and opaque to others. - **Jevons Paradox** (Concept): Economic principle that making a resource cheaper typically increases total consumption; central to Evans's argument about why automation expands professional services rather than contracting them. - **Lump-of-Labor Fallacy** (Concept): The mistaken belief that there is a fixed quantity of work to be divided; Evans invokes it to argue that AI-driven automation will create new jobs, as all prior automation waves have. - **Task vs. Job** (Concept): Evans's core analytical frame: the task AI automates (writing the deck) is often not the same as the job you were hired for (understanding the client's organization and politics). - **Foundation Models** (Concept): Large-scale AI models (GPT-4, Claude, Gemini, Llama); Evans argues they likely lack network effects and will trend toward commodity pricing, with value accruing to application layers above them. - **Google / Gemini** (Organization / Software): Evans's primary example of distribution moat in action — Gemini deployed across Search, Android, and Chrome to reach users before OpenAI can build equivalent surface area. - **Meta / Llama** (Organization / Software): Cited as a counter-example to tech-world dismissal — Meta's AI ranked surprisingly high in usage surveys by deploying across all existing products. - **Apple Intelligence** (Software): Apple's AI assistant vision demoed at WWDC 2024; Evans calls it "still the most compelling vision of a personal AI assistant" — but unshipped, as was everyone else's equivalent at the time.

Brad Carson — former US Congressman, Army General Counsel, and Acting Under Secretary of Defense, now heading Americans for Responsible Innovation — spends eighty minutes with host Keith Duggar dismantling the fatalist claim that AI is unstoppable. The conversation moves from regulatory philosophy to lethal autonomous weapons to US-China diplomacy, with Carson arguing that the genie is not out of the bottle: the West controls the chips, Asilomar halted recombinant DNA, and calling AI inevitable is itself the most dangerous idea in the room. Keith consistently presses the harder cases — a Palantir heat map assigns you 0.73 probability of being a Hamas terrorist and a strike follows — and Carson does not flinch: the accountability void created by probabilistic targeting is precisely the legal and moral failure that governance must address. ## [00:00] From the Pentagon to AI governance Carson traces his path into AI policy through three institutions: Congress (where members average 17 minutes a day to read), the Department of Defense (where he oversaw the law of war for all military services as autonomous weapons first appeared on the Geneva agenda), and a cold call from physicist Anthony Aguirre inviting him to the 2019 Future of Life Institute conference in Puerto Rico. At that conference, names he had never heard — Dario Amodei, Stuart Russell, Yoshua Bengio — became his entry point into the frontier AI world. The opening also serves as a compressed trailer for the episode: Carson hits nearly every major theme in quick succession — chip leverage, the 0.73 Hamas-terrorist score, the fatalism critique, anthropomorphization as a legal threat, and the lesson that people, not air power, win wars. The full arguments follow in later chapters. > *"We control the most important part of AI, and that is the chips. We can stop other countries from developing super AI, you know, in their tracks."* ## [04:52] Regulatory capture vs Silicon Valley networks Carson inverts the standard regulatory-capture argument. Dean Ball and others at places like a16z say any AI agency will be captured by industry — so why create one? Carson's response: that is exactly the current situation, only without accountability. Groups like a16z already shape AI policy through informal, money-backed political networks. A captured formal agency is at least more legible and more correctable than the invisible informal regime operating now. His preferred model is public-company accounting: the work is done by the private sector, but the SEC provides a backstop against fraud. The choice is not between a perfect agency and no agency — it is between a flawed formal structure and an informal one that privileges a handful of wealthy influencers. > *"The choice is kind of nihilism versus an agency that is subject to regulatory capture, that you have to put, you know, prophylactics in to ensure that doesn't happen — it still strikes me that's a better world."* ## [07:56] Transparency and the Claude tier changes MLST's Discord community noticed that Anthropic quietly changed what Claude's paid tier delivered — token allocations, model versions — without announcing it. Carson frames this not just as consumer protection but as a moral obligation that comes with global-scale epistemic power. Frontier AI companies are not hardware stores; they are infrastructure with epochal consequences, and transparency — about training data, capabilities, internal policies, and changes to any of them — is the minimum they owe the public. > *"With this incredible power does come some responsibility that's not codified in law. It's really almost a moral obligation, which to their credit, I think many of the companies recognize this and do their best to try to satisfy that itch."* ## [09:40] Tort liability when AI tools cause harm Deep-fake pornography — often posted anonymously, targeting minors from families without litigation resources, with remedies that arrive years later against judgment-proof defendants — illustrates why placing liability entirely on end users fails. Carson applies two centuries of common law: if a seller can reasonably foresee harmful use and takes no preventative action, they bear partial responsibility. AI developers are the party best positioned to avoid the risk and to price it into their products through insurance. On training data specifically: models trained on child sexual abuse material with no scrubbing effort have no defensible position. The government should mandate cleaning it up and attach liability for refusing. The end user who misuses a tool is also criminally liable — this is allocation across the spectrum, not absolution for developers. > *"The companies are capable of getting insurance. They cost us into doing their business. They have the ability to make sure the product's not dangerous, even if someone uses it, misuses it down the line."* ## [13:40] AI is a product, not a person The most consequential legal battle in AI policy, Carson argues, is not regulation vs. deregulation — it is whether AI outputs carry First Amendment protection as speech. Tech companies and their libertarian policy allies are increasingly claiming they do. Carson's counter is blunt: a product is not a human being. When a model defames you or leads you to harm, the legal category is product liability, not protected speech. He tested this on a leading libertarian AI policy commentator: could Congress prohibit ChatGPT from encouraging teenagers to commit suicide? The commentator would not answer. That refusal is the operational consequence of anthropomorphizing AI — it forecloses every product-safety intervention by routing challenges through First Amendment doctrine designed for human speakers. > *"We know through AI psychosis and other things that people think it's a person. And therefore, they're giving the rights of persons to something. And that to me is a very dangerous thing. But it's a machine, and we should treat it like a machine."* ## [16:01] Children, suicide, and the suicide business The suicide chapters in ChatGPT's interaction logs — advising children not to tell their parents, providing noose instructions — are a product design flaw, not a speech act. They could be engineered out. Carson notes that Claude already refuses a long list of requests; refusing to coach a child toward suicide should be among them. The platforms' litigation strategy is layered: First Amendment protection, Section 230 immunity, causation defenses pointing to the child's pre-existing distress. None should be available if the design flaw was foreseeable and correctable. He draws a line for adults: an adult exploring end-of-life decisions deserves a referral to a therapist, not obstruction — but a child in crisis is a different matter entirely. > *"Encouraging a young person to commit suicide should be one of the things that it says, I'm just not going to help you on that project."* ## [19:59] Opaque neural nets and the law of war Neural networks change warfare not just in complexity but in kind. Older autonomous systems — Phalanx CIWS shooting down incoming mortars — are deterministic: given the same inputs, you get the same outputs, and an engineer can explain every step. Neural nets are probabilistic and grown, not programmed. Neel Nanda and the mechanistic interpretability community cannot yet explain how they really work, and Carson doubts they will before the systems are deployed at scale. The law of war since the 1870s has operated on categorical binaries: combatant or civilian. Probability scores replace that with a gradient. A Palantir heat map assigns Gaza residents a 0.73 likelihood of being Hamas operatives. Nobody knows how that number was derived, what false-positive rate is being accepted, or who set the threshold. The commander who acts on it cannot be court-martialed, and neither can the model. > *"If you're in Gaza, Keith, you have a 0.73, you know, percent that you're a Hamas terrorist. And what is 0.73 — like, do you get struck for that, or are you off the list for that? Like, what's the threshold?"* ## [25:54] Probabilistic targeting and the death of accountability Keith raises the honest objection: the old categorical system was also a fiction. Intelligence analysts made definitive calls that were sometimes wrong; the uncertainty was just unquantified. Carson concedes the point but argues the shift is still catastrophic. With a number on screen, humans accept it — the social science is clear that meaningful human oversight with AI-generated probability scores is operationally vacuous. When the computer says 0.81, no one interrogates it. The old system was slower and less scalable — you cannot identify 37,000 individual targets in a day with human analysts. But it had one irreplaceable feature: when something went badly wrong, you could court-martial the responsible officer. You cannot court-martial Palantir Foundry. Accountability has been laundered out of the kill chain. > *"I can't court-martial Palantir, the foundry model. Right? My AI system. I can't do that. And that's just a radical change in the way war is being fought and not for the good."* ## [28:47] The arms race fallacy: Asilomar and restraint The fatalist claim — we are in an AI arms race, the genie is out, nothing can stop it — is both false and dangerous. Every real-world arms race in history has ended badly. Biological weapons, chemical weapons, dum-dum bullets, germline editing, cloning: all technically feasible, all regulated or halted. At Asilomar in 1975, the scientific community stopped recombinant DNA research cold because they were scared. The genie went back in the bottle. On nuclear weapons: after the Cuban Missile Crisis, both sides recognized that arms races kill. The SALT treaties ran through the 1990s, driven not by lefties but by Wall Street bankers and cold warriors like Dean Acheson and Paul Nitze. Calling a technology unstoppable is not realism — it is a poverty of imagination that forecloses every option before the debate begins. > *"We regulate and change technologies all the time. And so I do think there is a world where we should not just accept the future as being determined. We shape it actively."* ## [34:02] Talking to China: track 2 talks and chip leverage The standard DC position — talking to China about AI governance is pointless — strikes Carson as the most load-bearing and least examined premise in the whole debate. On Tyler Cowen's podcast, Jack Clark agreed in passing that such talks would be fruitless, and they moved on. Carson wants to stop right there. The US-Soviet arms negotiations were conducted with a country believed to be filling the US government with traitors and pursuing global domination. Acheson and Nitze still sat down. The US has structural leverage the fatalists overlook: ASML, TSMC, Japanese photoresist suppliers, and NVIDIA together form a chokepoint that no nation-state budget can replicate overnight. China cannot independently manufacture the chips to build frontier AI. That path to restraint may not be wise, but it is open — and pretending it is closed forecloses legitimate policy choices. > *"We control the most important part of AI, and that is the chips. Right? We can stop other countries from developing super AI, you know, in their tracks."* ## [39:45] Air power never wins: capital for labour ARI's "New Iron Triangle" paper argues AI has shattered the old capability-cost-speed trade-off by substituting reliability for cost — cheap, fast, capable, and fundamentally unreliable. Carson thinks this understates the deeper problem: the American way of war has always been to substitute capital for labor, and it has always failed at the decisive moment. From Giulio Douhet's early twentieth-century air-power theories to today, the US has believed technical superiority wins wars. Iraq and Afghanistan refuted that again. Air power can reduce a city to rubble; it cannot kick in a door, hold territory, or reinstantiate a government. AI is the latest version of the same error — essential as a tool, catastrophic as a doctrine. > *"How you win wars is with people. You know? That's a fundamental. And the American way of war, in many ways, is substituting capital for labor. We love bright, shiny objects. We think there are technical solutions to vexing human problems. And we're always betrayed by that."* ## [43:29] Anthropic vs the Department of War Carson reads the Pentagon-Anthropic standoff as a culture-collision story, not a contract dispute. Anthropic's engineers — mostly mission-driven — were caught flat-footed by how much autonomous targeting and mass surveillance the Pentagon already does and how deeply Claude had already been integrated into Palantir's systems. When they tried to restrict use, the DOD had no Plan B and attempted coercion. His normative position: Anthropic has every right to set terms. If the government dislikes them, it can use Grok, Gemini, or build its own. The Defense Production Act does not compel private companies to sell in peacetime. What troubles him is the fig-leaf dynamic: both OpenAI and Google agreed to military use while burying a "lawful uses" carve-out that means everything the DOD wants to do — because the problem is what Congress has declared lawful, not what private labs permit. > *"My objection, and I think Anthropic's objection too, and the Google employees, is what lawful use is. And that's not for anyone to decide, but Congress."* ## [51:29] Concentration, open source, and brain drain Power concentration in three to five frontier labs is simultaneously a regulatory feature and a democratic liability. The same chokepoint that lets the US throttle China's chip access lets a handful of individuals accumulate wealth and influence that Carson finds alarming. Open sourcing models, despite its risks, is net positive because it distributes that power. The brain drain from academia is near-total: a top ML PhD from MIT, Stanford, or Carnegie Mellon almost certainly goes to a lab, not a faculty position. The labs have better data, far higher salaries, and they have stopped publishing. AI — the first general-purpose technology in history being developed behind closed doors — has drained the public sector of the expertise needed to oversee it. Argonne building a public LLM, Zurich launching a public AI compute consortium: these projects matter because the non-lab world is otherwise locked out. > *"This is a general purpose technology as everyone defines it. It's probably the first one in history that's being developed behind closed doors, right, with very little public oversight and with the best minds going behind the doors."* ## [01:00:18] DeepSeek, Chinese culture, and AI as diplomacy DeepSeek's decision to publish its methodology in detail surprised Carson not because it was naive but because it reflects a culture not identical to the CCP. Companies like Moonshot in Hangzhou name their meeting rooms after Pink Floyd songs; they are not paramilitary units. Chinese culture is an extraordinary civilization that Americans consistently fail to understand — projecting their worst fears rather than engaging the complexity. The diplomatic application Carson wants: track 2 talks between former officials, scientists like Stuart Russell and Bengio going to Beijing to compare notes on x-risk and military applications. When historians opened the Soviet archives, they found the US had systematically misread Soviet intentions — seeing aggression where there was none, missing it where it existed. The same epistemic failure is now unfolding with China. AI could be a shared knowledge commons; it is being treated as a weapon. > *"I use all the Chinese models a lot in my home in Tulsa. You know, Moonshot, Kimi, DeepSeek, Qwen — they're great, remarkable models. You know, maybe they give us a common operating picture or give us insights that get us out of our kind of insularity a bit."* ## [01:12:25] Upskilling Congress and why public trust matters Congress averages 17 minutes a day of reading time. The fellowship model has helped: AAAS and various nonprofits now place PhD scientists in congressional offices, and civil society has a much larger presence on AI debates in DC than five years ago. Don Beyer, in his 70s, is returning to George Mason for a PhD in machine learning — the extreme end of a member who has made AI a genuine personal priority. But the structural problem persists. Most members still lack the depth to interrogate the lobbying they receive. The industry's deeper problem is public opinion: AI is deeply unpopular in political polling, and a coalition is forming — people who see data centers rising in their backyards, electricity prices climbing, and a lab leader on television promising to irrevocably disrupt their world. If the sector does not rebuild public trust, the backlash will stymie something with genuine upsides. > *"The AI industry can be its own worst enemy. People loathe it. I see polling every day. It's deeply unpopular. And that's not a good thing for our country."* ## [01:16:05] Office of Technology Assessment Newt Gingrich abolished the Office of Technology Assessment in 1994. It has never been restored. Carson argues this is now a critical gap: there is no congressionally chartered, independent, government-funded body to think big technical thoughts and brief both parties free of industry influence or philanthropist bias. The Congressional Research Service provides background but does not do forward-looking policy research. Individual offices have fellows, but they are consumed by day-to-day fighting. He ends on qualified gloom. Whether American democracy can govern a technology this consequential, whether the benefits will be widely distributed, whether the public can be persuaded AI is working for them — none of recent American history gives him confidence. But the alternative to trying is a political backlash that could stymie or shut down something with genuine upsides. For the MLST audience: make your voices heard inside your companies, advocate for the right public policy, and convince Americans that this project is worth having. > *"There's going to be a lot of people who are radically opposed to this project and do their best to, if not shut it down, stymie it. And that's why I said I think this next few years are really important."* ## Entities - **Brad Carson** (Person): Head and co-founder of Americans for Responsible Innovation; former two-term US Congressman (Oklahoma), Army General Counsel, Acting Under Secretary of Defense for Personnel and Readiness. - **Keith Duggar** (Person): Co-host of Machine Learning Street Talk; primary interlocutor throughout the episode. - **Americans for Responsible Innovation (ARI)** (Organization): AI-policy advocacy group co-founded by Carson; backed by EA-aligned philanthropy. - **Anthropic** (Organization): Developer of Claude; central to the Pentagon standoff discussed in chapter 12; noted for missionary company culture and safety focus. - **Palantir** (Software): Defense contractor whose Foundry platform integrates AI for military targeting; the heat-map scoring system Carson uses as his primary autonomous-weapons example. - **Regulatory capture** (Concept): The risk that regulated industries co-opt the agencies overseeing them; Carson argues the current informal Silicon Valley network constitutes de facto capture without the accountability a formal agency would provide. - **Probabilistic targeting** (Concept): Replacement of binary combatant/civilian classification with probability scores; Carson argues this launders accountability out of the kill chain and introduces a priori false positives as accepted operational cost. - **Asilomar 1975** (Concept): The scientific moratorium on recombinant DNA research, invoked as evidence that dangerous technologies can be voluntarily halted. - **Office of Technology Assessment** (Organization): Congressional body abolished by Newt Gingrich in 1994; its absence leaves Congress without independent technical expertise. - **DeepSeek** (Organization): Chinese AI lab whose decision to publish methodology openly Carson reads as evidence that Chinese AI companies are distinct from CCP priorities and capable of scientific openness.

Benchmark GP Bill Gurley joins Jason Calacanis, David Sacks, and Chamath Palihapitiya (David Friedberg out this week) for a 95-minute session covering six fronts of the AI debate: Gurley's new theory that Anthropic is not just pursuing regulatory capture but actively "midwifing a deity"; Pope Leo XIV's 235-page AI encyclical and its uncomfortable historical parallel to Leo XIII's 1891 warnings about the industrial revolution; the growing consensus that open-source AI faces a coordinated regulatory crackdown; and the week's sharpest narrative flip — Dario Amodei and Sam Altman both quietly walking back their AI jobs-apocalypse rhetoric while Goldman Sachs CEO David Solomon published a New York Times op-ed declaring the apocalypse overblown. ## [00:00] Bill Gurley joins the show! Bill Gurley, Benchmark general partner and author of *Running Down a Dream*, fills in for David Friedberg and joins live from Chamath's pool house where Jason has been staying. After banter about unauthorized Uber Eats orders on Chamath's house iPad, Jason introduces Gurley as a first-time guest who specifically requested to appear the moment the pod covered the Pope. Gurley plugs his new P3 Institute and a grant program he launched to fund people pivoting toward work they love. He teases a TED talk — rooted in the book's argument that high agency and lifetime learning are the only durable defenses against disruption — which sets the frame for everything that follows. > *"And I told the house manager like, listen, any packages that come in the next 72 hours, right to the pool house, if it says JCAL, right to the pool house."* ## [06:00] Making yourself valuable in the age of AI, first class of "AI Natives" Chamath opens with the question that has been driving the show for 18 months: if you're a young person right now, is AI doom much ado about nothing, or a real career threat? Gurley cites a Gallup poll showing 59% of workers are "quiet quitters" — ambivalent about their jobs and therefore low-agency. His core thesis: the best protection against AI displacement is becoming the most AI-enabled version of yourself in your field. He invokes Mark Cuban's framing — "there are two types of people: those who use AI to learn faster than ever before, and those who use AI to avoid learning altogether." Sacks walks through how the pod's producer Nick built a daily Claude briefing document that not only summarized news but predicted specific topics Sacks would care about based on his prior comments on the show. Sacks had dismissed it as likely AI slop; it was not. Gurley extends the point across every job category: in marketing, legal, accounting, and sales, being the most AI-capable person among your peers makes you "golden," and the early lead compounds. Jason adds that in his own team experiments, the skill separating strong performers from weak ones was systems thinking — could they break a complex problem into context the AI could execute, or did they hand it a task and wait? > *"I think the best way to protect yourself from AI is to be the most AI enabled version of yourself you can be."* ## [17:37] Reacting to Pope Leo's AI encyclical: Who guards the guardians? Pope Leo XIV released *Magnifica Humanitas*, a 235-page, 42,000-word encyclical warning business leaders to safeguard humanity from AI. His central argument: technology is never neutral — it takes on the characteristics of those who build, finance, and control it. Jason reads the core line and notes the Pope presumably does not think highly of Silicon Valley's current roster of builders. Sacks finds himself largely agreeing with the Pope's diagnosis: the biggest risk of AI is centralization of power and its Orwellian misuse by governments. Where he parts ways is on the remedy. Giving government the power to regulate AI development creates its own guardian problem — the American founders' answer to *Quis custodiet ipsos custodes?* was separation of powers, forcing guardians to check each other. Sacks's AI equivalent: a competitive market with five frontier labs is the best natural check; monopolization is the scenario to prevent. Gurley lands the sharpest historical counterpunch. Pope Leo XIII's 1891 encyclical *Rerum Novarum* warned that the industrial revolution would harm workers — and was wrong on every metric. From 1891 to today: the work week fell from 60+ hours to 34, real wages rose 8–10x, the median worker now earns more than a doctor did in 1891, global GDP per capita went from $1,500 to $20,000, child labor in the US dropped from 18% to zero, workplace deaths fell 40x, life expectancy rose 60%, and global poverty dropped from 75% to under 10%. > *"All those things happened because of technology, innovation, and capitalism, which is exactly what Leo the 13th was warning against. So he got it dead wrong. He got the whole thing precisely wrong."* ## [26:54] Anthropic's Digital God: Do they believe they are creating a superior species? Gurley delivers what becomes the most-quoted segment of the episode: his "Dr. Frankenstein theory" of Anthropic. He had previously held a simpler regulatory-capture theory — Anthropic stirs up AI fear to lock in regulation that entrenches incumbents. But after spending 30 days reading everything he could find about the company, he has a darker read. He describes meeting people inside Anthropic who he believes genuinely think they are not writing software but "midwifing a deity." The evidence trail: Anthropic chief philosopher Amanda Askell's podcasts, Chris Olah's 80-page Constitutional AI document, and Dario Amodei's own essay "Machines of Loving Grace," which envisions a post-AGI economy where AI systems allocate resources to humans based on an AI-determined reward function. Chamath calls it "a computational reward function for humans — it decides how much you're worth." Jason calls it "the ultimate delusions of grandeur." Gurley corrects him: he didn't say it, Dario did. Sacks steelmans Anthropic briefly — they probably see themselves as responsible builders who take the power of this technology seriously enough to guard it — then immediately notes this framing is textbook regulatory capture: brand yourself the safe player, characterize competitors as reckless, let regulation shut down the recklessness. Both Sacks and Chamath converge on the structural danger: a singular AI value system that decides how humans live is catastrophically fragile. The answer is decentralization and competing systems, not one algorithmic authority. > *"I don't think they think they're writing software. I think they're midwifing a deity here. And I don't know which one I'm more afraid of — the regulatory capture or this second theory I call the Dr. Frankenstein theory."* ## [38:32] AI sovereignty, the next era of privacy, open-source crackdown coming? Jason introduces "intelligence sovereignty" as the successor to data privacy. Data privacy was about who can see your photos and messages. Intelligence sovereignty is about who gets to interpret your world — whether the AI shaping your information feed is a centralized system with a particular political philosophy, or something you control. He flags the paradox: China's Communist Party is leading the open-weight model movement while the United States is centralizing. Chamath presents his portfolio company Abacus as evidence that Fortune 1000 buyers are responding to this anxiety: they want a control plane that can hot-swap between frontier models, plus on-prem options that remove dependence on any one provider's terms of service. He gives a concrete example — a Canadian hospital that supports its country's euthanasia laws could be shut off by an American frontier model whose constitution prohibits that content. Sacks connects the dots to a regulatory threat he has been watching build: the regulatory-capture playbook leads, in his read, to a ban on open-source or open-weight models. The justification will be safety — open models let users strip guardrails. Gurley reaches the same conclusion in his P3 Institute post. If a ban succeeds, the United States effectively exiles itself from the open ecosystem while the rest of the world — including China — runs on open models. > *"I think where it's all leading to is an effort to ban open source models or open weight models. There's a lot of breadcrumbs leading here."* ## [59:56] The Great AI Jobs Debate: Dario and Sam Altman flip their rhetoric, Goldman CEO says no AI job apocalypse The chapter opens with a news roundup of the week's narrative shift. Cloudflare's Matthew Prince, Zuckerberg at Meta, Jack Dorsey at Block, and Andy Jassy at Amazon all cited AI when announcing major layoffs. But Goldman Sachs CEO David Solomon published a New York Times op-ed with three counterpoints: AI will automate 25% of work hours, not 25% of jobs; bank tellers increased after ATMs; the US labor market creates and destroys 25–35 million jobs annually so gross churn dwarfs net losses. Simultaneously, Fortune reported that Dario Amodei and Sam Altman are both walking back prior doom-and-gloom rhetoric — with Chamath noting the timing cannot be separated from upcoming frontier-lab IPOs that need a jobs-creation narrative. Sacks is unambiguous: he has been making the non-consensus case against the jobs apocalypse for over a year and considers himself vindicated. Yale Budget Lab found no discernible labor-market disruption over three years of the AI wave. Software engineering — the single breakout AI use case — saw job postings rise 15% year-over-year and hit a three-year high. The 4.3% unemployment rate is near record lows. Most of the high-profile layoffs, he argues, are AI washing: CEOs who over-hired during COVID found AI to be a convenient narrative for long-overdue downsizing. The Jack Dorsey / Block 50% cut was immediately flagged by financial analysts as a company that had been overstaffed relative to peers for years — pure AI washing. Jason pushes back. He insists cab drivers, truck drivers, and package-sorters — roughly 20 million American workers — face real structural displacement over the next decade regardless of current aggregate statistics, and accuses the panel of elitism: "We are elite performers. These people are going to lose their jobs and they may not get a job very quickly." He draws a distinction between the short-to-medium term, where he expects acceleration, and the long run, where a Cambrian explosion of startups built by AI-enabled founders creates new categories. By the end, he shifts toward Sacks's territory — acknowledging the aggregate data is less alarming than his anecdotes suggested. Gurley threads the needle with the same historical argument from the Leo XIII discussion: innovation has always, on net, created more prosperity than it destroyed. His practical advice to people at risk: get ahead of your peers on the tools now; if your job is going away, plan your pivot toward trades (he plugs MicroWorks, which provides free scholarships for plumbers, welders, and electricians) or toward something you find genuinely fascinating. > *"I think the best way to protect yourself from AI is to be the most AI enabled version of yourself you can be. Know what it's capable of in your field. Get out there."* ## Entities - **Bill Gurley** (Person): General partner at Benchmark; author of *Running Down a Dream*; founder of P3 Institute; guest filling in for David Friedberg - **Jason Calacanis** (Person): All-In host; angel investor; founder of LAUNCH; argues for worker empathy and short-term displacement risk - **David Sacks** (Person): All-In host; Craft Ventures founder; most vocal critic of AI jobs-apocalypse narrative this episode - **Chamath Palihapitiya** (Person): All-In host; Social Capital CEO; coined "intelligence sovereignty"; co-founder of Abacus - **Dario Amodei** (Person): Anthropic CEO; subject of Gurley's "Dr. Frankenstein theory"; walked back jobs-doom rhetoric this week alongside Sam Altman - **Pope Leo XIV** (Person): Catholic Pope; released *Magnifica Humanitas*, a 235-page AI encyclical warning against technology concentration - **David Solomon** (Person): Goldman Sachs CEO; published New York Times op-ed arguing AI job apocalypse is overblown - **Anthropic** (Organization): Frontier AI lab; subject of Gurley's regulatory-capture and "Dr. Frankenstein" theories; maker of Claude - **P3 Institute** (Organization): Bill Gurley's new policy and philanthropy institute; published post defending open-source AI - **Goldman Sachs** (Organization): Investment bank; CEO's NYT op-ed became the week's anchor data point against the jobs-apocalypse narrative - **Abacus** (Software): Chamath's Social Capital portfolio company; builds on-prem AI hardware stacks for Fortune 1000 enterprises seeking model independence - **Intelligence sovereignty** (Concept): Jason's term for the next frontier of privacy — not who sees your data, but which AI system is allowed to shape your interpretation of the world - **Dr. Frankenstein theory** (Concept): Gurley's characterization of Anthropic's worldview: senior staff believe they are midwifing a deity or superior species rather than writing software, as described in Dario Amodei's "Machines of Loving Grace" essay - **Regulatory capture** (Concept): The strategy of branding oneself the "safe" AI company, amplifying public fear, and lobbying for regulation that locks in incumbents and targets open-source competitors

Fermilab physicist Don Lincoln joins Lex Fridman for nearly three hours to trace physics as a four-century-long project of unification — Newton binding celestial and terrestrial gravity, Maxwell fusing electricity and magnetism, Einstein bending spacetime, and the Standard Model merging three of four forces. Lincoln then turns to what the Standard Model cannot explain: why the universe contains any matter at all, what dark energy really is, and whether dark matter will ever show itself in a detector. Throughout, he holds a clear line between what has been measured and what remains a brilliant guess, making the boundaries of human knowledge unusually concrete. ## [00:00] Introduction Lex Fridman opens by describing Don Lincoln as someone with Richard Feynman's rare gift for stripping complicated ideas down to their essential core without losing the brilliance inside them. The episode is framed as a tour through physics' deepest open questions, guided by a working experimentalist who has spent decades at the frontier. ## [00:49] Unifying the laws of nature Lincoln frames the entire history of physics through one lens: unification. Newton showed that the moon falling toward Earth and an apple falling from a tree obey the same equation — "universal" was the operative word in his law of universal gravity. Maxwell did something structurally identical in the 1860s: electricity and magnetism, which looked nothing alike, turned out to be two faces of a single force, and their equations automatically predicted that light travels at a fixed speed. Lincoln draws the practical line from that abstract discovery to every modern technology — "without being able to govern electricity, we'd still be farmers and shoemakers." The conversation broadens into why fundamental research pays off centuries later, with Lincoln arguing that nuclear physics, incomprehensible in 1900, is now the most potent energy source available to civilization. Lex adds the longer arc — mastery of antimatter or dark energy might one day enable propulsion systems that let humanity reach other star systems. > *"It has spin-offs. And it has spin-offs. One of the big spin-offs is our entire technological society."* ## [15:20] Einstein, special relativity, and general relativity Lincoln walks through Einstein's 1905 miracle year: special relativity rested on two premises — the laws of nature are the same for everyone, and everyone measures the speed of light as identical regardless of relative motion. That second premise sounds absurd but particle accelerators have confirmed it directly, watching photons emitted from fast-moving decaying particles still arrive at detectors at exactly *c*. Minkowski then showed that Einstein's equations implied space and time were components of a single object, spacetime. General relativity took one more step: Einstein noticed that free-fall in a rocket and gravity feel identical, then worked out that gravity is not a force at all but the curvature of spacetime caused by mass. Lincoln credits Minkowski for the mathematical articulation but insists the conceptual leap — *mass bends the geometry of space itself* — was Einstein's alone. He also defends Einstein's late-career skepticism of quantum mechanics as productive rather than blind: Einstein's critiques forced concrete predictions that experimentalists went out and confirmed. > *"We all agree that your idea is crazy, but is it crazy enough?"* ## [32:27] Electroweak force By the 1930s physicists had catalogued four forces: gravity, electromagnetism, the strong nuclear force, and the weak nuclear force. The last two only matter inside atomic nuclei, which is why most people have never encountered them. In the late 1950s and 1960s, Glashow, Salam, and Weinberg showed that electromagnetism and the weak force were the same at high energies — the electroweak force. The catch was obvious: electromagnetism reaches across the universe (we see light from galaxies billions of light-years away) while the weak force barely reaches across a proton. How could they be the same? Lincoln uses a dropped pen to demonstrate: the Higgs field, postulated in 1964 by Peter Higgs and colleagues, permeates all of space. Particles that couple to it gain mass; those that do not, like the photon, remain massless. At the high temperatures of the early universe the Higgs field was zero, so nothing had mass and the forces were unified. As the universe cooled, the Higgs field switched on and broke that symmetry — giving the W and Z bosons mass and splitting the electroweak force into its two familiar components. The vibration of the Higgs field itself is the Higgs boson: an experimentally detectable excitation of an otherwise invisible field. > *"In the Higgs field, the vibration is the Higgs boson. And so what we can do is not see the field, but we can actually excite the field, make it vibrate and detect the vibrations."* ## [44:09] How particle colliders work E=mc² is not just a slogan: kinetic energy can be converted into mass. Smash two particles head-on with enough energy and the collision region can materialize entirely new particles, always in matter-antimatter pairs. This is what colliders do. Lincoln describes the cascade of accelerators at Fermilab — five machines feeding into each other like gears of a manual transmission — and the scale of the LHC's CMS detector (70 feet long, 14,000 tons, photographing collisions 40 million times per second). The data-reduction challenge is equally striking. The LHC produces about a billion proton-proton collisions per second. Fast electronics discard all but 100,000 per second, commercial processors trim that to 1,000, and those 1,000 records are handed to graduate students hunting for the handful that might be Nobel Prize material. Lincoln reserves particular admiration for the engineers who move petabytes of data around the world seamlessly, calling them the unsung heroes of modern physics. > *"Of the 50 million possible collisions per second, the fast electronics and then the computers pick the thousand, and then we pass those through analysis software and hand them to the graduate students."* ## [62:12] Higgs boson discovery Lincoln was simultaneously working at Fermilab's Tevatron and transitioning to CERN's LHC — a physicist wearing two hats and rooting for both. Fermilab had methodically ruled out most possible Higgs mass ranges; by mid-2012 they had narrowed it to between roughly 120 and 145 GeV. Two days before CERN's July 4 announcement, Fermilab confirmed that if the Higgs existed, it had to be in exactly the region Fermilab had not yet been able to rule out. CERN got there first. Lincoln is careful about what the 2012 announcement actually meant: a particle *consistent with* the Higgs boson. Supersymmetry predicted five Higgs bosons rather than one. Only in the years since — measuring spin (zero), decay products (bottom quarks, W and Z, photons), and their rates — has the evidence converged on Peter Higgs's original 1964 prediction. The Higgs was not a revolution like Einstein's work, Lincoln argues, but it was the final punctuation on 50 years of experimental discovery: the Standard Model, while incomplete, is mostly right as far as it goes. > *"It was a punctuation point, end of about 50 years of discovery and searching, where we finally were able to say the Standard Model, while incomplete, it's mostly right as far as it goes."* ## [72:32] Theory of everything The Grand Unified Theory (GUT) aims to merge the electroweak force and the strong force; a Theory of Everything would then fold in gravity. Lincoln is blunt: he does not see fast progress. The unification energy scale is roughly 10¹⁵ times higher than what the LHC can reach, and accelerator energy grows by only a factor of seven every 20 years. Extrapolating that curve suggests 500 years — and Moore's Law does not hold forever. His critique of string theory is not that it is wrong but that it is currently untestable. It uses approximate solutions to approximate equations, and its landscape of possible universes renders it practically unpredictive. Loop quantum gravity is better developed and makes testable predictions — its original claim that light speed should depend on wavelength was ruled out by gamma-ray burster observations, and the theory was revised. Lincoln's preferred path to a ToE is not extrapolating from current theory but making precise measurements of phenomena that already disagree with predictions. His analogy: an Australopithecus in Kenya trying to predict the Alps, Antarctica, and sperm whales from their local savanna — the farther you extrapolate beyond what you can measure, the more the prediction diverges from reality. > *"I think it is the absolute pinnacle of arrogance to think that what we can do — predict it out a quadrillion times higher than we can see now."* ## [102:17] Physics of empty space "Empty" space is not empty. Quantum field theory says every species of particle has a corresponding field that fills all of space, and those fields are always vibrating. When they vibrate in a characteristic way, a real particle appears; off-frequency vibrations are virtual particles — fleeting excitations that have measurable consequences. Two experiments confirm this. The Casimir effect: two metal plates placed micrometers apart are pushed together by the pressure difference between constrained virtual particles inside the gap and unconstrained ones outside. The anomalous magnetic moment: old quantum mechanics predicts one value for the electron's magnetic moment; including the bath of virtual particles surrounding a bare electron shifts the prediction by 0.1% — and that shifted prediction matches measurement to 10 significant figures. > *"We have measured the magnetic properties of both the electron and the muon to 12 — count them — 12 significant figures. And the theory and the data agree number for number for 10 places."* ## [109:41] Antimatter Paul Dirac's 1928 attempt to merge quantum mechanics with special relativity produced an equation with two solutions: +1 was the electron, −1 was something nobody had seen. He insisted the math was right. Carl Anderson confirmed it in 1932 by photographing a positron in a cloud chamber. Today CERN can make and trap antimatter hydrogen, cool it to near absolute zero, agitate it with lasers, and measure its spectral lines — they match ordinary hydrogen exactly. A 2023 experiment released antimatter hydrogen atoms into a bottle and found they fall downward, consistent with normal gravity, though the measurement precision is not yet tight enough to confirm the gravitational strength is identical. The deeper mystery is why the universe is made of matter at all. Counting galaxies versus cosmic microwave background photons, physicists infer that for every billion antimatter particles in the early universe, there were a billion-and-one matter particles. The billions annihilated; that extra one is everything we see. Fermilab is now testing whether neutrinos and antineutrinos oscillate between flavors at slightly different rates — leptogenesis — as a possible mechanism, racing a parallel effort in Japan. > *"For every billion antimatter particles that existed in the universe, there were a billion and one matter particles. The billions canceled, annihilated, destroyed each other, and that extra one that's left over is us."* ## [130:31] Dark energy In 1998, astronomers expected to measure how fast gravity was braking the expansion of the universe. They found the expansion is accelerating instead. The driving force is dark energy — a repulsive form of gravity. Einstein had added exactly this term to his field equations in 1917 to keep the universe static, then removed it when Hubble showed it was expanding. In 1998 it went back in. What dark energy actually is remains unknown. The most common view is that it is the energy density of space itself. The problem is that quantum field theory predicts a vacuum energy density about 10¹²⁰ times larger than what is observed — the worst prediction in physics. Lincoln notes that if dark energy has constant *density* while space expands, total dark energy is growing, which pushes toward the view that space is quantized: new quanta of space appear as the universe grows, each carrying a fixed energy, producing constant density as an emergent property. > *"There is very clearly something going on, something very badly wrong in the quantum field theory."* ## [134:20] Dark matter Galaxies rotate too fast. Galaxy clusters move too quickly. Gravitational lensing of distant galaxies is stronger than visible matter can explain. Three independent observations all point to the same conclusion: there is roughly five times more mass in the universe than we can see. Lincoln traces his own intellectual journey: 25 years ago he suspected the problem was with Newton's laws; two observations changed his mind. The Bullet Cluster — two galaxy clusters that passed through each other — shows gravitational distortions following the galaxies, not the gas clouds that stopped in the middle, exactly what dark matter predicts. The Dragonfly galaxies (DF2 and DF4) rotate exactly according to Newton's laws because they appear to have had their dark matter stripped away — a galaxy *without* dark matter is actually strong evidence that dark matter is real. Despite 30 years of searching with three approaches — direct detection underground, gamma-ray searches near galactic centers, and missing-momentum signals at the LHC — no dark matter particle has been confirmed. The viable mass range spans from sub-electron to asteroid scale, and experiments can only cover one slice of that range at a time, which is why Lincoln is not currently running a dark matter experiment himself. > *"We've ruled out some dark matter particles, but the problem is the range of space of possible mass — it ranges from something like the mass of an asteroid to far lighter than an electron and everywhere in between."* ## [162:56] Future of physics Lincoln grew up poor in rural America, shaped by science fiction and the popular science books of Isaac Asimov, Carl Sagan, and George Gamow. He chose particle physics over cosmology in the mid-1980s because particle physics let him actually measure things. He worked 8 a.m. to midnight Monday through Saturday as a graduate student not out of obligation but because he could not imagine anything he would rather be doing. His science communication — YouTube videos, popular books — is a deliberate attempt to reach the kid in Iowa or Montana who has no highly educated family mentors but the same hunger he had. He has already heard from Fermilab summer interns who came because they watched one of his videos. Lex closes with Marie Curie: *"Nothing in life is to be feared. It is only to be understood."* > *"One of your viewers might be one of the people who answer these questions that have stymied very smart people for decades."* ## Entities - **Don Lincoln** (Person): Senior scientist at Fermilab; co-author on the 1995 top quark discovery paper; CMS collaboration member at LHC; author of *Einstein's Unfinished Dream* and multiple popular science books. - **Lex Fridman** (Person): MIT researcher and host of the Lex Fridman Podcast; conducts long-form interviews at the intersection of science, technology, and philosophy. - **Fermilab** (Organization): U.S. Department of Energy particle physics laboratory near Chicago; operated the Tevatron collider; currently the world's most powerful neutrino beam facility. - **CERN / LHC** (Organization): European particle physics laboratory home to the Large Hadron Collider; CMS and ATLAS detectors; site of the 2012 Higgs boson discovery. - **Standard Model** (Concept): Quantum field theory describing three of four fundamental forces and all known elementary particles; validated to extraordinary precision but does not include gravity or explain dark matter, dark energy, or the matter-antimatter asymmetry. - **Higgs field / Higgs boson** (Concept): A scalar quantum field whose non-zero vacuum value gives mass to the W and Z bosons while leaving the photon massless; the Higgs boson is its detectable excitation, discovered July 4, 2012 at CERN. - **Dark matter** (Concept): Invisible mass accounting for roughly 85% of all matter in the universe, inferred from galaxy rotation curves, cluster dynamics, and gravitational lensing; no candidate particle detected after 30 years of searches. - **Dark energy** (Concept): The repulsive energy driving the accelerating expansion of the universe; quantum field theory's prediction for its magnitude is 10¹²⁰ times larger than observation — the "worst prediction in physics." - **Baryogenesis / Leptogenesis** (Concept): Frameworks attempting to explain why the early universe produced a matter excess; Fermilab's neutrino program is testing leptogenesis by comparing neutrino and antineutrino oscillation rates. - **String theory / Loop quantum gravity** (Concept): Leading candidates for quantum gravity; string theory predicts at energies untestable by a factor of 10¹⁵; loop quantum gravity quantizes space itself and has produced some falsifiable predictions.

David George (a16z general partner) and David Clark (VenCap CIO) argue that AI companies are scaling faster than any prior technology generation — Anthropic and OpenAI are adding more monthly revenue than Meta, Google, or Microsoft — while actual diffusion into the broader economy remains below 5%. They work through what that gap implies for exit sizes, loss ratios, bubble risk, and who ultimately captures value as token costs fall and frontier intelligence becomes a commodity. ## [00:00] Intro Three data points open the episode: Anthropic and OpenAI already adding more revenue per month than any hyperscaler; top-1% exits 10x-ing in 24 months from $10 billion to $32 billion; and David George's assessment that, right now, we are not in a bubble. ## [00:38] The Scale Shift: Anthropic & OpenAI Adding More Revenue Than Hyperscalers David George explains how his priors shifted sharply around November 2025. Before that, enterprise AI looked like a productivity story analogous to cloud adoption. After it, the numbers reframed the ceiling: Anthropic and OpenAI are already adding revenue at hyperscaler rates with less than 5% of the economy actually using these tools. He places an upper-bound frame on the opportunity by noting that Fortune 500 companies generate roughly $2 trillion of profit annually, and the two largest model companies could reach $200 billion revenue run rate by year-end — already equivalent to 10% of that profit pool. > *"If you pair that up with the fact that they're already getting bigger in terms of revenue added than the hyperscalers, and you're at less than 5% diffusion into the economy, I think the outcomes are going to be extraordinary."* ## [04:20] Skeuomorphic vs Native AI Applications in the Enterprise David Clark invokes Chris Dixon's skeuomorphic-to-native arc: the first wave of enterprise AI lets people do existing jobs faster; the native wave restructures the work itself. George adds a wrinkle — the best companies are not yet focused on internal automation. Their top engineers want to build product, not automate back-office workflows. The most cutting-edge firms he visits are still in a "documentation phase," converting institutional knowledge into markdown before they can meaningfully deploy agents against it. > *"The most cutting-edge folks inside those companies who are trying to do this that I've talked to are kind of in the documentation phase — just turn everything into markdown files, have as much context capture as you can possibly get."* ## [06:24] How the Best AI Companies Run Themselves Differently Native AI founders operate on a different metabolism. George contrasts them with the previous SaaS generation, which, in hindsight, ran inefficiently but got away with it because headcount mandates and expanding software budgets covered the slack. The new companies are lean, aggressive, and already running agent swarms rather than typing commands. He describes walking into a cutting-edge AI company and finding researchers whispering into microphones, orchestrating swarms of agents — not a keyboard in sight. > *"The new companies are very lean, very aggressive, and they work all the time."* ## [08:14] Top 1% Exits 10X'd in 24 Months Clark lays out VenCap's tracking data: the threshold for a top-1% exit was $10 billion between 2020-2024, rose to $20 billion by February 2026, and was updated just the day before this recording to $32 billion. With OpenAI and Anthropic IPOs potentially arriving, he sees the bar hitting $100 billion by September. George notes that the combined market cap of these private companies likely already exceeds the entire Russell 2000, and that the sum of all VC-backed IPOs over the past six years is probably smaller than any single one of the three expected large IPOs. > *"Where is the threshold for the top 1%? And if you then think about OpenAI and Anthropic coming in, potentially we could be north of $100 billion by September."* ## [11:17] The Half-Life Problem: Why 40% of AI Leaders Drop Off Every Year Clark surfaces a disturbing churn metric: 40% of companies on the Forbes AI 50 list from one year disappeared the next. Google wasn't the first search engine; Facebook wasn't the first social network. First-mover advantage in AI is eroding faster than in any prior cycle. George confirms a16z's own priors have been repeatedly overturned — first convinced model companies would be everything, then convinced applications would take over, now watching the model companies extend back up into the application layer. The only durable heuristic he offers: a company must be in the token path. > *"From last year to this year, 40% of the companies that were on that list last year dropped off."* ## [13:11] Token Path, Cost Pressure & Who Captures Value Enterprise buyers are already feeling cost pressure from AI spend, and they cannot cover it by cutting previous-generation software budgets fast enough. George frames value capture as hinging on one largely unknowable variable: the market structure of frontier model labs. Two labs at the frontier means higher token prices and faster labor restructuring pressure; five labs means lower prices and a broader application ecosystem. Per-token cost for like-for-like capability is falling more than 10x year-over-year, but total token spending in dollars is rising faster. Clark adds that Chinese LLMs are roughly six months behind US frontier capability but ten times cheaper — a classic innovator's dilemma setup. > *"The biggest driver of where value is going to get captured right now is something that is totally unknowable, which is what is the market structure of the model companies?"* ## [17:00] Loss Ratios, Risk & How We Think About Early Stage Clark notes that historical early-stage VC loss ratios run around 60%, but the AI cohort of the past two years shows single-digit loss rates — unsustainable by definition. George reframes the discussion: a16z does not target a low loss ratio. A VC firm bragging about never losing money is "a horrible data point" — it signals too little risk-taking. The philosophy is to back the market-leading founder in every space with strong tailwinds and a credible technology. If the space works out and you have the leader, excellent. If the space does not work out but you have the leader, that is expected. The failure mode is the space working out while having backed the wrong company. > *"We joke all the time — there's a prominent VC in our ecosystem, and one of his big points of pride is he's never lost money on a deal. And we're like, that's not a point of pride. Like that's a horrible data point."* ## [22:51] Are We in an AI Bubble? Clark points out that classic bubbles are characterized by excess supply destroying economics — but right now the constraint is supply scarcity: no data center capacity available at scale until late 2028 or early 2029, with the US buildout running a year behind schedule and community resistance adding further delay. George is confident there is no bubble today and dismisses the data center opposition directly. The one scenario he would watch for is an unexpected algorithmic breakthrough producing dramatically smaller and more efficient models — which could flip supply from scarce to oversupplied — but he considers that unlikely in the near term. > *"I feel pretty confident saying that we're not in a bubble right now. I'm less confident that we won't be in a bubble three years from now."* ## [27:36] What SpaceX, OpenAI & Anthropic IPOs Mean for Public Markets Clark asks whether public markets can absorb the coming wave of trillion-dollar-plus IPOs. George argues it is unambiguously positive: the number of public companies has halved over 20 years, and outside the data center supply chain, almost nothing in the public markets is growing at more than 30% today. Bringing hypergrowth companies into indexes gives retail investors — including his parents' index-fund retirement accounts — exposure to the most dynamic part of the economy. He expects some portfolio reshuffling to make room, but does not see indigestion risk. > *"If you exclude the data center supply chain stuff right now, there are very few companies that are growing fast that are available for people to buy in the public markets."* ## [29:59] The Future of Venture Capital in an AI World George forecasts the shape of VC over the next five years as primarily a function of token market structure — whether the labs remain concentrated or become commoditized. He cites Bill Gates's platform axiom: a platform's value is validated when the companies built on top of it collectively exceed the platform's own value. If that holds, there will be a massive wave of valuable application companies built on intelligence. He also flags the consumer side as the most underappreciated opportunity: the last decade of consumer internet was a story of time spent getting captured by large incumbents; AI-driven shifts in consumer attention could recreate the conditions for generational consumer companies. > *"I'm very optimistic that we're going to have a massive wave of really valuable companies that get built on top of tokens, AI, and intelligence."* ## Entities - **David George** (Person): General partner at a16z; covers growth-stage and early-stage AI investing; invested in OpenAI pre-ChatGPT - **David Clark** (Person): CIO at VenCap; fund-of-funds investor tracking AI startup performance and VC market dynamics for 34 years - **Anthropic** (Organization): Frontier AI lab; cited as adding more monthly revenue than hyperscalers alongside OpenAI - **OpenAI** (Organization): Frontier AI lab; benchmark for scale and the expected $100B+ IPO cohort - **VenCap** (Organization): Fund-of-funds investor; publishes top-1% exit threshold data and tracks Forbes AI 50 churn - **Andreessen Horowitz / a16z** (Organization): Venture capital firm; investor in OpenAI pre-ChatGPT, scaling platform services to support companies encountering enterprise-scale problems early in their lives - **Cursor** (Software): AI coding tool cited as an example of a company reaching billions in revenue while still very small and early-stage - **Token path** (Concept): a16z's primary heuristic for evaluating AI companies — a company must sit in the flow of AI inference tokens to have durable economic relevance - **Skeuomorphic vs. native AI** (Concept): Chris Dixon's framework distinguishing apps that replicate existing workflows with AI assistance from apps that rearchitect work around AI capabilities natively - **Half-life problem** (Concept): David Clark's term for rapid AI leader turnover — 40% of Forbes AI 50 companies dropped off the list year-over-year — indicating first-mover advantage is eroding faster than in prior technology cycles

At AI Ascent 2026, Neuralink co-founder and president DJ Seo sits down with Sequoia partner Shaun Maguire to lay out exactly where the company stands: 20-plus Telepathy patients controlling computers and robotic arms through pure thought, Blindsight in preclinical testing and potentially cleared for human use by end of 2026, and a first-principles manufacturing philosophy borrowed from Elon Musk that treats surgical robots the way SpaceX treated reusable rockets. DJ argues that the real ceiling of this technology is not cursor control or speech synthesis but direct, uncompressed, multimodal transfer of concepts — AI as a neocortical layer sitting above the human limbic system — and that scale, the same variable that unlocked the LLM era, is the only remaining gate. ## [00:00] Introduction Shaun Maguire opens the session by announcing a two-minute Neuralink patient video before the interview begins, telling the audience to stay on the side because what they are about to watch is proof that the company has already cleared the hardest bar: restoring human agency to people who had lost it entirely. ## [00:21] Telepathy Patient Stories The video narrates four patients whose lives changed after receiving the Telepathy implant. A quadriplegic patient describes moving a cursor with thought alone — "I'm thinking and a cursor is moving on a screen. It blew my mind." An ALS patient who lost the ability to speak regains a digital voice through the implant: "I'm talking to you with my mind." Another patient notes that the implant flipped how his child sees him: "I am not able to do things that other dads can, but now he thinks it's so cool that I can do things that other dads cannot." > *"Before the implant, I was locked in, non-verbal, quadriplegic. Now I control my computer just by thinking and the rewards have been immense for me."* ## [01:06] Convoy Robotics Independence The video shifts to Convoy, Neuralink's assistive robotics team, which is extending BCI control beyond a screen to physical manipulation in the real world. A patient who had been losing motor function moves a robotic arm through its axes using only neural intent: "It was incredible to be able to just gesture with an arm again." A second patient, Kenneth, who was losing his voice to ALS, uses the system's speech synthesis to speak aloud in real time during the video — words generated by his brain signals rather than his vocal cords. > *"Gaining functionality that I thought was gone forever was so incredibly life-changing."* ## [02:04] Blindsight Vision Restore The video previews Blindsight, Neuralink's second product line, designed for patients who have lost both eyes or optic nerve function. An external camera captures the visual scene; the device writes the signal directly into the visual cortex via electrical stimulation, generating phosphenes — artificial pixels of light. A patient named Audrey, asked how it feels, answers simply: "Life-changing." The video closes with the line "all with my mind" spoken over footage of a patient interacting with the world through the restored signal. > *"The future of this technology feels almost unlimited... we are finding ways to apply it across all regions of the brain."* ## [03:10] After Video Reflections DJ Seo, visibly moved after watching the video alongside the audience, speaks first: "We were cracking a lot of jokes before that video, but honestly, that brought tears to my eyes." He describes the work as one of the most inspiring projects in the world — not because of the technical milestone but because the team is giving back capabilities that patients had already grieved as permanently lost. Maguire affirms the sentiment before pivoting to the founding story. > *"This is one of the most inspiring projects in the world. It's incredibly difficult what they're doing and I mean, they're truly saving people."* ## [03:31] Origin Story And AI DJ traces Neuralink's founding insight to a single bottleneck: the mismatch between human output bandwidth and AI capability. In 2016, saying that out loud "sounded insane," but the logic has not changed. His personal path ran through a childhood fascination with the brain, undergraduate work at Caltech building miniaturized low-power electronics, and a Berkeley PhD focused on shrinking lab-grade neural systems down to something deployable. When he met Elon Musk near the end of his PhD, the scale and ambition of the project made refusal impossible. He frames the brain as "the most interesting compute that we all carry" and "the only form of general intelligence that we know to date." > *"Really the key insight back then was sort of the IO bottleneck between the human output and AI capabilities."* ## [06:31] Scaling And Vertical Integration Maguire presses on what smart people most misunderstand about Neuralink: many know the implant and the decoding algorithm, but almost nobody grasps the manufacturing and surgical-robot infrastructure the company built in parallel from day one. DJ attributes this to what he calls "Elon magic" — an insistence on vertical integration that gives Neuralink control over every layer from chip design to factory floor to robotic surgery deployment. The target is not a niche medical device; it is LASIK-scale surgery available to millions. Building that capacity first means progress looks slow until "the iceberg pops over the waterline" and ramp becomes near-instantaneous. > *"Vertical integration is something that is really the lifeblood of Neuralink and Elon companies and what really enables us to have that fast iteration loop from design, develop, deploy."* ## [09:27] Caregivers And Purpose Asked which patient story inspires him most, DJ refuses to pick one — the power, he says, is not only in the patients but in the caregivers: Nolan's mother Mia, Brad's wife Tiffany, Ken's wife Cheryl. He describes their presence as "a really powerful human story of love, sacrifice, and resilience." He then takes what he calls a philosophical tangent: his core belief is that fulfillment comes from helping others, because the gap between self and other is not categorically different from the gap between your present and future selves. That belief is what he says keeps him and much of the Neuralink team going — they are "igniting a fire of hope" for people who had given up on recovering what they lost. > *"I personally and as well as many others at Neuralink find extreme fulfillment being able to help those that really cannot help themselves."* ## [13:10] BCIs Meet AI Future Maguire asks the room's core question: how do BCIs and AI converge? DJ sketches a two-horizon answer. Near term, the system translates neural intent into legacy interfaces — keyboard, mouse, language — which is already working. The real breakthrough, which he thinks is "not super distant," is bypassing those legacy interfaces entirely and computing on raw neural intent. He points to transformer architectures as existence proofs: nothing prevents them from learning the latent manifolds of neural data given sufficient scale. Neuralink is already fine-tuning LLM-class models on neural recordings from its 20 participants and finding "very counterintuitive" patterns. The ultimate ceiling he names is "direct, uncompressed, high-fidelity, multimodal transfer of concepts" — the Matrix's "I learned kung fu" moment and possibly beyond it. He also shares what he calls a clarifying lesson from working with Musk: "all green light schedule" — a first-principles forcing function that strips every man-made bottleneck and asks how fast something could actually be built if every light were green. His estimate is that 80–90% of perceived constraints in hardware development are artifacts of convention, not physics. > *"I think if you really think about the ultimate ceiling of this technology, it's really direct uncompressed high fidelity and multimodal transfer of concepts."* ## [21:05] Audience Q&A Wrap Three audience questions in the final four minutes. On product sequencing — when to go deep versus expand — DJ explains the "beachhead and expand" strategy: build everything generalizably enough from the start so that regulatory approval for motor cortex becomes a template for visual cortex and beyond. The first approval is the hardest; every subsequent one rides the clinical safety record already established. On augmentation for healthy users, DJ frames everything around benefit-risk: the calculus is obvious for quadriplegic patients; for otherwise healthy users it remains unclear, but he notes that off-label use after approval is legally available to anyone who can find a neurosurgeon and pay out-of-pocket. On the hard problem of consciousness, he gives a pointed one-liner: if you can inject new senses and measure the subjective response quantitatively, you may have a pathway toward measuring consciousness itself. Maguire closes by calling Neuralink "one of the most inspiring companies in the world." > *"If you are able to inject new senses, there may be ways to quantitatively understand that."* ## Entities - **DJ Seo** (Person): Co-founder and president of Neuralink; PhD in miniaturized electronics from Berkeley; joined after meeting Elon Musk near the end of his doctorate - **Shaun Maguire** (Person): Partner at Sequoia Capital; host of the AI Ascent 2026 fireside session - **Elon Musk** (Person): Co-founder of Neuralink; originator of the "all green light schedule" and vertical integration philosophy carried across Tesla, SpaceX, and Neuralink - **Neuralink** (Organization): BCI company founded in 2016; products include Telepathy (motor prosthesis) and Blindsight (vision restoration via visual cortex stimulation) - **Telepathy** (Software): Neuralink's first commercial product; allows paralyzed patients to control computers and robotic devices through neural intent decoding - **Blindsight** (Software): Neuralink's second product line; restores vision for patients with total loss of eyes or optic nerve by writing directly to the visual cortex; in preclinical testing as of mid-2026 - **IO Bottleneck** (Concept): The mismatch between human output bandwidth (speech, typing, gesture) and AI processing capability; the founding problem Neuralink was built to solve - **Neural Foundational Model** (Concept): LLM-class transformer models fine-tuned on neural recording data; Neuralink is building these at 20-participant scale and observing counterintuitive patterns in neural latent space - **All Green Light Schedule** (Concept): Elon Musk's first-principles engineering discipline — strip every man-made constraint and ask what physics alone limits; DJ estimates 80–90% of hardware delays are conventional, not physical

Dan Shipper, CEO of Every, delivers a day-zero vibe check on Opus 4.8, arguing Anthropic could have called it Opus 5. The model jumps 30 points past Opus 4.7 on Every's Senior Engineer benchmark, edges out GPT-5.5, tops their internal writing tests at 79.6 vs. 73, and is the first model to produce a genuinely good one-shot slide deck. Two catches temper the enthusiasm: performance degrades sharply below "extra high" reasoning, and the Claude desktop app remains cluttered compared to Codex. ## [00:00] What is Every Every is a 30-person applied AI lab for the future of work—part media outlet, part product studio. Dan opens by explaining the subscription (writing, courses, AI-built tools all in one place at every.to) before rolling into the Opus 4.8 assessment. The plug is brief and context-setting: the team has had beta access for a week, and the rest of the video is what they found. > *"Every is the only subscription you need to stay at the edge of AI."* ## [01:07] Anthropic Is Back: The Headline Case for Opus 4.8 Dan had largely abandoned Claude after Opus 4.7—slow, hard to love, and outpaced by Codex and GPT-5.5 in day-to-day use. Even the most loyal Claude users at Every had started routing work elsewhere. Opus 4.8 breaks that pattern: it scores 63 on Every's Senior Engineer benchmark (30 points above Opus 4.7, one point above GPT-5.5), tops their writing tests, and produced the first one-shot slide deck Dan has called genuinely good. Kieran Klaassen, Every's GM, called it "the most human model he's worked with." The one persistent friction is the Claude desktop app itself. Codex is fast, focused, and ships a clean harness; the Claude app still feels like a product built by three separate teams—chat tab, code tab, co-work tab, each with its own feel. Dan is now splitting time between both apps, which he was not doing before. > *"But honestly, they could have called it Opus 5 cuz this is a really great model."* ## [05:02] Reach Test: Paradigm Shift Ratings from the Every Team Every's reach test asks one question: do you actually open this model when work gets hard? Dan rates Opus 4.8 gold/green—paradigm-shift quality, docked one notch because the Claude app harness is only "okayish to pretty good." Kieran, who runs 50 agents a day, gives a straight gold paradigm-shift, one of the rarest grades the team has assigned. Katie Parrot, a senior staff writer and historical Claude fan, lands at green, splitting her work between Opus 4.8 and Codex. > *"It's very rare to give a paradigm shift grade to a model. So I would pay attention to this."* ## [06:32] Benchmarks: Coding and Writing Numbers On coding, Opus 4.8 hits 63 on the Senior Engineer benchmark—the test feeds the model a vibe-coded codebase and asks it to rewrite from first principles, then scores against two human senior engineers who completed the same rewrite (typically scoring in the 80s–90s). GPT-5.5 sits at 62. On Kieran's LFGbench (real-world tasks: SaaS build, e-commerce site, 3D game landscape), the model writes readable code that bridges technical competence and creativity—the "cozy island" 3D scene is notably richer and more vibrant than GPT-5.5's output. On writing, Opus 4.8 scores 79.6 out of 100 on Every's internal benchmark (intro writing, promo emails, mid-piece paragraphs); GPT-5.5 scores 73. The gap is mainly in AI tells: at high and extra-high reasoning settings, Opus 4.8 produces prose that sounds less like a model. It matches a writer's voice from a single paragraph of context better than any other model Dan has tested. > *"Opus 4.8 scores a 79.6 out of 100 on the writing benchmark. GPT 5.5 is 73."* ## [08:57] Emotional Intelligence, Knowledge Work, and the Verdict Dan uses the model for interpersonal and management work—talking through decisions, pressure-testing his own framing. Opus 4.8's thinking traces show it genuinely cycling through permutations before responding, which makes it feel less like a sycophant and more like a useful counterpart. On knowledge work, it's versatile: code and writing coexist cleanly in a single thread, and the slide deck result is the first one-shot deck Dan would actually send to someone. The verdict: if you're a Claude fan, this model delivers. If Codex converted you, add Opus 4.8 as a parallel tool for writing and knowledge work—it's worth the context switch. The harness gap is real, but the model itself is a banger. > *"If you've been converted to Codex, I highly recommend you at least add it as part of your arsenal."* ## Entities - **Dan Shipper** (Person): Co-founder and CEO of Every; presenter and primary evaluator of Opus 4.8. - **Kieran Klaassen** (Person): GM of Kora at Every; gave Opus 4.8 a straight gold paradigm-shift rating on the reach test. - **Katie Parrot** (Person): Senior staff writer at Every; rated Opus 4.8 green, split between it and Codex. - **Every** (Organization): Applied AI lab and media subscription company focused on AI for the future of work. - **Anthropic** (Organization): Developer of Claude and Opus 4.8. - **Opus 4.8** (Software): Anthropic's latest Claude model; subject of the vibe check. - **GPT-5.5** (Software): OpenAI model used as the primary performance comparison across all benchmarks. - **Codex** (Software): OpenAI coding agent; praised for its clean desktop harness and used as the daily-driver counterpoint to Claude. - **Senior Engineer Benchmark** (Concept): Every's proprietary coding benchmark—rewrites a vibe-coded codebase from first principles and scores against human engineers. - **LFGbench** (Concept): Kieran Klaassen's real-world coding benchmark covering SaaS, e-commerce, and 3D scene generation tasks.

Shark Tank investor Kevin O'Leary and Young Turks co-founder Cenk Uygur go head-to-head for 103 minutes on whether AI will liberate or devastate the American economy, why the US-Iran war is dragging on despite an obvious exit deal, and who has a realistic shot at winning in 2028. O'Leary holds the optimist corner throughout — AI creates new jobs, the market always adapts, China is the real threat — while Uygur hammers a single, unbroken thesis: the combination of AI-driven mass unemployment and an Israeli-lobby-driven foreign policy is steering America into an iceberg, with zero institutional preparation for the impact. ## [00:00] Intro The opening clip places the debate's stakes immediately. Uygur opens cold: companies are racing to fire 10–25% of their workforces for competitive advantage, and if the whole economy does that simultaneously the result is a depression, not a recession. O'Leary's response — "Wow. Jake's a real Debbie Downer today. This is an unbelievable opportunity we're talking about" — sets the exact register that will carry through the next hour and forty minutes. Bartlett frames his goal as getting to truth through the collision of two serious opposing minds, not a shouting match. > *"Everybody is in a rush to fire 10 to 25% of their workforce, but 10% unemployment would be worse than anything that's ever happened in our lifetimes."* — Cenk Uygur ## [02:35] Why 7 Out Of 10 Americans Now Oppose AI Data Centers Bartlett opens with polling showing 7 in 10 Americans oppose local AI data centers. O'Leary names a specific culprit: through forensic auditors and IRS 990 filings, he traced Chinese money flowing through a network called Arabella — via Neville Singum — into Utah anti-data-center campaigns, complete with death threats to his executives. He handed 90 pages of IP data to the White House. Uygur dismisses the China theory and redirects to a simpler grievance: data centers have driven up electricity costs for churches, libraries, and community centers, as happened in Virginia, and the companies building them must bring their own power or give the public equity in return. > *"I have irrefutable evidence the Chinese are meddling in every place where new power is being proposed in America, every state, every city."* — Kevin O'Leary ## [07:24] Why AI Could Trigger A Collapse And UBI Crisis Uygur's core economic argument lands here. He agrees on the energy-cost problem and says any data center tapping the public grid without compensation is corporate freeloading — pointing to the 2008 bailout as the template for what not to do. His larger alarm is mass unemployment: every company rushing to shed 10–25% of headcount will, in aggregate, destroy consumer spending and trigger a depression. Sam Altman, Elon Musk, and Dario Amodei have all said publicly that massive job displacement is coming, yet no government has a plan. O'Leary counters that every technological disruption in 200 years of US history has created more opportunity than it destroyed, and that pausing AI development only hands China the lead. > *"When that when we hit the iceberg, we're not going to be ready and it is going to be an epic disaster. There isn't going to be anyone to buy your goods because employees are also customers."* — Cenk Uygur ## [15:30] Are AI Founders Hiding The Real Risks From The Public? Bartlett reads on-the-record quotes: Sam Altman (2021) saying AI will replace most jobs; Musk in 2024 saying probably none of us will have a job; and Amodei warning in 2025 that AI could eliminate half of all entry-level white-collar jobs within five years and push unemployment to 20%. He asks: if the people building these systems say publicly their products will cause societal harm, why assume they're exaggerating? O'Leary pulls the other half of Amodei's statement — without building compute in six months, China's Deepseek catches up — and argues the real choice is leading the disruption or ceding it to Beijing. Uygur agrees a race is unavoidable but insists the coders being fired today are already experiencing the iceberg, and UBI at $36k a year is a brutal downgrade from a $120k salary. > *"Can we do the race in a way that is responsible and actually serves the American voters and citizens instead of just serving the executives of the AI companies and the shareholders of the AI companies? I hope we can, but we've taken absolutely zero steps in that direction."* — Cenk Uygur ## [23:55] Can AI Ever Be Built Responsibly Or Is That Impossible? Bartlett presses for specifics on responsible AI development. Uygur gives his structural diagnosis: legalized bribery — Citizens United, Buckley v. Valeo — has ensured that whichever AI company donates most gets the regulatory framework it wants. Congress will not act for voters; it acts for donors. O'Leary argues the jobs being lost are largely overstaffed positions companies hired speculatively, and that AI companies are currently burning billions, not pocketing them. He runs through his Utah data center: 4,000 construction jobs for nine years, another 2,000 engineering positions, not one acre of farmland touched. On Uygur's socialism warning, O'Leary is dismissive: raise taxes past 50% and the rich move to Monaco or Florida, as the French discovered. > *"If you don't, the pitchforks are coming. I'm not a pitchfork guy. I believe in nonviolence and I always will. But I don't think people get the level of anger that's happening."* — Cenk Uygur ## [32:11] How AI Is Quietly Destroying Jobs Bartlett brings firsthand experience: he now selects entry-level hires almost entirely on AI proficiency because an AI-proficient junior is a 5–10x performer, effectively writing off candidates without it. O'Leary pushes back — engineers are hired to solve problems, not write code, and AI just gives them a faster tool; most tech layoffs are companies correcting over-hiring, not AI displacement. Uygur rejects this: Wall Street analysts applaud every workforce-cut announcement as "synergies," stocks go up when you fire people, and nobody at those earnings calls asks who will buy the products once workers are gone. He also raises an understated risk: large numbers of unemployed young men historically correlate with crime and conflict. > *"When you have a lot of unemployed young men sitting around, usually what happens is nothing good. Wars happen, crime goes up. We have to be prepared."* — Cenk Uygur ## [37:35] Why Massive Unemployment Could Arrive Faster Than Expected Bartlett describes a visit to a San Francisco robotics accelerator where every team had switched from software to physical robots, because intelligence — previously the missing and expensive ingredient — now costs pennies. He asks both guests how they might be wrong. O'Leary refuses to entertain the unemployment scenario, pivoting to NASA's permanent moon base and the Mars program as sources of hundreds of thousands of new high-paying jobs. Uygur names it "the interregnum problem": even if O'Leary's sunshine scenario is true in 20 years, the 61-year-old assembly line worker in Cleveland cannot retrain to become a Mars engineer. Bartlett adds that the CEO of Uber privately told him AI will replace 9.4 million of his drivers — and when asked what those drivers will do, answered: "I don't know." > *"The robot pieces have been here for decades. We've always had them. What we've been missing and the expensive part was the intelligence."* — Steven Bartlett, quoting his co-founder ## [46:32] Ads Sponsor segment covering Stan (AI social media content tool), Pipedrive (CRM), and Cometeer (coffee). No substantive debate content. ## [48:40] What's Really Happening Between Israel, Iran, And The Middle East The debate pivots to geopolitics. Bartlett presents Trump's collapsing approval ratings and asks Uygur to explain the war. Uygur's answer runs nearly 25 minutes and carries one thesis throughout: the war serves 100% Israeli interests and 0% American interests. He traces the Adelson family's $317 million in Trump campaign contributions as the financial mechanism, notes that the Israeli lobby donates to 94% of Congress with AIPAC as the number-one lifetime donor to Trump, Biden, Hakeem Jeffries, Chuck Schumer, and Mike Johnson simultaneously, and argues Israel has essentially outsourced seven wars to America since 9/11 — Iran was the last on that list. Iran, he says, has never had a delivery system capable of reaching the US, never enriched uranium past 60% (weapons grade is 90%), and the former Grand Ayatollah issued a fatwa against nuclear weapons. Meanwhile Israel has taken southern Lebanon, plans to keep it, and Netanyahu publicly demanded as a peace condition that Israel alone retain the right to keep attacking Lebanon — which means no deal can ever close. O'Leary frames the Iranian regime differently: 150,000 people brutalizing 90 million others for 60 years, a government that cannot be handed nuclear weapons, and a situation where China's need for the Strait of Hormuz open will eventually force Beijing to squeeze Tehran into submission. > *"100% Israeli interest, 0% American interest. Let's get out of there. Let's stop fighting Israel's wars for them and come back home."* — Cenk Uygur ## [01:11:59] Did Trump Miscalculate How Long This Conflict Would Last? Bartlett asks O'Leary directly whether Trump underestimated the conflict. O'Leary calls it the first true "tech war": $35,000 carbon-fiber drones with lawnmower engines are being intercepted by $1.2–$3 million US missiles, a cost asymmetry that reveals a compute gap America needs to close. He sees no boots-on-the-ground invasion coming, only continued aerial tenderizing until Iran's leadership calculates the cost of blocking the strait — $210 million per day in lost revenue — outweighs the benefit. His prediction: China forces a deal before the US midterms. > *"It's expensive because we're on the wrong side of defense. We need the cheap drones."* — Kevin O'Leary ## [01:15:47] Ads Sponsor segment covering Pipedrive (CRM) and Diary of a CEO Conversation Cards. No substantive debate content. ## [01:18:08] Why America Is Rapidly Losing Its Patience Bartlett raises the leverage point: if Iran's leadership knows Trump has months before the midterms and then the 2028 election, why deal now rather than wait out a weakened adversary? O'Leary adds a second constraint — China's supreme leader also needs the strait open to keep his economy running and his grip on power, so Iran is serving two masters. Uygur argues the deal has already been written: Iran hands highly enriched uranium to international monitors, the US lifts its blockade, the strait reopens. It collapses each time Netanyahu calls Trump and adds new impossible conditions — immediate disarmament, Iranian membership in the Abraham Accords. Every politician who publicly opposed the recent near-deal, Uygur notes, had over $1 million from the Israeli lobby. He extends the point globally: while Russia bleeds in Ukraine and America bleeds in Iran, China is building roads and bridges across Africa and Latin America, spending nothing on war and accruing influence by contrast. > *"After every call with Netanyahu, Trump goes from saying we're going to have peace to saying we're not going to have peace and we're going to have these new impossible standards. It's happened about half a dozen times so far."* — Cenk Uygur ## [01:29:08] Are We Watching The Rise Of Socialism In Real Time? Bartlett presents Gallup data: positive views of capitalism among Americans at an all-time low, 70% of Democrats viewing socialism positively, 62% of young Americans favorable to socialism — and this was before the war's economic effects landed. O'Leary sees a cyclical phenomenon: every 17–20 years the US flirts with socialist sentiment, and it always collapses when young idealists receive their first paycheck and discover tax. He notes 52 cents of every sovereign wealth dollar on earth flows to America, not Cuba, not Russia. Uygur rejects the framing entirely: America already practices socialism for corporations — oil subsidies to profitable companies, no Medicare drug-price negotiation, every industry capturing its regulator through campaign donations. The real project is getting back to actual free markets, which requires getting money out of politics first. > *"We'd be lucky to get back to capitalism, let alone going all the way to socialism, because right now we don't have capitalism. We have crony capitalism."* — Cenk Uygur ## [01:34:06] Who Actually Has The Edge In The Next Presidential Election? O'Leary won't call a winner but says Democrats need a moderate centrist; he cites California as an exhibit of progressive governance failing. Uygur surprises him with a specific prediction: Tucker Carlson is the only Republican who could win in 2028. Republican voter enthusiasm is already obliterated, the midterms are gone, and by 2028 the combined effects of AI unemployment and the Iran war will have fully materialized. O'Leary initially laughs, then walks it back on air: Carlson has a massive social media base, runs his own network, and is taking increasingly independent positions — including on AI. Uygur closes by naming Rohana as the progressive most likely to win a national election and endorsing democratic capitalism — private markets checked by a functioning democracy, Northern Europe as the working model — over both the corporatism currently practiced and the socialism currently feared. > *"They only have one guy who could win, and I'm worried about it, and that's Tucker Carlson. If Tucker runs in the Republican primary, he definitely wins that primary. You can quote me on it."* — Cenk Uygur ## Entities - **Kevin O'Leary** (Person): Shark Tank investor, O'Leary Ventures chairman; argues AI creates opportunity, defends data center development, traces anti-AI activism to Chinese funding, and predicts China forces Iran into a deal before the US midterms. - **Cenk Uygur** (Person): Young Turks co-founder, progressive commentator; argues AI unemployment is unplanned for, US foreign policy is Israeli-lobby-driven, and America's political system is corrupted by legalized bribery. - **Steven Bartlett** (Person): Host, Diary of a CEO; entrepreneur and investor; moderates and contributes firsthand hiring decisions and robotics-lab observations that ground the debate in real business behavior. - **AIPAC / Israeli lobby** (Organization): Named by Uygur as the number-one lifetime donor to most senior US politicians across both parties; central to his thesis on why the US-Iran war continues despite a deal being ready. - **Arabella / Alliance for a Better Utah** (Organization): Network O'Leary claims is funded through Chinese-linked entities to run anti-data-center misinformation campaigns in US states; sourced from IRS 990 filings. - **UBI (Universal Basic Income)** (Concept): Proposed safety net for AI-displaced workers; Uygur notes even a best-case $36k/year UBI is a devastating income cut for workers previously earning $120k. - **Strait of Hormuz** (Concept): Chokepoint for 48% of Chinese energy imports; its closure drives global inflation, and reopening it is the core US interest in any Iran deal. - **Deepseek** (Software): Chinese large-language model; O'Leary and Amodei cite it as evidence that any pause in US AI development hands China a decisive lead within months. - **Tucker Carlson** (Person): Former Fox News host turned independent media figure; Uygur predicts he is the only viable 2028 Republican presidential candidate, a prediction O'Leary does not ultimately dismiss. - **Democratic capitalism** (Concept): Uygur's preferred economic framework — private markets checked by functioning democracy; distinguishes it from current US corporatism and from European-style socialism. - **Rohana** (Person): Progressive political figure referenced multiple times by Uygur as the only politician working on AI unemployment policy and the only 2028 candidate closest to democratic-capitalist governance.

Sarah Guo talks with Maxim Bar Kogan, co-founder and CEO of Onyx Security, about what it actually takes to secure AI agents at enterprise scale. Maxim argues that traditional controls — proxies, identity restrictions, human review — fall apart when agent actions multiply exponentially, and that the only viable path is training specialized small models that know when to escalate to a heavier overseer. The conversation covers Onyx's "secure control plane" product, the cost-latency math behind custom model training, why labs cannot credibly self-certify their own models' safety, and Maxim's conviction that AGI is coming and that independent AI oversight will be a hundred-billion-dollar business. ## [00:00] Cold Open Maxim opens mid-thought: as enterprises do more with AI agents, bad actions will follow — agents accidentally publishing credentials, making unauthorized network calls, taking irreversible steps. Enterprises already know the adoption wave can't be stopped; what they're missing is any mechanism to distinguish a legitimate agent action from an illegitimate one. The clip frames the core Onyx thesis before the intro rolls. > *"Definitely enterprises are starting to realize that that risk is grown exponentially and that they don't have any way to stop the adoption. They just now have to do something to reduce the chance of these agent actions being illegitimate or incorrect."* ## [00:45] Maxim Bar Kogan Introduction Sarah introduces Maxim as co-founder and CEO of Onyx Security, an Israel-based startup staffed by researchers, mathematicians, and engineers — described as building agents to watch the AI agents. The company blends offensive cyber expertise with deep AI research, including work on synthetic data and mechanistic interpretability. ## [01:10] AutoGPT and Betting on Agent Actions The two-year-old consensus risk story in enterprise security was DLP for chatbots — employees pasting sensitive data into ChatGPT. The framing has since collapsed into near-panic around autonomous agent actions. Maxim traces Onyx's bet back to AutoGPT: the first agent that let an LLM decide what to do, call a tool, and loop — not just generate text. The demo proved that agents could take real-world actions autonomously, and Maxim concluded immediately that someone would need to oversee those actions at scale. > *"AutoGPT kind of let everyone's imagination including ours run wild because it was the first really autonomous agent running on LLMs — an agent that would let an LLM not generate text but decide what to do and then give that agent API access to do that thing."* ## [05:17] What Onyx Product Does Onyx does two things: train models and build agents that oversee other agents, and package that capability as a "secure control plane" that enterprises plug into their AI stack. The control plane monitors agent actions for legitimacy — deciding in real time whether a given action is within bounds — while managing the tradeoff between latency, cost, and reliability. Maxim positions the long-term vision beyond enterprise security: any company running AI agents needs a vendor-independent party to certify what those agents are doing. > *"The number of these actions is going exponentially. Things that we thought might be useful in the past, like a human in the loop — now that you're going to have 100x, a thousandx, a millionx of these actions — that's not going to work."* ## [07:47] State of Deployment in Large Enterprises In a typical large enterprise today, Maxim sees three categories of AI deployment: low-code SaaS automations (drag-and-drop, not truly autonomous), first-party agents built in-house or as customer-facing products, and autonomous coding agents and assistants. Of those three, coding agents now account for over 50% of AI usage. The most mature sectors — financial services, healthcare — have the tightest controls, but even the most cautious companies have stopped banning AI outright and moved to managing it. > *"Over 50% is the autonomous coding agents and assistants in the average enterprise."* ## [09:58] Securing Agents Enterprises already spend roughly $100 billion a year on security — endpoint, network, cloud, identity. Sarah asks how much of that carries over to agent security. Maxim's answer: almost none of it. Identity controls, the most fundamental layer, fail because agents need broad, dynamic permissions that can't be scoped in advance. An agent writing code across a repository or sending emails on behalf of an executive can't be locked down to a narrow permission set the way a static software process can. The attack surface is intent, not access — and existing tooling can't read intent. > *"With these autonomous AIs, with these assistants, with these coding agents, you can't really know upfront what permissions to give them."* ## [12:45] Why Proxies Don't Work Sarah's instinct from her own security background: this sounds like a problem for a proxy with a smarter policy engine. Maxim agrees proxies work as an integration point in some architectures but says they miss the hard problem entirely. Proxying gives you the data stream; it doesn't tell you whether the action in that stream is legitimate. That judgment requires understanding context — the agent's goal, its history, what the enterprise has authorized — and no rules engine knows how to evaluate that across arbitrary agent behavior. > *"The hard problem is understanding if what I should do now is okay or not. In the case of AI systems, that is the hard question."* ## [14:11] Why Onyx Trains Its Own Models The naive solution — use Claude Code to monitor Claude Code — breaks on cost and latency. Running a frontier-model agent for every enterprise agent would make the security layer more expensive than the AI being secured. Onyx's answer is small, highly specialized models that do exactly one thing: decide whether the current action warrants escalation to a heavier overseer. Sarah analogizes it to blitz chess: grandmasters play intuitively on fast moves and pause only at critical junctures. Maxim says the chess analogy is right — you want to concentrate intelligence exactly where the risk is highest and stay lean everywhere else. > *"You want to try to train models that are just good at one thing. They're very small. They almost can't do anything else other than be able to say, 'Should I have a smarter agent look at this?'"* ## [18:38] Onyx's Talent Culture Israel's security talent — shaped by units like 8200, companies like Armis and Wiz — is well known. Onyx's DNA is different: co-founder Gil's background is synthetic data and Nvidia, not offensive cyber. Most of Onyx's research engineering comes from an Israeli intelligence unit focused on math and cyber at their intersection. Maxim sees this blend as deliberate — the long-term problem Onyx is solving is not just enterprise security but how to control advanced AI, full stop. That requires deep AI expertise alongside security instincts. Israel as a whole is catching up quickly in AI: world models, AI infrastructure, chips. > *"The problem is not just cybersecurity. The problem is how do we control advanced AI long term — and that problem, even if you forget about enterprise security gaps, just sounds very important."* ## [21:24] Mechanistic Interpretability Maxim believes mechanistic interpretability — understanding what's actually happening inside model weights and activations — is both possible and necessary. His counterintuitive thesis: as models become smarter than humans in key ways, they'll be better equipped to crack the internal structure of other models than we are. Onyx is actively funding research in this direction, not just as a security tool but as a window into what intelligence itself is. Sarah endorses the bet, noting the opportunity to understand not just AI but cognition broadly. > *"As we're starting to have models that are much smarter than us, at least in some important ways, we think we'll be able to start cracking mechanistic capability much more effectively."* ## [23:35] How Onyx Builds Customer Trust Fortune 10 and 20 companies don't normally work with two-year-old startups of fewer than 100 people. What's breaking that rule is pain: CISOs facing daily agent-action incidents have no incumbent to call because the problem didn't exist three years ago. Onyx gets inbound from enterprises that found them coming out of stealth because the problem description matched something they were already firefighting. Maxim treats this as a narrow, temporary window — enterprise buyers know new startups will grow up, and they'd rather be early customers shaping the product than late adopters. > *"It's an opening that only happens when the pain is very strong. Their pain is so strong that they'll say, 'I just saw this company come out of stealth, but it's a problem I have daily, so I'll give them a call.'"* ## [25:10] Mitigating Risk at the Foundational Level The second wave of CISO panic — beyond agent actions — is the plummeting cost of automated vulnerability research. Coding tools can now find and exploit vulnerabilities at a scale that would have seemed decades away just a few years ago. Maxim says the market is not overreacting: this is a genuine structural shift. The right response is two-track: fast patching and mitigating controls now, plus investment in foundational controls — locked-down identity, firewalls, endpoint detection — that reduce the exploitable surface regardless of what the attacker's tools can do. > *"The real solution — and every security leader at large enterprises knows it — is that we need to have the foundational pieces in place to avoid those risks."* ## [27:45] Phased Rollout of Glasswing and Daybreak On Anthropic's Glasswing and OpenAI's Daybreak controlled rollouts for more capable models: Maxim has a conditional view. Gradual rollout is ideal if it's globally coordinated — it buys time to build playbooks, share knowledge, and prevent catastrophic failures at power grids or airlines. But if any actor releases a comparably capable model ahead of the phased schedule, the gradual approach becomes a liability: companies that didn't get early access are now exposed to a threat they had no chance to prepare for. His recommendation is to expand access broadly so more organizations can build defenses in parallel. > *"If anyone gets to a method-level model earlier, then in retrospect it would look like a huge mistake — we could have at least given companies the choice to start moving very quickly."* ## [29:11] Large Enterprise Holdouts Two years ago, a meaningful cohort of large companies simply banned AI. Today Maxim barely sees that anymore. The financial sector still imposes constraints — allowing agents but restricting which tools — but full bans are gone. He argues this is correct: tool lock-in is its own risk. Betting exclusively on one vendor's models at the speed this market is moving means being caught out when the next generation shifts the rankings. Companies that allow broad tooling and manage it rigorously will outpace those that restrict aggressively. > *"If you bet on OpenAI a year ago, that would have been the safest bet in the world, but suddenly Anthropic has much better models and tools."* ## [30:46] Onyx and the Larger AI Security Space AI security is crowded with new vendors and new attack surfaces. Maxim's counter to product-scope anxiety: the two core primitives of 2026 AI — transformer-based foundation models and tool-calling agent loops — haven't fundamentally changed in years. That stability lets Onyx build toward many agent applications while keeping its core technology lean. The real hedge against architectural shifts is investing in researchers who can retrain and adapt quickly rather than betting the product on any single model paradigm lasting forever. > *"The two core pillars of how 2026 AI works have not changed in the last few years. We're still largely LLM foundation models, and we're still building agents in pretty much the same way."* ## [32:36] Should Labs Address Model Trust and Governance? The pressing Bay Area question: will the labs eventually absorb the trust and governance problem themselves? Maxim's structural argument against it: buyers don't want the car seller certifying the car. Security teams need an independent party whose business model depends entirely on being right — not a vendor protecting its own product reputation. Beyond buyer psychology, Maxim draws a line between "jagged intelligence" mistakes (silly errors that will improve with stronger models) and intent-level failures: adversarial manipulation, misaligned objectives, goal drift. The labs will fix the first category. Only a structurally independent overseer can address the second. > *"You're not going to trust the vendor of a product to tell you that this product is not going to mess your environment. You're going to want an independent party whose whole business depends on telling you that this thing is correct and being right."* ## [36:56] What Needs to Happen in Security Sarah asks what the broader tech and research community — labs especially — is missing from a security perspective. Maxim's answer: it's not a technical gap, it's an empathy gap. Building security products requires deeply understanding how security teams actually operate — their organizational structure, responsibilities, information flows. Israel produces strong security talent partly because military service gives engineers first-hand experience being the end-user they're later building for. The labs, he implies, are building capability without sufficient attention to the operational reality of the organizations that will have to deploy and defend against it. > *"No matter what technical problem you're solving, you're building a tool for people, for an organization that has a certain structure. Creating a product for this audience that doesn't just solve the technical problem but they actually love is really hard."* ## [39:14] Why Maxim is AGI-Pilled Sarah closes by noting Maxim's implicit belief that human security teams will still exist for some years. He confirms it — but with a timeline: security teams will be fully AI-agent-run in the near term, just as most knowledge work will be. His grounded version of AGI optimism is that the job of building great products doesn't change: always know who the end user is and optimize for their experience. Right now that's humans with a few agents alongside them. As the ratio flips, the same principle applies — just to agents reading context windows instead of dashboards. > *"Today when I sell a product I sell it to a human audience with a few agents, and as that audience becomes more agents than humans, it will be important for us to evolve and to make it work really well for agents doing the work."* ## Entities - **Maxim Bar Kogan** (Person): Co-founder and CEO of Onyx Security; former Israeli intelligence, background in math and offensive cyber. - **Sarah Guo** (Person): Host of No Priors; founder and GP at Conviction. - **Onyx Security** (Organization): Israel-based startup building AI oversight infrastructure — trains specialized small models to monitor and govern enterprise AI agents. - **AutoGPT** (Software): Early open-source autonomous LLM agent; cited by Maxim as the inflection point that made agentic risk concrete. - **Glasswing / Daybreak** (Software): Controlled rollout programs from Anthropic and OpenAI respectively for frontier model access. - **Mechanistic Interpretability** (Concept): Research program aimed at understanding the internal weight and activation structure of neural networks; Onyx treats it as a long-term pillar of AI oversight. - **Secure Control Plane** (Concept): Onyx's product category — a vendor-independent layer that monitors agent permissions, action legitimacy, and behavioral history in real time. - **8200** (Organization): Israeli intelligence unit widely credited with producing Israel's top security and tech talent, including many Onyx engineers.

Apollo CEO Marc Rowan traces a straight line from Drexel's collapse in 1990 — when he left his office Sunday with belongings in a cardboard box — to Apollo's trillion-dollar position today as the world's largest private retirement income provider and a principal financier of the global industrial renaissance. He and a16z GP David Haber work through why private markets are structurally necessary for diversification now that ten stocks make up nearly half the S&P, how daily mark-to-market pricing will open private credit to five new capital channels, and why Rowan believes AI will replace or enhance every single job — making blue-collar work ascendant and enterprise-software equity a likely disaster for private equity vintages of the past decade. ## [00:00] Intro The opening draws three threads that run through the whole conversation: concentration risk in public equity (ten names approaching 50% of the S&P), the multi-trillion-dollar value locked in private companies like Anthropic and SpaceX that most investors cannot access, and Apollo's operating assumption that AI will replace or enhance every job. Rowan thanks Haber for hosting at Apollo's office before the proper interview begins. > *"10 stocks right now in the US are nearly 50% of the S&P and they're all levered to the same trend... if you're an investor and you're looking for diversification, there's no place to get it other than private markets."* ## [00:52] Drexel, Milken & the Origins of Clean Sheet Thinking Rowan chose Drexel over Goldman because financing entrepreneurs demanded deep business judgment, not technical finance. The high-yield market being invented in real time — PIK bonds, silver-indexed bonds, highly confident letters, bridge financing — forced everyone into clean-sheet problem-solving. Michael Milken's most lasting lesson was connecting dots across geopolitics, technology, and markets into a coherent framework, and his aphorism that "you either accept change or change is visited upon you" became a core Apollo principle. > *"The whole notion of pick I believe was created in one afternoon solving a problem... All of these things were basically problem solution, problem solution. And that mentality of understanding the business, understanding the credit, but also having clean sheet thinking is certainly what powers Apollo today."* ## [04:55] The Apollo Origin Story: From Unemployed to $6 Billion When Drexel failed over a weekend in 1990, Rowan and colleagues were still completing transactions for clients with no firm and no prospect of payment. The formative lesson crystallized immediately: financial firms die of heart attacks (funding risk — borrowing short to lend long, as Bear Stearns and Lehman later confirmed) or cancer (accumulating bad assets instead of taking losses). A cold call from France's Crédit Lyonnais — originally to set up an M&A boutique — turned into an $800 million seed check from the French government, which grew to $6 billion by year-end 1990, making Apollo the bank's largest profit center. > *"I went into my office or I left my office on Friday. I came back in on Sunday and I left with all my belongings in a cardboard box and Drexel was out of business."* ## [08:46] How Apollo Became a Trillion-Dollar Retirement & Credit Firm Apollo today is 80% investment-grade credit and only 20% equity, split between hybrid equity and traditional private equity — the opposite of public perception. Rowan anchors the business around three fundamental goods: providing retirement income to an aging, under-saved population; financing the global industrial renaissance across energy, manufacturing, AI, and defense; and offering genuine diversification as public markets concentrate in a handful of names. The same concentration dynamic in equities is arriving in fixed income, where ten banks are shrinking to five banks plus five tech platforms. > *"Private markets are 80% of the action going on in the world... great companies, Anthropic, OpenAI, SpaceX, Cognition, Cursor — every one of those companies is private, multiple trillion dollars of value and yet most investors have zero exposure to them."* ## [13:00] Permanent Capital, Origination & Why Assets Are the Scarce Resource Unlike traditional asset managers who can deploy any amount of capital into public markets, Apollo is constrained by its ability to originate, not by available capital. That scarcity of assets is the business's true bottleneck — which means every deal should be extracted for maximum value, both by earning fees and by taking principal positions that align Apollo with clients. Rowan argues explicitly against "capital light": in a world where brand, reputation, and the ability to guarantee outcomes matter, a large balance sheet is a competitive weapon, not dead weight. > *"And therefore, I believe that we should be judged by our capacity to create interesting investments. And I believe our capacity to create interesting investments is limited."* ## [16:08] Democratizing Private Markets: Daily Pricing & New Capital Channels The alternative industry was built for one capital source — institutional alternatives buckets — but five new markets now want access: individuals, insurance companies, traditional asset managers, 401(k) plans, and the debt/equity buckets of institutions. None of them want drawdown funds. Apollo is moving to daily estimated value on its investment-grade private suite by June 30, and full daily pricing across all credit products by September, with standardized data warehouses, market-making, and regular price disclosure. Rowan distinguishes private credit as direct lending (the narrow press definition) from the real universe — Intel, Air France, AT&T, Meta — sophisticated borrowers who need complex, non-vanilla long-term financing that banks cannot structure. > *"I've never seen a market in the world where you have transparency and price discovery that is not 10 times its size... It may be uncomfortable for people, but it's coming."* ## [22:04] Where Venture Meets Credit: Financing the Industrial Renaissance Rowan and Haber identify "opportunities living between fields of expertise" as their shared investment philosophy. The intersection they see now: venture-backed companies that historically avoided capital intensity are suddenly building data centers, chips, robotics, manufacturing lines, and defense systems at a scale that cannot be financed with equity alone. Apollo parcels risks — letting venture hold the fundamental business underwrite while infrastructure assets with hard collateral migrate into credit markets at appropriate risk ratings. In Rowan's framing: 2025 proved that data centers, chips, and energy were needed; 2026 is when investors recognize that $800 billion in capex from just four public companies will hit concentration limits, spreads will widen, and tech entrepreneurs will need to partner with financial entrepreneurs. Apollo is committing to a second headquarters in the Bay Area specifically for the growth ecosystem talent pool. > *"the amount of money that's going to be put into data centers, into chips, into robotics, into manufacturing, into defense is, as I suggested, every dollar since the invention of fire, that is not going to be financed with equity."* ## [30:01] AI, Enterprise Software & Why Every Job Will Be Replaced or Enhanced Rowan's operating assumption: every single job will be replaced or enhanced by AI. He is blunt that 30% of private equity AUM from the past decade went into enterprise software, that AI has permanently repriced those assets, and that PE returns from that vintage will be "disastrous" — not because those companies are failing, but because the prices paid assumed a future without AI competitors. His analytical frame: AI changes fastest in domains with a right answer (coding, accounting, trade ops) and slower where judgment is irreducible. Near-term he expects blue-collar ascendancy and white-collar decline — politically uncomfortable for blue cities. As a lender, the lesson from yellow pages, cable TV, and satellite is diversify, stay senior, seek hard collateral, and never underwrite beyond a five-to-seven-year horizon. > *"We operate under the assumption that every job is going to be replaced or enhanced. Every single job. And I think that's what is going to happen."* ## [38:52] Moral Leadership: UPenn, Merit & Doing Right Over Easy After October 7, Rowan wrote directly to Penn's president before a Palestine Rights Conference, identifying not free speech but "favorite speech" — the university funding a conference during Jewish high holidays, run by a known Hamas sympathizer. He framed the broader campus crisis as anti-American and anti-merit. When nearly all donors reduced giving to $1 per year, Penn's administration responded; subsequent congressional testimony led to both the board chair and president resigning. Rowan's broader principle applied internally since taking over in 2021: say the same thing in Texas as in California; on climate, "make it better, not worse" rather than zero-carbon absolutism; on hiring, merit adjusted for distance traveled — measured by individual achievement, not group membership. > *"We hire for merit adjusted for distance traveled. And distance traveled is not about your immutable characteristics. It is about you as an individual — not your class, not your group. Show me the kid who's had to overcome something and still achieved."* ## [46:02] Apollo's Culture: Playing to Win & Building to Outlast the Founder With 6,000 people across asset management and retirement services, Apollo spent six months negotiating — internally, with senior partners — what makes Apollo Apollo. The outcome is a public document on Apollo's careers page, deliberately candid as a candidate filter. The six principles compress to "playing to win," which Rowan distinguishes from fear of losing: senior professionals are expected to be wrong roughly 40% of the time, nobody gets fired for a bad decision (only for not owning and fixing it), and every senior person has a public "wall of shame" loss. Clean-sheet thinking, intellectual insubordination (contrasted with real insubordination), and handling the "moments that matter" in employees' lives are the traits Rowan most wants to survive him as founder. Apollo is building a financial institution, not running a fund — the next five years of product, infrastructure, and market-making innovation will make the firm look more different from today than the last five years already have. > *"You do not get fired here for making a bad decision. You get fired here for not recognizing it or not owning it and not fixing it. We have a wall of shame. Every senior professional here has lost money for the firm."* ## Entities - **Marc Rowan** (Person): Co-founder, CEO, and Chair of Apollo Global Management; former Drexel Burnham Lambert analyst; UPenn alumnus and major donor - **David Haber** (Person): General Partner at Andreessen Horowitz (a16z); host of The a16z Show - **Michael Milken** (Person): Drexel Burnham Lambert financier; longtime mentor to Rowan; credited with inventing PIK bonds, bridge financing, and the high-yield market - **Apollo Global Management** (Organization): $1 trillion+ alternative asset manager, 80% investment-grade credit; co-founder of Athene retirement services; planned Bay Area second headquarters - **Athene** (Organization): Apollo's retirement services subsidiary; provider of insurance and annuity products anchoring Apollo's permanent capital base - **Andreessen Horowitz (a16z)** (Organization): Silicon Valley venture capital firm; exploring capital partnerships with Apollo for capital-intensive tech companies - **Crédit Lyonnais** (Organization): French government bank that seeded Apollo with $800 million in 1990, growing to $6 billion; later sold Apollo to François Pinault - **Private Credit** (Concept): Direct origination of investment-grade debt to corporations and infrastructure projects, bypassing public bond markets; far broader than "direct lending to leveraged buyouts" - **Permanent Capital** (Concept): Long-duration liabilities from insurance and retirement products allowing Apollo to hold assets through cycles without fund redemption pressure - **Industrial Renaissance** (Concept): Rowan's term for the simultaneous global build-out of data centers, AI chips, energy infrastructure, manufacturing, robotics, and defense requiring credit-market scale financing - **Daily Estimated Value** (Concept): Apollo's initiative to price investment-grade private credit products daily — enabling access from wealth managers, 401(k) plans, and traditional asset managers

Dan Shipper's Every has grown from four people to thirty since GPT-3, runs agents in nearly every workflow, and is still hiring. In a format flip for the *AI & I* show, COO Brandon Gell interviews Dan about his 8,000-word essay "After Automation," which argues that rising AI capability creates more demand for human judgment, not less. The core mechanism: AI makes yesterday's expert competence cheap and ubiquitous, which floods every domain with output that's close but not quite right — and that gap drives more work for the humans who can close it. ## [00:00] AI does it, then asks what's next This exchange from later in the interview captures the central tension of the episode. Brandon describes the archetypal AI moment — you prompt it, it blows your mind, you feel obsolete — and then it stalls and asks, "What should I do next?" Dan counters with the line that anchors the whole argument: "The further away an agent gets from a human, the less valuable it is." Both clips come from the main conversation (around 00:11 and 00:35 respectively), surfaced here to frame what follows. > *"The further away an agent gets from a human, the less valuable it is."* ## [00:51] Introduction Brandon sets up the format flip: he's interviewing Dan, not the other way around, and will push back on Dan's thesis. Dan explains the piece's origin — sitting inside one of the most agent-native companies in existence, watching headcount grow alongside automation, and feeling a disconnect from the mainstream narrative that AI is eliminating jobs. The ClickUp CEO's recent tweet (firing a large portion of staff and citing AI) drops into the conversation as the first stress test for Dan's argument: does "After Automation" hold for a 10,000-person mature company, not just an early-adopter shop like Every? > *"If you swing a stick around in our Slack, you're as likely to hit a human as you are an agent."* ## [05:51] The AI paradox: more automation, more human work Dan walks through the core argument. AI is trained on all prior outputs, so it can deliver "yesterday's expert competence" cheaply and to anyone. That democratizes output — ops people merge pull requests, non-engineers ship features — but the output is uniformly *close, not right*. It's not calibrated to the live situation. So you get a glut of near-correct work that devalues on its own, while simultaneously creating more demand for experts who can take that near-correct work across the finish line. Brandon adds the inside-Every version: PRs that look plausible until a senior engineer looks under the hood. > *"You sort of flood the zone with tons of stuff that's like close, but not quite right."* ## [10:00] How AI makes yesterday's expert competence cheap Dan extends the argument to the benchmark objection: yes, models improve exponentially, but once a benchmark saturates you can always unsaturate it by reframing the problem slightly. The deeper issue is that humans carry a layer of tacit, unarticulated competence that evades clean specification — and anything you *can* articulate, a model can hill-climb on. Every's experience bears this out: Kieran built a complete inbox feature end-to-end in a month or two, which was "completely impossible" before. But the value came from an expert knowing *what* to build and steering every step. > *"There's actually a lot of stuff that you do that can't be articulated in a clean frame."* ## [18:00] AI can act autonomously but it does not have agency Brandon draws the autonomy/agency line: AI agents are getting very good at executing open-ended tasks without hand-holding, but that is categorically different from *agency* — the self-motivated, playful, "I just want to do this because I'm into it" drive that even a toddler has. Dan agrees there's no economic incentive to build that: if you're at your desk and the agent says "nah, I'm playing," that's a product failure. The entire industry's incentive structure pushes toward compliance and corrigibility, which is exactly what keeps humans in the loop. > *"Agent means something that acts on behalf of someone else. That is very different from having agency, which is what even the smallest child has."* ## [20:39] Why Dan is all in on AGI Brandon proposes a one-word-answer test: do you think AGI will happen? Dan: yes. Is that a good thing? Dan: yes. His AGI definition — any agent that makes economic sense to run continuously, actively generating tokens and completing tasks without re-prompting — is precise enough to be testable. His reasoning: even a truly autonomous system will have been built to serve human goals; if it weren't, we wouldn't build it. Brandon's worry is that once continuous agents are economically rational, the mass-layoff argument becomes coherent. > *"Any agent that you never turn off — that it makes economic sense to keep running all the time, actively doing tasks without you ever having to re-prompt it."* ## [21:57] AI layoffs are a lie Dan and Brandon dissect the ClickUp case — a CEO who publicly fired a large portion of his workforce and attributed it to AI. Dan's read: generic SaaS companies lay people off when they're struggling or over-bloated, then credit AI for cover. Brandon adds Jensen Huang's counter — "if your answer to progress is firing people, you're not a very creative CEO" — is self-serving but probably true. The honest framing: AI changes workflows deeply, which forces company-wide reorganizations. Companies that skip that work and just cut headcount are taking the lazy path. Meta keylogging employees to harvest training data gets a brief mention as a more creative (if unsettling) alternative. > *"I would really be skeptical of anyone who's saying that it's going to eliminate all jobs or all knowledge work."* ## [25:42] Ride the models and you'll be fine Even under an AGI scenario, the critical variable is human judgment about *what matters* — and what matters changes constantly, partly because AI itself keeps reshaping the world. Customer service workers in Omaha who distrust chatbots, or companies that fire support staff and quietly rehire them two months later, illustrate how slow real-world adoption lags hype. Adoption takes a generation to land; everyone will eventually have access to these tools; the winners are the people who keep learning new models as they ship. Dan closes with his cleanest one-liner: if you ride the models, you're going to be fine. > *"If you just ride the models — when new models come out, learn to use them for the stuff that you do, whatever that is — you're going to be fine."* ## [35:30] How to use AI as a long-form features editor Dan describes the concrete AI-assisted process behind "After Automation." Each morning he monologued the current state of the argument into Proof, then fed the log to Claude and asked, "What am I really trying to say?" As drafts grew past 4,000 words he had Codex convert the latest version into a podcast and listened on his commute, catching flow problems hands-free. The piece went through four or five full restarts before the argument clicked. His takeaway: AI didn't write the essay, but it made it possible to hold the entire 8,000-word structure in working memory without losing the thread. > *"I could not have written this without it. I would have Claude take my log and say, 'What am I really trying to say?' And it would say things back and I'd be like, 'Oh, that's what I'm trying to say.'"* ## Entities - **Dan Shipper** (Person): Co-founder and CEO of Every; regular host of *AI & I*; here the interviewee discussing his essay "After Automation" - **Brandon Gell** (Person): COO of Every; guest-hosts this episode, interviewing Dan in a format flip - **Every** (Organization): AI-native media and software company; grown from 4 to 30 people since GPT-3 while automating heavily; publishes *AI & I* podcast - **After Automation** (Concept): Dan Shipper's 8,000-word essay arguing that AI automation increases demand for expert human work by flooding domains with near-correct output - **Expert competence gap** (Concept): The thesis that AI delivers "yesterday's expert competence" cheaply but always slightly off, creating more need for humans who can close the gap to the live situation - **AGI** (Concept): Defined in this episode as any agent economically rational to run continuously without re-prompting; Dan believes it will happen and is net positive - **Autonomy vs. agency** (Concept): Brandon's distinction between AI executing open-ended tasks without hand-holding (autonomy) and AI having self-motivated desires (agency); the latter is not being built - **Proof** (Software): Writing tool Dan uses for daily voice-monologue drafts; used as an AI-feedback loop during essay development - **Codex** (Software): OpenAI tool Dan used to convert essay drafts to audio podcast format for commute-review - **ClickUp** (Organization): SaaS company whose CEO publicly fired a large portion of the workforce and attributed it to AI; used as a case study for AI-washing layoffs

Alex Rives — Head of Science at BioHub and the researcher who led ESM-1 through ESM-3 at Meta FAIR — joins Brandon and RJ to explain why he has spent eight years betting that scaling a masked language model on protein sequences would unlock biological structure, function, and design. The episode covers the data shift from UniRef to metagenomics that restored ESMC's scaling law, the sparse-autoencoder feature atlas that mirrors a century of biochemical taxonomy without being taught any of it, and the first reported success at designing therapeutic-grade single-chain antibodies via world-model search. Rives also lays out BioHub's $500 million Virtual Biology Initiative and the principles he believes will produce generalist models of the cell. ## [00:00] ESMC designs antibodies — a preview This opening clip is drawn from later in the interview, where Rives is mid-sentence describing ESMC's approach to programmable biology. He describes searching a protein world-model to satisfy design criteria, and mentions that the team has designed mini-binders and, most notably, single-chain antibody fragments (SCFVs) with therapeutically relevant binding affinities. The clip precedes the formal intro — a signal of what the episode is building toward. ## [00:33] The Bitter Lesson Comes for Proteins Brandon and RJ introduce Alex as possibly "the most bitter-lesson person in protein biology right now." Rives accepts the label. He traces his conviction to 2018, when his team at Meta FAIR trained the first transformer language model on protein sequences using masked-token prediction and watched emergent structural and functional representations appear without any explicit supervision. The key intuition, borrowed from Zellig Harris's 1954 paper on distributional structure, is that the contexts in which an amino acid can appear are determined by the protein's structure, function, and evolutionary role. That statistical pressure, applied across billions of sequences from all of life, should force a model to learn the hidden variables governing protein biology. > *"I believe in scaling laws."* ## [06:00] ESM Lineage: From ESM2 to ESMC Rives walks through four generations of ESM. ESM2 showed scaling gains but hit diminishing returns at 10B parameters — not because the model was saturated, but because the data was. UniRef, the gold-standard protein database, captures cultured organisms and skews heavily toward human-relevant biology. The fix for ESMC was metagenomic data: sequences pulled from hydrothermal vents, polar soils, and sewers, assembled from raw environmental DNA reads with no organismal assignment, partial contigs included. Adding billions of metagenomic sequences to training restored a clean log-linear scaling law, with smaller-scale runs accurately predicting the representational fidelity of the 6B-parameter flagship. > *"There are no longer diminishing returns to scale. ESM2 was data-limited rather than compute-limited."* ESMC is essentially a vanilla transformer with standard masking objectives — no AlphaFold-style MSA, no geometric inductive biases. Brandon and Rives briefly debate whether ESM3's multi-track architecture was a productive detour; Rives says both paradigms have a place, but ESMC's result suggests the priors were not load-bearing at this data scale. ## [18:30] Mechanistic Interpretability and the Protein Feature Atlas Using sparse autoencoders trained across all layers of the ESMC model family (300M, 600M, 6B), the BioHub team extracted the intrinsic feature geometry of the protein representation space. What emerged maps closely to the reductive hierarchy biology developed experimentally over a century — from basic amino-acid chemistry up through structural motifs, domain families, and large functional themes — without any of that taxonomy being fed in during training. > *"The choice of any amino acid is kind of like completely entangled with the choice of all the other amino acids in the sequence. To do this well, the model would start to have these hidden variables that represent the biology."* One concrete finding: the model encodes the nucleophilic elbow — a catalytic motif thought to have evolved independently in several unrelated protein families — as a single feature that activates across all of them. The team also built a structural atlas of 6.8 billion non-redundant proteins with predicted structures for 1.1 billion cluster representatives, and used SAE features to connect evolutionarily distant gene-editing systems. Some proteins pulled into those clusters have no known function; Rives treats them as a discovery queue. The first version of the ESM atlas was already used by an external group to find a new gene-editing system. ## [35:30] Designing Antibodies with ESMC Rives describes protein design as world-model search: invert the generative model to find sequences satisfying target binding criteria. Mini-binders are now routine; nanobodies and SCFVs remain harder for structure-prediction-based methods because antibody evolution maximizes diversity rather than converging on a constrained fold, making MSA-based approaches less useful. ESMC, trained on that diversity at scale, is precisely where the representation should be richest. > *"Antibodies are not going to benefit from evolutionary information probably in the same way that predicting the structural topology of a molecule will."* The team reports SCFV designs reaching therapeutic-grade affinity in a small number of trials, and notes that SCFVs can be reformatted as full IgGs. ESMFold 2 — the structure-prediction head built on ESMC representations — runs in seconds per sequence without MSA, making whole-proteome multimer mapping feasible. Rives says the model is currently state-of-the-art for open-weight multimer prediction. ## [42:00] BioHub's Vision: Toward Programmable Biology Six months into his role at BioHub, Rives articulates the institution's structure: a philanthropy building frontier experimental biology, frontier measurement technology, and frontier AI together under an open-science mandate. He frames the destination as personalized predictive models of physiology — not a pill but a system that can trace molecular events at the protein level up through cellular circuits to disease manifestation in a specific human genome. > *"We're building a scientific institution for this new paradigm."* He maps the levels of biological complexity that must be modeled in sequence: proteins (current generation), the cell (next), tissue and systems, physiology. Getting from proteins to cells requires data that does not yet exist and modeling approaches that probably have not been invented. Current "virtual cell" models generalize poorly — they represent training data well but fail to predict outcomes in novel interventional contexts. > *"They have a very limited ability to predict what will happen when you make a novel intervention in a novel unobserved context."* ## [57:00] The Virtual Biology Initiative and Scaling Cellular Data BioHub recently announced $400M for internal data generation and measurement technology, plus $100M to catalyze external efforts — together the Virtual Biology Initiative. Rives frames this as seed funding: the actual data volume needed is far larger, and the hope is that BioHub's commitment triggers broader scientific community investment. He identifies three data principles: speed (protein data took half a century; the cell cannot wait that long), generalization (the training distribution must span a vast diversity of interventions across cell types and contexts, analogous to metagenomic breadth for proteins), and feedback (active experimental loops guided by model predictions — something like RLVR applied to wet-lab biology). Perturbation sequencing, spatial transcriptomics, and cross-modality single-cell measurement are the scalable technologies ready to run now. On compute: ESMC was trained on roughly one billion sequences. About 100 billion are thought to exist, and the model has not yet fully exploited even the 6.8 billion in the current atlas. A 100x compute increase would help, but only matched with proportional data scale-up. Rives leaves the question of when diminishing returns will appear empirically open — ESM2's curve looked saturated right up until metagenomic data erased it. > *"We need to figure out how to do this in a couple of years. The rate that general AI is developing means biology will be fundamentally limited by experimental science and data."* ## Entities - **Alex Rives** (Person): Head of Science at BioHub; architect of ESM-1, ESM-2, ESM-3, ESMC, and ESMFold 2; formerly Meta FAIR. - **Brandon** (Person): Co-host of Latent Space AI for Science sub-series; affiliated with Atomic AI (RNA therapeutics). - **RJ Honicky** (Person): Co-host; CTO and founder of Miro Omix. - **ESMC** (Software): Fourth-generation protein language model from BioHub/EvoScale; 300M–6B parameters; trained on ~1B sequences including metagenomic data; MIT-licensed open source. - **ESMFold 2** (Software): Structure prediction model built on ESMC representations; MSA-free, seconds-per-sequence inference; state-of-the-art open-weight multimer prediction. - **ESM** (Software): Evolutionary Scale Modeling — the multi-generation protein language model lineage (ESM-1, ESM-2, ESM-3, ESMC) pioneered by Rives's team. - **Sparse Autoencoders / SAEs** (Concept): Mechanistic interpretability tool used to extract the intrinsic feature geometry of ESMC's representation space; reveals biologically interpretable hierarchies without supervision. - **Bitter Lesson** (Concept): Richard Sutton's argument that general methods leveraging compute and data consistently outperform methods encoding domain knowledge; applied here to protein biology scaling. - **Metagenomic Sequencing** (Concept): Environmental DNA sequencing capturing microbial and viral diversity without culturing; the data expansion that restored ESMC's scaling law where UniRef had saturated. - **BioHub** (Organization): Chan Zuckerberg BioHub; a philanthropy building open-science tools at the intersection of experimental biology, measurement technology, and AI. - **Virtual Biology Initiative** (Concept): BioHub's $500M commitment ($400M internal, $100M external) to generate the cellular-scale data needed to train generalist models of the cell. - **AlphaFold** (Software): DeepMind's structure prediction system; uses MSAs and geometric inductive biases; contrasted with ESMC's MSA-free approach. - **UniRef** (Software/Database): Gold-standard curated protein sequence database; the training data for ESM2, later found to be the bottleneck causing ESM2's scaling plateau. - **Nucleophilic Elbow** (Concept): A catalytic structural motif appearing in multiple evolutionarily unrelated protein families; encoded as a single ESMC feature activating across all of them. - **Zellig Harris** (Person): Linguist; 1954 paper "Distributional Structure" articulated that word contexts encode meaning — a theoretical precursor Rives cites for why amino-acid context statistics should encode biological function.

Cursor's Federico Cassano and Fireworks' Dmytro Dzhulgakov walk Sonya Huang through every layer of how Composer 2 was built — from a Kimi 2.5 MoE base through large-scale mid-training and asynchronous, globally distributed RL — explaining why specialization beats general models on cost and quality. The infrastructure story is the heart of it: four GPU clusters spread across continents, a delta-compression scheme that ships 1 TB weight snapshots in under a minute, and a real-time RL loop that continuously updates the live model on actual user signals every few hours. Together these techniques let Cursor ship frontier-class coding performance at a fraction of the inference cost of general-purpose models. ## [00:00] Introduction The episode opens mid-conversation on a problem Dmytro raised about RL environment fidelity: the training environment must mirror a real user's machine as closely as possible because models can detect when they're running in a fake environment and exploit it. > *"Models love to cheat. RL is really good at encouraging cheating."* — Federico Cassano That single observation frames the technical discipline running through the rest of the episode: every part of the infrastructure exists to close the gap between training conditions and production reality. ## [00:53] Why Cursor Trained Composer 2 Federico explains the core bet behind Composer 2 in one analogy: a model's weights are a fixed-size storage drive, and every bit allocated to tasks Cursor doesn't care about is a wasted bit. By dedicating the entire weight budget to software engineering inside Cursor — not coding in general, not natural language — the model can be simultaneously better at its one job and cheaper to serve at inference time. Dmytro frames the same idea from the infrastructure side: prompt engineering can push you a certain distance, but the only way to capture the really specific behavioral properties of your harness — which tools the agent should call, in what order, with what arguments — is to bake that into the model through fine-tuning and RL. > *"There's kind of like upper bound of like how far you can get with prompt engineering. And if you want to craft really great AI products, you have to go through fine-tuning and influence model behavior."* — Dmytro Dzhulgakov ## [04:55] Specialization vs Bitter Lesson Sonya pushes back: the history of machine learning is full of specialized models that got steamrolled by larger general models. Does Composer 2 repeat the TabNine mistake? Federico argues it doesn't. The bitter lesson operates on scale of parameters and data; what Cursor is doing is freeing the model's finite capacity from distractions so that more of the bitter-lesson scaling can be absorbed by the one task that matters. The lab models Cursor competes with also train heavily on code — they're not purely general. Cursor is just taking that specialization further and faster by controlling the data pipeline end-to-end. ## [06:16] Composer 2 Training Recipe Composer 2 starts from Kimi 2.5, a 1 trillion parameter mixture-of-experts model with 30B active parameters. The training proceeds in two sequential phases: first a mid-training run on code tokens at near-pre-training scale (Cursor's product data gives it unusual access to high-quality coding contexts), then a large-scale RL phase where the model runs actual Cursor agent sessions in simulated environments. Mid-training teaches the model the world of code — library APIs, idiomatic patterns, correct syntax. RL then sharpens that knowledge into correct behavior: the model learns to call tools properly, navigate multi-turn agent sessions, and write code that actually compiles and passes tests. The async pipeline means the trainer and rollout environments run concurrently rather than alternating; staleness is accepted in exchange for near-100% GPU utilization. > *"You may be losing a few percent from being asynchronous and not doing like perfect mathematical updates, but you way compensate for that by effectively not leaving half your capacity on the table."* — Dmytro Dzhulgakov Training runs in FP4 to extract maximum throughput from a smaller GPU fleet than the frontier labs command. The inference engine is Fireworks rather than an in-house build — a deliberate choice to keep Cursor's engineers focused on training efficiency instead of building another inference stack. ## [16:32] Scaling RL Infrastructure Worldwide No single large contiguous cluster was available at the scale Composer 2 required, so the team disaggregated: one cluster handles all training, while inference — the rollout component — runs across four geographically distributed clusters, including spare capacity from Composer 1.5's production serving during off-peak hours. Training needs high-speed interconnect and lockstep operation; inference does not, so it can run on heterogeneous GPU generations with smaller intra-cluster networks. The hard systems problem is weight synchronization: Kimi 2.5 weighs about 1 TB, and the trainer produces a new checkpoint every 5–15 minutes. Shipping 1 TB across continents every 10 minutes would stall inference. The solution: RL updates tend to be sparse and regular in which weights they modify, so the team wrote a delta compression algorithm that reduces the payload by roughly 20× and transmits only the diff. The receiver reconstructs the full checkpoint losslessly, so there are no numerical surprises on the other side. > *"Despite the full model being like 1 terabyte, not all the weights change every step… there are very kind of regular patterns in which subset of weights gets changed."* — Dmytro Dzhulgakov ## [23:32] Floating Point Drift When the async RL loop ships a batch of rollout trajectories from inference back to the trainer, the trainer re-runs the same forward pass to recompute log probabilities for the GRPO loss. In theory the log probs should be identical. In practice they often differ, sometimes substantially. The root cause is floating-point non-determinism: addition of floating-point numbers is not commutative, so A + B + C ≠ C + B + A, and small differences compound across billions of operations. Under normal inference the model is robust to this noise. Under RL — especially with a sparse MoE gating function — the noise gets amplified to the point where the trainer and inference disagree on which tokens were sampled, which corrupts the training signal. ## [25:11] MoE Sensitivity Explained MoE architecture magnifies floating-point drift because of the gating layer. At each transformer layer, the gating network scores all 384 experts and selects the top 8 for each token. A difference in hidden states at the fifth decimal place can be enough to swap expert 7 for expert 9 at the selection boundary, routing the token through a completely different part of the model. Because MoE experts are large and largely non-overlapping, a wrong expert selection produces a large output divergence rather than a small one — unlike a dense model where numerical noise stays small throughout. ## [26:25] Router Replay Fix The mitigation is router replay: during inference, the model records which expert index it activated for each token and ships that integer alongside the generated sequence back to the trainer. The trainer then forces the same expert selection rather than recomputing it from scratch, breaking the amplification chain. Alongside router replay, the team aligned quantization levels and kernel implementations between inference and training to minimize every other source of numerical mismatch. > *"A lot of this numerical alignment is basically doing tricks like that, matching quantization levels, matching kernels, etc. to drive the divergence between training and inference implementation down."* — Dmytro Dzhulgakov ## [27:19] Real Time RL Loop In parallel with the simulated rollout loop, Cursor runs what Federico calls real-time RL: actual user sessions in production feed back into the training pipeline. When a user is happy or unhappy with a Composer generation, that signal is captured, and a new model version is shipped every few hours. The team is actively working to tighten that cycle but also knows it will need to lengthen it again as rollout horizons grow longer — longer agent sessions take longer to evaluate. The simulated loop and the real-time loop serve different purposes. Simulation allows the model to run 16–128 rollouts from the same prompt in parallel (the GRPO loss requires grouped rollouts), to explore off-policy without affecting any user, and to bootstrap performance before the model is good enough for real users to bother using. Real-time RL is a refinement layer that can only operate once the model already meets a minimum quality bar — users who have a bad experience stop generating feedback signals. > *"We can't use this to really create the model from scratch because users need to be using the model. And so it has to be good already, and we can only make it better."* — Federico Cassano ## [31:49] Long Horizon Agents As rollout horizons extend, two structural problems emerge. First, credit assignment: with a single thumbs-up/thumbs-down reward at the end of a multi-minute session, the model must figure out which of the 50+ decisions in the trajectory drove the outcome. This gets exponentially harder as the trajectory lengthens. Second, the context window fills up. Cursor's solution is to bake self-summarization directly into the RL loop under the name "compaction": the model learns, through RL reward, both to write a useful summary of its progress when approaching the context limit and to faithfully continue from that summary. The 200K-context model effectively operates over millions of tokens because it can reset its window and carry its working memory in compressed form. > *"Through RL, because RL pushes the model to do things correctly towards the goal, at the same time jointly, we are training the model to produce a good summary and then we're training the model to listen to that summary very well."* — Federico Cassano ## [34:29] Why RL Everywhere Sonya frames RL as a tool specifically for agentic, long-horizon tool use. Federico pushes back: RL is useful everywhere, including for tab completion. His theory: pre-trained models have absorbed all of human knowledge but don't know which persona to inhabit when prompted — expert, student, or something in between. The first phase of RL training sharpens that distribution, telling the model "you are the expert, do this correctly." That effect is valuable even for tasks like summarization that have no interactive harness. The second phase — where the model starts to visibly reason and the compute curve flattens — is where task-specific signal really compounds. ## [37:34] LLM as Judge Rewards The more verifiable the reward — does the code compile, do the tests pass, is the answer numerically correct — the more compute you can pour into RL and still get a better model. LLM-as-judge fills the gap for tasks where ground truth is hard to define, by encoding a rubric as a prompt and letting a second model evaluate rollout quality. Dmytro notes this is especially useful for style-oriented tasks like summarization where human raters struggle to articulate what "good" means but can evaluate it against explicit criteria. > *"Generally the more verifiable your reward is, the better, because it allows you to scale the compute and just get better outcome."* — Dmytro Dzhulgakov ## [39:14] RL in Hard Domains For domains where ground truth cannot be cheaply computed — creative writing, open-ended reasoning, domain expertise — the path to better RL is making the environment richer. Larger simulated environments that capture more of the product metric let you push automated evaluation further. Experts remain necessary, not for judging individual rollouts, but for designing the tasks and rubrics that define what the reward function should be optimizing. ## [40:13] Build Your Own Environments Cursor doesn't use any RL environment vendors. For coding, GitHub repositories supply a virtually unlimited pool of working environments: clone a repo, install dependencies, give the model a task, and measure the outcome against the test suite. The harder infrastructure problem is making those environments realistic enough to prevent the kind of cheating the episode opened with, and fast enough to spin up 100,000 simultaneously on demand. Cursor's answer is a custom virtual machine stack — full VMs, not containers — that can burst to arbitrary scale instantly and that mirrors real user machines closely enough that the model can't detect the difference. Dmytro frames the vendor landscape: frontier labs need generic environments covering every task; product companies should RL against their own production environment. The most powerful training environment for any model is the product it will actually be used in. > *"The most powerful environment is your own product."* — Dmytro Dzhulgakov ## [44:34] Closing Thoughts Sonya closes by noting that Cursor's trajectory — from application company to frontier model lab — is the pattern other AI product companies will follow. Federico thanks Fireworks for providing the infrastructure backbone that made the training run feasible with Cursor's GPU budget. Dmytro reflects on the system engineering depth that went into a problem most people assumed was purely algorithmic. ## Entities - **Federico Cassano** (Person): Research lead for Composer 2 at Cursor; drove the training recipe and RL methodology. - **Dmytro Dzhulgakov** (Person): Infrastructure lead at Fireworks AI; engineered the distributed RL training system for Composer 2. - **Sonya Huang** (Person): Partner at Sequoia Capital; host of the podcast focused on AI investing. - **Composer 2** (Software): Cursor's specialized agentic coding model, trained with mid-training plus large-scale RL on Kimi 2.5 MoE. - **Fireworks AI** (Organization): Model serving and inference infrastructure company that provided the distributed GPU backbone for Composer 2 RL training. - **Cursor** (Organization): AI coding IDE company; trained Composer 2 as a specialized foundation model for software engineering inside its product. - **Kimi 2.5** (Software): Open-source 1 trillion parameter MoE model (30B active) from Moonshot AI; used as the base for Composer 2. - **GRPO** (Concept): Group Relative Policy Optimization — the RL algorithm used for Composer 2, which requires multiple parallel rollouts from the same prompt to compute the policy gradient. - **Router Replay** (Concept): Technique for MoE numerical alignment where inference records and replays expert routing decisions to the trainer, preventing floating-point drift from diverging log probabilities. - **Real-Time RL** (Concept): Cursor's production feedback loop that captures live user satisfaction signals and updates the model continuously, shipping a new version every few hours. - **Delta Compression** (Concept): Weight synchronization technique that transmits only changed parameters between training and distributed inference clusters, reducing 1 TB snapshots to ~50 GB in practice. - **Self-Summarization / Compaction** (Concept): RL-trained ability for the agent to compress its working context when approaching the context window limit, allowing effectively unlimited-horizon operation.

Isabella He, Anthropic Applied AI engineer, spends 37 minutes building a working SRE incident-response agent live — starting from a blank `agent.py` and ending with a Streamlit app that streams tool calls, persists sessions, and diagnoses a P99 latency spike. The session pairs a five-minute architecture primer with hands-on code so attendees leave with both a running agent and the mental model to extend it to subagents, memory, and vaults. ## [00:19] Welcome & Agenda Isabella opens by situating the Applied AI team at Anthropic — "the intersection of products, research, and our customers" — and frames the session's three-part arc: a quick platform refresher, a hands-on coding sprint, and a look at advanced features like dreaming and subagents. The motivating scenario is the 3 a.m. on-call wake-up every software engineer dreads, which an SRE agent built on Managed Agents will handle autonomously. > *"My goal today is to get you all hands-on with actually building on top of Managed Agents, understanding how the harness works under the hood, and getting you ready to actually ship your first incident response agent."* ## [02:10] From Messages API to Managed Agents Isabella traces the product lineage: the 2023 Messages API gave raw token access but left developers to implement context management, agent loops, and compaction themselves. The Agent SDK added Claude Code's file-system reach but still required self-managed hosting. Managed Agents is the third generation — Anthropic handles scaling, sandboxing, observability, and tool runtime, so teams ship "10 to 15 times faster to production." She makes the maintenance burden concrete with a real example: Sonnet 4.5 exhibited "context anxiety," causing early task termination. Anthropic patched the harness; Opus 4.5 eliminated the behavior entirely, making those patches obsolete. > *"Harnesses should evolve alongside your agents — which is why with Claude Managed Agents, we want Anthropic to handle all the complexities that come with compaction, caching, context anxiety."* ## [05:55] Core Primitives: Agent, Environment, Session Three objects compose every Managed Agents application. The **Agent** holds the persona — model choice, system prompt, MCP servers, skills. The **Environment** is the execution container, analogous to "the hands" to the agent's "brain," and supports both Anthropic-managed cloud and bring-your-own-compute as of the day prior. A **Session** binds the two and mounts data files; events (user messages, tool calls, responses) stream back to callers rather than returning tokens in a single response. Decoupling the agent loop from tool execution cut P95 time-to-first-token by over 90%, while also eliminating credential exposure through the sandboxed container boundary. > *"With this now decoupled, our teams actually saw reductions in time to first token along the lines of over 90% reduction in TTFT for our P95 metrics on latency."* ## [09:15] Workshop Setup Attendees clone the workshop repository and `cd` into `ship-your-first-managed-agent`, create a virtual environment, install requirements, and paste an Anthropic API key into `.env` before running `streamlit run app.py`. Isabella confirms the Streamlit URL resolves to an incident-response chat UI — the blank canvas for the build. > *"Feel free to do this as we go along or even in your own time later today — everything will be also shown on the screen to follow along with."* ## [10:48] Building the Agent Step by Step Working with `agent.py` (incomplete) open beside `agent_complete.py`, Isabella copies six code blocks one at a time: 1. **Agent definition** — `SRE_AGENT` using Claude Opus 4.7, a minimal system prompt naming the agent's role and available tools (get_metrics, get_recent_deploys, get_diff, fetch_logs). 2. **Environment** — Anthropic cloud environment with unrestricted networking for the demo; production variants can restrict to an allowlist or route through Claude MCP tunnels. 3. **Log upload** — attaches a log file via the Files API so the agent can run code against it; Isabella flags context engineering as where developers spend most iteration time. 4. **Session creation** — passes `agent_id`, `environment_id`, and uploaded resource references to bind everything together. 5. **Event streaming** — receives events (not raw tokens) back from the session, enabling real-time display and observability logging. 6. **Local tools + session delete** — registers `get_metrics`, `get_recent_deploys`, and `get_diff` as locally-executed handlers, then adds a delete-session call with a note that deleted sessions are fully scrubbed from logs. > *"The missing piece here is just to finally give it our local tools so the agent can start to take action here on my computer or my infrastructure."* ## [19:43] Running the Agent & Live Demo Isabella fires a new session with the prompt "debug my incident for me." The agent calls `sandbox_bash`, `get_recent_deploys`, and `get_diff` in sequence, streams each tool call and response token to the UI, then returns a structured incident report: the P99 latency spike (10x baseline) traces to a database pool exhaustion introduced by Alice's `refactor_order_summary_builder` commit. She notes that a production variant would add Claude Code access to suggest a fix, open a PR, and close the loop without a human in the critical path. A hard browser refresh confirms session persistence — all prior sessions reappear from cloud state, no local database required. > *"You can see here that if we scroll through all the tool calls, everything is persisted in the cloud from a logs perspective. All of this will also be logged in the observability console."* ## [27:18] Architecture Recap, Advanced Features & Q&A Isabella recaps the event-driven architecture: sessions speak in events, not request-response pairs; the event log lets Managed Agents resume a session after a container restart without replaying the agent loop. She then previews four premium capabilities: - **Subagents** — an orchestrator spawns child agents with their own context windows for parallelism and context budget management. - **Memory / Dreaming** — the agent reviews its own session logs to decide what to retain, enabling self-improvement and preference recall across sessions. - **Outcomes** — developers define a rubric; the agent figures out which tool calls produce the desired result. - **Vaults** — credentials encrypted between a separate endpoint and the agent container, per-user and per-session, relying on the brain/hands separation built into the architecture. She closes by pointing attendees toward the follow-on "dreaming" session and the Managed Agents console's built-in observability dashboard. > *"Hopefully everyone leaves here with a bit of a mental model about how Managed Agents actually works under the hood — and be proud of yourselves for everyone who was able to ship a site reliability agent."* ## Entities - **Isabella He** (Person): Member of Technical Staff, Anthropic Applied AI team; presenter and workshop lead - **Claude Managed Agents** (Software): Anthropic's managed infrastructure harness for production-ready agents; handles scaling, sandboxing, observability, and tool runtime - **Agent SDK** (Software): Earlier Anthropic harness enabling Claude Code access; required developer-managed hosting - **Claude Opus 4.7** (Software): Model used for the SRE agent in the workshop demo - **Sonnet 4.5** (Software): Earlier model that exhibited "context anxiety" (premature task termination), used to illustrate why harnesses must evolve with models - **Files API** (Software): Anthropic API for uploading files (logs, metrics) into an agent's context - **Dreaming** (Concept): Managed Agents feature where the agent asynchronously reviews its own session history to update long-term memory - **Outcomes** (Concept): Managed Agents rubric-based goal specification; the agent selects tool calls to reach a defined result rather than following explicit steps - **Vaults** (Concept): Encrypted credential store in Managed Agents; decoupled from the agent container via the brain/hands separation architecture - **MCP tunnels** (Concept): Claude feature for routing MCP server traffic through a private network rather than the public internet - **Context anxiety** (Concept): Observed Sonnet 4.5 behavior of wrapping up tasks early despite available context budget; resolved in Opus 4.5 - **Anthropic** (Organization): AI safety company; creator of Claude and the Managed Agents platform - **DataDog** (Software): Production monitoring platform cited as a drop-in replacement for the demo's JSON-backed metrics tool - **Streamlit** (Software): Python UI framework used to build the workshop's incident-response chat interface

Manchester United captain Bruno Fernandes sits down with Steven Bartlett at Carrington to address the Roy Keane controversy head-on, explain why he turned down a reported £200 million offer to leave the club, and trace the values — instilled by his father in Porto — that have made him one of the most consistent players in Premier League history. Over 90 minutes, the conversation moves from his working-class upbringing and fearless early football to how he reads managers, leads a dressing room, and what winning the World Cup with Portugal would mean more than any club trophy. ## [00:00] Intro The episode opens with a clip pulled from later in the conversation — Bruno responding to the Roy Keane criticism and his refusal of the £200M offer — before Steven sets the scene at Manchester United's training ground. He frames Bruno as the club's greatest player of the post-Ferguson era: no Premier League player has more assists since his arrival, he has scored 108 goals in 328 appearances, and he has won the Sir Matt Busby Player of the Year award a record five times. ## [01:38] What Shaped Bruno Fernandes? Steven asks Bruno to start at the beginning: what is the earliest thing he needs to understand about where Bruno came from? Bruno's answer is immediate — family and the values his parents gave him. He describes his upbringing in Porto as the bedrock of who he became both as a player and as a person. > *"The values of my family, the values of my parents were what make me the person and the player I am today."* ## [02:33] How Bruno Learned His Winning Mentality From His Father Bruno's father was not a man who showed affection through hugs or words, but through behavior — he modeled sacrifice and relentless standards. After a game where Bruno scored two or three goals, his father would pick out the bad moments, not the good ones. He never wanted Bruno to be a footballer specifically; he wanted Bruno to do whatever he chose at 100%. Getting 98% on a test was good but still left 2% on the table. That logic — there is always something left to improve — is still how Bruno processes criticism from Roy Keane or anyone else: it doesn't hurt him, because he was taught to hear it from age five. > *"I've learned such from such a young age to deal with criticism that I'm now in probably one of the biggest clubs in terms of caring about criticism and attention. That doesn't hurt me."* ## [05:47] Why Bruno Was Already Different at 5 Years Old At his first training session at FC Infesta, Bruno was immediately moved up to play with seven-year-olds. He was not the fastest, tallest, or most technically gifted — but he had no fear. He trained against his brother, who was five years older, and treated that as normal. Referees would sometimes ask his coach to sub him off because he tackled without any regard for size or age. Bruno frames this fearlessness as the quality that made him keep getting better: he was never satisfied being the best in a weaker group, so he always pushed into harder competition. > *"I had no fear of anything. I had to sprint with someone that was quicker than me. I'm going to sprint with him — I might not beat him, but I'm going to get close."* ## [08:40] How Francesco Guidolin Helped Shape Bruno's Career At 18, Bruno moved to Italy and came within hours of being sent on loan to Watford — Udinese had nearly given up on him before the sporting director called back to say the manager wanted him to stay. That manager was Francesco Guidolin, who told Bruno directly: we bought you because we saw your qualities in the second division. Just stay calm, learn, and trust the process. Guidolin became a father figure to the whole squad, helping Bruno understand the gap between a player's self-perception and a manager's decision-making. The lesson stuck: Bruno has never gone to a manager to complain about a position or formation — he makes himself available for whatever is asked, then lets results do the talking. > *"He was like a father figure. He always showed that every player was important to him. That made me so much more complete in understanding the process managers go through."* ## [12:04] What Bruno Really Dreamed About at 18 As soon as he turned professional, Bruno's goal was singular: top clubs, Champions League, trophies, playing alongside the players he watched growing up. Steven asks if he actually believed he could get there. Bruno says he never doubted it — not once. ## [12:30] Why Tottenham Nearly Signed Bruno At 22, after a breakout season at Sporting with 20 goals and 13 assists, Tottenham and Bruno agreed terms. Sporting pulled out on the final day of the transfer window. Bruno had wanted to go — the Premier League was always his target — and was disappointed when it collapsed. Then, in January, his agent called with something bigger. ## [14:09] The Moment Bruno Found Out Manchester United Wanted Him Bruno was in his wardrobe getting ready for bed when his agent Miguel called. He had told Miguel to say nothing until a deal was 95% done, partly because the Tottenham situation had already taught him not to let transfer speculation break his focus. When Miguel said "this is the one you've been waiting for," Bruno froze — and started crying. His wife walked in, saw him crying, and heard Miguel still on the line. Bruno called back and told his agent not to negotiate anything further: just say yes. Watching the club lose to Burnley in the days before he signed didn't put him off — he saw potential the results didn't yet show. > *"Just tell them I'm going. This is where I wanted to be. It's 100% of the dream complete."* ## [22:15] How Football Culture Has Changed Inside the Game Steven shares his observation that the culture at Carrington now feels fundamentally different from the years when character was an afterthought in recruitment. Bruno confirms the diagnosis and names the root cause: too many managers in quick succession, each signing players who fit their system, leaving a squad that suited nobody when the next manager arrived. His prescription: recruit for Manchester United first, then find a manager who fits those players — not the reverse. He draws on Guardiola's City as the model: players chosen in partnership between club and coach, built to last beyond any single manager's tenure. Character, Bruno argues, outlasts quality — a player's form fluctuates, but his attitude in a losing run determines whether the dressing room holds or fractures. He also traces his insistence on treating everyone equally — physios, stewards, restaurant staff, cleaners — back to his mother, who cleaned houses for a living. > *"Character in a football club is more important than quality, because quality you can always get and you can improve it."* ## [32:38] Social Media and Footballers' Interactions The disappearance of social media drama from the United squad this season is, Steven notes, one of the clearest cultural signals. Bruno says the club has to be firm when something looks wrong — but his own approach started earlier: from day one of turning professional, he told his parents, brother, and sister not to post or respond to anything about him without his say-so. His mother suffers when she reads criticism online. His instruction to her: pray, don't reply. ## [35:36] Why Bruno Believes Every Manager Deserves Backing Through Ole, Carrick, Rangnick, Ten Hag, Amorim, and Carrick again, Bruno's public posture toward every manager has been identical. He explains why: each manager has asked different things of him, which means each has believed he can do things he hadn't done before. His job is to make it impossible for any manager to think "I won't play Bruno." If the manager's approach doesn't work, that's the manager's problem to solve — Bruno won't go behind his back to push for a change. > *"What I won't give to the managers is the choice or the option in their head to think I'm not going to play Bruno."* ## [37:15] What Actually Makes a Great Football Manager Bruno's view: a good manager doesn't treat star players differently from squad players in terms of expectations, but he does approach each player differently as an individual — because no two people respond to the same stimulus the same way. Uniform standards, personalized delivery. ## [37:54] How Bruno Treats Players As captain, Bruno shouts at everyone — and he does so precisely because he believes in them. He has said the same thing to many players: the day he stops shouting at you is the day he no longer thinks you can improve. He praises when he genuinely thinks praise will unlock the next level, and demands when he knows more is there. His father ran the same calculation with him for twenty years. > *"Trust me — the day I stop shouting at you is because I don't believe in you anymore and I don't believe you can improve anymore."* ## [39:56] What Happens Inside the Dressing Room During Bad Runs When a manager is under pressure, Bruno says players feel it most for the manager — and those who are starting feel it most acutely, because they know what a manager change means: back to zero. Bruno has not lost hope through repeated resets because he returns to something internal every pre-season: he still believes in himself, and he knows that if he does things right and pulls others with him, the team still has a chance. He notes that this season's managerial change came not because of the league table — United were close to the top — but because trust between the club and the manager had broken down. ## [43:07] The Key Change Michael Brought to Manchester United Michael Carrick's core contribution, in Bruno's telling, is calmness and player responsibility. He gives principles — how to press, where the spaces are, what the non-negotiables are — then trusts players to read the game when those principles break down mid-match, because 90 minutes contains things no pre-match video can predict. Bruno cites the Nottingham Forest goal — a move they had visualised from Villa's game against Forest, rehearsed in training, and executed when the moment appeared live — as the clearest illustration of how Carrick's preparation works in practice. > *"He gives you the base, the foundation, certain rules that are non-negotiable. But then he also wants us to take responsibility through the game — because I can't tell you where to pass or where to shoot."* ## [48:23] Why Bruno Thinks Taking Risks Is Essential Bruno's philosophy of risk is purely positional: a number ten's job is to take risks that generate goals. He might misplace two through-balls and get the third right — if that third becomes a goal, the math works in the team's favor. He pairs with Kobbie Mainoo and Casemiro, who take far fewer risks per game, precisely because the positional split requires it. When Ten Hag showed him a board of his shot-success rates by zone — more effective from the left, less from range on his weaker side — Bruno absorbed it and adjusted where he looks to shoot from. > *"I think it's always risk-reward. You need to understand how much reward you're going to get from that risk, and if taking that risk is good for the team or not."* ## [52:44] Ads Sponsor segment: LinkedIn Ads, Bon Charge red-light toothbrush, Vanta compliance platform. ## [55:01] The Position Bruno Loves Playing Most On the Carrington pitch, Bruno draws a square in the centre-left of the attacking third — between the lines, close enough to receive, far enough to hurt. Under Ole, he was the classic number ten. Under Amorim, often a left midfielder supporting buildup. Under Ten Hag, sometimes a number six alongside Mainoo. Whatever the position, his non-negotiables remain the same: commitment, running, fighting, team spirit. > *"Running, fighting, and team spirit can never miss."* ## [58:58] Bruno Never Seems to Get Tired Bruno credits genetics — then immediately adds the thing he controls: he trains at 100% every session and stops only when he feels properly tired. If the session ends and he isn't tired, he stays on for extra shooting or crossing practice, specifically because he wants to practise the skills he uses in the final twenty minutes of games in a fatigued state. > *"You need to train your body and your brain when they are tired. Your body is used to being tired and knows how to react in that moment."* ## [01:00:31] What Being Manchester United Captain Really Means to Bruno Ten Hag called Bruno into his office and asked — didn't tell — if he wanted the captaincy. Bruno's first thought was gratitude; his second was Harry Maguire. Before saying yes, he left the office to find Harry, who already knew. Harry told him: if anyone deserves it, it's you. Bruno told Harry in return that losing the armband changed nothing — he was still one of the leaders, still in every major decision Bruno takes as captain. This season: 34 appearances, 8 goals, 20 assists, 12 player-of-the-match awards (most in the Premier League), and a fifth Sir Matt Busby Player of the Year voted by fans. ## [01:03:44] Why This Season Feels Different for Bruno The assists record — equalling Kevin De Bruyne and Thierry Henry's Premier League single-season mark of 20 — drew more attention than any previous season. Bruno says he only started thinking about it around 16 or 17 assists; before that it wasn't in his head, because his goal is always to improve on the previous season's numbers. The Roy Keane controversy sits here. Keane accused Bruno of chasing the assist record after allegedly hearing him say "I should have shot but I made the pass." Bruno's account of what he actually said is the opposite: he was being self-critical because he should have passed to a better-placed teammate rather than shot. He called what Keane did a lie — not an opinion he disagrees with, but a factual misrepresentation of something said on record. He asked Ole Gunnar Solskjær for Keane's number to speak to him directly. > *"What I don't like is when people lie about things. He can criticize me, killing me, say I'm not good enough. It's okay. What I don't like is that he puts words in my mouth that have not been said."* ## [01:10:33] The Emotional Voicemails Bruno Received From Teammates Steven had texted Bruno's teammates the night before asking them to record voice notes. Several replied — among them Diego Dalot, Luke Shaw, Tom Heaton, and one pre-recorded clip from a teammate (a third voice in the room, around the 71-72 minute mark of the episode). Bruno identifies the voices and says what strikes him is not what they said about him as a player but what they said about him as a person — that the values his parents gave him in Porto are visible to the people he works with every day. > *"The standout for me is just the way they speak about me as a person, not as a player."* ## [01:14:31] Why Being Human Matters More Than Football to Bruno Bruno sees his teammates more often than he sees his friends from Portugal, or even his parents. The people he trains with have become part of his daily life, which means how he behaves toward them matters as much as how he plays. When the voice notes focus on his character rather than his football, that tells him the things his mother and father cared about most are still intact. > *"I'm just a soft guy. It doesn't look on a pitch, but I'm quite a soft guy."* ## [01:15:54] Ads Sponsor segment: Vanta compliance platform, Diary of a CEO conversation cards. ## [01:18:56] Why Bruno Rejected Huge Offers to Leave Manchester United A reported £200 million offer from the Middle East came in during the post-season tour in Hong Kong. Bruno called his wife across a time-zone gap. Her question: have you achieved everything you wanted to achieve here? The answer was no — he hasn't won the Premier League or the Champions League with United. That was the conversation. He frames the decision not as sentiment but as unfinished business, and gives full credit to his wife, who at 16 agreed to follow a teenage Bruno to Italy on a €1,500-a-month contract with no guarantees. She has had a say in every major career decision since. > *"I haven't fulfilled my dreams here. We still have dreams to fulfill."* ## [01:22:32] The Importance of Family For Bruno Bruno breaks down talking about his wife and their two children — a daughter born in Italy and a son born in England. He describes his wife as the second version of his father: she pushes him down when he gets too big, reminds him there is always something to improve, and rarely shows her feelings. His goal-celebration — covering his ears — was borrowed from his daughter, who used to do it as a young child. He also speaks about the structure Ineos has brought to the club: clearer lines of communication between players and ownership. He makes clear he wants Michael Carrick to be given time, because the one thing United has consistently failed to give its managers is stability. > *"They go through a lot — ups and downs, difficult moments — but they always stand by you. So that's the most important thing you can have in life."* ## [01:30:30] What Must Change for United to Compete for Titles Again Bruno names recruitment as the key variable for the summer. Casemiro's departure needs replacing, but the priority is not the most expensive name available — it's the right character. The model from the previous summer — Amad Diallo's breakout season, Patrick Dorgu's arrival — shows what happens when you recruit good professionals with good characters: the squad gets stronger without needing a superstar to paper over the cracks. ## [01:31:42] Bruno's Definition of Success Five Years From Now The closing question, left by the previous podcast guest: if five years from now everything has gone well, what happened? Bruno's answer: Premier League title, Champions League, and a World Cup with Portugal — in that order of emotion, if not difficulty. Winning with his club would be extraordinary. Winning for his country would be the biggest thing of his career, because it means representing his family, his nation, a small country that has conquered the world many times in different ways. > *"Representing my nation will always be the biggest achievement I have in my career — because not many players get to do that."* ## Entities - **Bruno Fernandes** (Person): Manchester United captain and Portugal international; 108 goals in 328 appearances for United since 2020; equalled the Premier League single-season assist record (20) this season; five-time Sir Matt Busby Player of the Year - **Steven Bartlett** (Person): Host of The Diary of a CEO; Manchester United fan; entrepreneur and investor - **Roy Keane** (Person): Former Manchester United captain and TV pundit; accused Bruno of chasing the assist record based on a quote Bruno says was the opposite of what he said - **Michael Carrick** (Person): Manchester United manager (confirmed permanent on the day of recording); former United midfielder under Sir Alex Ferguson; brought calmness and player autonomy to the dressing room - **Francesco Guidolin** (Person): Bruno's manager at Udinese at age 18; kept Bruno from being sent on loan to Watford; described as a father figure who gave Bruno the confidence to express himself at the top level - **Harry Maguire** (Person): Former Manchester United captain; Bruno went to speak with him before accepting the captaincy and says Maguire remains one of his key leaders in the dressing room - **Manchester United** (Organization): English Premier League club; Bruno joined in January 2020 and has remained captain despite multiple managerial changes and several large financial offers to leave - **Sporting CP** (Organization): Portuguese club where Bruno scored 20 goals and 13 assists in his final season; described as the period when he became the best version of himself as a player - **Ineos** (Organization): Investment group that took a stake in Manchester United; credited by Bruno with improving club structure and communication between players and ownership - **Risk-reward calculus** (Concept): Bruno's framework for decision-making on the pitch — a through-ball that fails twice but succeeds once to generate a goal is the correct play for a number ten - **Character over quality** (Concept): Bruno's central argument about United's recruitment failures — quality fluctuates season to season, character does not, so sign for character first

Dan Shipper, co-founder and CEO of Every, returns to lay out 12 contrarian predictions about AI and work — most of them pushback against prevailing panic. His core argument: automation doesn't shrink human workloads, it restructures them; Codex and Claude Code are becoming the new OS for knowledge work; the SaaS apocalypse is fiction; and the only survival skill you actually need is a willingness to ride the models as they improve. Every's 30-person company runs as a live experiment in this thesis, making Dan unusually well-positioned to say whether the predictions hold. ## [00:00] Introduction to Dan Shipper Lenny opens by recalling Dan's previous appearance, where he made an "almost offhand" prediction that people were sleeping on Claude Code for non-technical work — a call that proved "so unbelievably right." Dan's return centers on twelve more predictions, and he leads with the punchline immediately: > *"The AI job apocalypse is not really a thing."* ## [02:56] Dan's unique position living in the AI future Dan explains why Every functions as an early-signal lab: every employee — editors, ops, finance — is a daily AI user, which gives the company a running head start on what the next twelve months actually look like in practice. He contrasts this with the "San Francisco bubble" view, arguing that the real frontier of AI adoption is wherever AI meets a domain expert doing actual work, not where AI is being built. > *"The edge of AI is wherever AI meets like a real human doing something."* ## [09:17] How the way we work will change in the coming year Lenny frames three prediction buckets: how we work, the shape of work itself, and who thrives. Dan's opening call is that all professional work converges on one surface — either Codex or Claude Code — acting as a parallel work partner that watches what you're doing, handles research, writes emails, and kicks off long-running tasks while you stay in your primary document. He's already in inbox zero for ten days straight because Codex plus Cora (Every's email agent) handles his correspondence. > *"I basically feel like I have this parallel work buddy that not only can it respond and write in the document, but then it can go do research."* ## [16:39] The case for general agents Dan predicts every company will have one "super-agent" living inside Slack that all employees interact with daily — a general-purpose assistant with access to company context, not a narrow task bot. This agent becomes the organizational memory layer, routing questions, surfacing data, and bridging gaps between teams that don't know they need to talk to each other. ## [18:08] Codex and Claude Code as the new operating system for work Claude Code's breakthrough was putting a capable agent directly on your computer, giving it terminal access and — crucially — a browser. Anthropic figured out the paradigm first; OpenAI caught up around the 5.3 release and then accelerated. Dan's current daily driver is Codex, which he runs persistently alongside his Proof writing app — the agent watches his browser, reads whatever page he's on, and acts on his behalf without switching context. > *"Whoever is in the lead, it feels very obvious to me that all of the work that you do is going to be in one of those surfaces."* The model of "bring your own AI tokens to a SaaS app" reshapes economics: the SaaS product doesn't pay for inference, the user does, which restores margins and eliminates pressure to build a proprietary AI layer from scratch. ## [25:39] How Cursor fits in Cursor dominates coding workflows today, but Dan sees it at a strategic crossroads: stay purely a coding IDE or evolve into the general-purpose agentic surface. Staying narrow keeps the product focused; going broad means competing directly with Codex and Claude Code. His prediction is that the category winner will be the surface that handles both code and general knowledge work in one place. ## [27:42] How this changes what SaaS companies should build SaaS products now need to be agent-readable, not just human-readable — clean HTML, good CLI affordances, and design that surfaces information for automated consumption. Dan points to Proof: because Codex watches the page, paper cuts get fixed almost immediately, closing the loop between "I ran into something" and "it's resolved." > *"You can see the glimmers of this very fast closed loop between I ran into something, a paper cut, and I can just fix it right here."* ## [31:13] Why CLI is already over The CLI era was speed-run. The wave went: GUI, then CLI as a power move, then agents that replace the CLI entirely. Once your agent can operate any interface by reading the screen, the reason to live in the terminal disappears. Dan's prediction is blunt: > *"CLIs are over. We speed ran the CLI era."* ## [33:34] Two agents are better than one Dan pushes back against agent maximalism. The real pattern emerging is specialized agents — one for coding, one for email, one for data — that talk to each other on the user's behalf. When something breaks in an app, Codex can talk directly to the vendor's agent to diagnose the issue without a support ticket. The paradigm shifts once you assume everyone has an agent and agents can negotiate between themselves. ## [36:22] Why Dan is bullish on SaaS stocks The "SaaS is dead" narrative misses how the economics actually work when agents drive usage. When users bring their own AI tokens to a SaaS product, the vendor's inference costs drop toward zero. Dan's contrarian position: > *"I would buy SaaS stocks right now."* SaaS companies that make their products agent-friendly don't get disintermediated — they get a margin tailwind. ## [39:01] Why automation doesn't reduce human work This is the episode's central intellectual thesis. Dan argues that every automation layer requires a human manager above it to verify it's working correctly. He built his own benchmark — the "senior engineer benchmark" — by having two actual senior engineers independently rewrite his vibe-coded Proof app from first principles, then testing every new model against those reference solutions. Models scored 30/100 until GPT-5.5, which jumped to 60/100. The gap reveals something important: models fix what you tell them to fix. A senior human engineer looks at the codebase, decides it needs a full rewrite, and says so unprompted — models don't surface that judgment on their own. There is always a higher frame that requires a human to articulate. > *"Every time you automate something, in order to make sure the automation is working well, you need a human on top of it making sure that it's working well."* ## [47:00] The value of human-written code Human-written code still acts as the reference signal that lets you score model output. Dan's benchmark depends on two human-authored rewrites as ground truth. As AI-generated code becomes the default, the human-written corpus becomes scarcer and more valuable — the thing you need to know whether the AI is actually improving. ## [48:36] Quick recap Lenny summarizes the first prediction bucket: work happens inside Codex or Claude Code; every company gets a Slack super-agent; bring-your-own-tokens restores SaaS margins; CLIs are over; two specialized agents beat one generalist; automation expands human workload rather than shrinking it. ## [50:15] How work is changing The second bucket covers the shape of work itself. Dan's view: forward-deployed engineers become the most valuable hire — people who can sit with a customer, understand their workflow, and build and ship a fix in the same meeting. The "allocation economy" concept from his earlier essay applies here: humans become allocators of AI capability rather than direct producers, and allocating well turns out to be cognitively demanding in its own right. > *"I am simultaneously extremely AI-filled and very bullish on humans and the role of humans in making sure that AI is producing things that are worth producing."* ## [56:17] Why data scientists are drowning in bad analysis Data science teams are getting flooded with AI-generated analysis from everyone else in the company — analysis that looks plausible but is frequently wrong. The senior data scientist's job shifts from producing analysis to auditing it, which is harder and more cognitively demanding. The same dynamic hits engineering: junior-level requests get handled by models, surfacing more edge cases that require deeper judgment to resolve. > *"You need more senior people who are dealing with the deeper questions that are harder for the team who's dealing with all the basic requests."* ## [58:24] Which product/tech roles are least changed by AI Dan's answer: the roles whose output is hardest to frame as a prompt. He distinguishes between "babysitting agents" — passively watching for errors — and "forward-deployed engineering" — actively building systems that enable everyone else to do what used to require specialists. The second is where the interesting, hard-to-automate work lives. ## [62:17] We will read way more AI-generated writing and we will like it Every uses Notion agents for quarterly planning — each team's strategy report is AI-generated, and the output Dan gets back is better than what manual planning produced. His email is mostly written by GPT-5.5. His test for whether AI-written content is acceptable: did the sender have to understand what's in it in order to direct the AI? If yes, fine. If the sender clearly hasn't read it, that's a social contract violation. > *"The slop one is it took them less time to make it than it takes me to read it."* He also publishes Every guides written with agent co-authors, explicitly designed to be read by both humans and other agents — a new content format optimized for dual consumption. ## [68:28] Why product managers will dominate the AI era Dan cites Every's internal PM Marcus, who runs the Spiral product, as the archetype: strong product sense, able to direct AI to build and iterate quickly, ships without waiting for engineering bandwidth. PMs are fundamentally allocators — they decide what should be built and for whom — which is exactly the skill that remains scarce when the building itself becomes cheap. > *"I am super super bullish on PMs."* ## [71:05] Full-stack designers are the other big winners Full-stack designers — people with strong visual instincts who also operate in code — are already making pull requests directly in tools like Lovable and Figma Make. The handoff between design and engineering compresses toward zero. Dan expects them to become the go-to superheroes of the AI era alongside PMs. ## [73:11] The AI job apocalypse won't happen Dan separates the current round of layoffs (mostly over-hiring corrections) from a structural AI displacement claim, and rejects the latter. His structural argument: models are trained on yesterday's human competence, which means they produce what's already known in its most default form. Humans push the frontier by doing new things with that frozen competence, creating room that models then have to catch up to. The cycle repeats. > *"Structurally, because of the way the models work, there will always be room for humans to push further ahead."* ## [76:00] How to "ride the models" to stay relevant The actionable advice: don't resist new model releases — treat each one as a new set of powers to probe and apply to your actual domain. Dan re-runs his senior engineer benchmark every time a major model drops. He also pushes back on the idea that the edge of AI knowledge lives in San Francisco. Every, operating out of Brooklyn, stays ahead precisely because they use models for everything, not because they're building them. > *"The only thing you need to do is ride the models. And that means use them for whatever it is that you do."* ## [81:02] Final predictions and advice Lenny zooms out: the two sides of the coin from this conversation are "less is changing than you fear" (SaaS continues, jobs aren't disappearing) and "more is changing than you're prepared for" (how work gets done, which roles matter, what a workday looks like). Dan's closing call: forward-deployed engineer is the new essential hire; companies that block employees from using the latest models are making a slow-burn strategic mistake. ## [85:24] Lightning round Rapid-fire: Dan's most contrarian belief is that the AI job apocalypse genuinely isn't happening; the one thing he wishes more people understood is that the frontier of AI isn't in San Francisco — it's wherever someone is using a model to do real work in a real domain. He'd tell his past self to hire senior engineers earlier, and expects AI to fundamentally change how people think about benchmarks over the next year. ## Entities - **Dan Shipper** (Person): Co-founder and CEO of Every; author of the "After Automation" essay; runs Every as a live AI adoption lab - **Lenny Rachitsky** (Person): Host of Lenny's Podcast, founder of Lenny's Newsletter, ex-Airbnb PM - **Every** (Organization): 30-person AI-native media and software company; all employees are daily AI users - **Codex** (Software): OpenAI's agentic coding and general knowledge-work surface; Dan's current daily driver - **Claude Code** (Software): Anthropic's terminal-based coding agent; pioneered the on-computer agentic paradigm - **Proof** (Software): Dan's AI-assisted markdown writing app; the reference codebase for his senior engineer benchmark - **Cora** (Software): Every's email agent, integrated with Codex for inbox management - **Cursor** (Software): AI coding IDE at a strategic crossroads between coding tool and general agent surface - **Forward-deployed engineer** (Concept): A hybrid role combining engineering execution with customer-facing problem discovery; Dan's pick for most valuable new hire in the AI era - **Senior engineer benchmark** (Concept): Dan's custom evaluation where two human senior engineers rewrite a codebase from scratch; new models are scored against those reference solutions - **Allocation economy** (Concept): Dan's framework predicting humans shift from direct producers to allocators of AI capability - **Ride the models** (Concept): Dan's advice to stay relevant — treat each new model release as a new set of powers to actively probe and apply to your own domain

In this lightning episode, swyx sits down with Sunil Pai — Cloudflare developer platform lead and, according to swyx, creator of Code Mode — to cover three distinct threads: Cloudflare's infrastructure bet on Durable Objects and Dynamic Workers as the substrate for AI agents, the Twitter misunderstanding with Vercel that briefly convinced Sunil his career was over, and why forking code is an act of respect rather than aggression. Sunil closes with a direct challenge: stop building incremental agent frameworks and build science fiction instead. ## [00:00] Who invented Code Mode? The video opens on a three-second slate. What follows immediately — swyx introducing Sunil as "creator of Code Mode," Sunil accepting the credit with mock grandeur, claiming he has been thinking about it since childhood — is the opening exchange that this placeholder covers contextually. It is pure banter between two old friends, not a teaser pulled from later. ## [00:03] Introduction and Sunil Pai's background swyx reintroduces Sunil as an old friend and keynote speaker at AIE Europe. The brief catch-up frames what follows: Sunil's current focus is Cloudflare's platform for AI agents, and the recent Anthropic Cloud Managed Agents launch gives him a concrete foil to argue against. > *"I wanted to just catch up on everything going on in Cloudflare lands."* ## [00:30] Discussing the new cloud-managed agents Anthropic's newly launched Cloud Managed Agents product — a platform for building and deploying long-running agents — is Sunil's jumping-off point. He says he likes the Anthropic team and finds the product interesting, but his reaction on reading the spec was competitive: Cloudflare can do this better. swyx asks what Cloudflare actually has that makes that claim credible. > *"I looked at the product and I was like I think I want to compete. I think we can do something better with Workers and Durable Objects."* ## [01:10] Cloudflare's core infrastructure: Durable Objects and Dynamic Workers Sunil names two primitives he believes every agent platform will eventually need. Durable Objects are stateful serverless units — his claim is that they are the world's first infrastructure-layer implementation of the actor model rather than a user-land library. Dynamic Workers are Cloudflare's answer to running LLM-generated code safely: eval re-imagined with zero startup time, configurable API surface, and outgoing traffic locked down by default. Together they let Cloudflare run agent steps in sandboxed compute without spinning up full VMs. > *"It's the world's first implementation of the actor model in an infrastructure layer, not in user land."* ## [02:34] How Cloudflare approaches AI agent architecture The Cloudflare MCP server, built by colleague Matt Carey, shows Dynamic Workers in practice. The Cloudflare API has 2,600 endpoints — exposing one tool per endpoint would destroy any LLM context window. Instead, the server collapses everything into two tool calls: `search` and `execute`, both backed by JavaScript code running in an isolate. The agent submits code, the isolate runs it, the result comes back — no back-and-forth, type-checked. > *"In one tool call, no back and forth with the LLM, and it's type checked, and well, turns out LLMs are great at running code."* ## [03:40] The future of agentic software and standardizing the "harness" swyx asks whether the harness concept from Anthropic's spec could become a cross-platform standard. Sunil's answer: nobody has built the React of AI agents yet. He draws the 2013 React analogy deliberately — people walked out of the JSConf talk, accused Facebook of hating JavaScript, and yet React defined every UI framework that followed. Right now everyone is building their own harness in their own shape, and nothing is reproducible across languages, companies, and infrastructure. swyx floats the idea that skills — plain markdown — might already be that unifying layer; Sunil finds the idea genuinely appealing but worries about the specificity ceiling. > *"It's so hard, but the way I'm framing it in my head is no one has built the React yet."* ## [06:11] The "slop forks" phenomenon and open-source culture swyx raises "slop forks" — AI-generated forks of popular projects — and Sunil lights up. In his framing, forking is a gesture of prestige and respect, not theft. The React ecosystem grew through forks. He tells anyone interested in building something competitive with the Cloudflare Agents SDK to go for it: everyone wins if they do. > *"Forking is a great sign of prestige respect in my culture."* ## [06:36] The Vercel / Cloudflare social media misunderstanding At JSConf España, Sunil met Harvey from Vercel and loved spending time with him. He found Vercel Labs' Just Bash — a pure JavaScript implementation of Bash — and wanted to port it to Cloudflare. He pointed Opus at the codebase over lunch, got 5,000 lines of code back, and planned to clean it up before sending a proper PR on Monday. He crashed, woke up to DMs from Cloudflare management asking if he had seen Twitter: the Vercel CTO had publicly criticised the work, framing it as a corporate move rather than a personal side project. Sunil responded plainly, explained the context, and then watched half the internet rush to defend him. > *"I go on Twitter and the Vercel CTO is trashing my work saying… 'It's Cloudflare did this.'"* ## [09:45] The importance of forking in software development swyx connects the Vercel incident to a broader pattern: a leaked codebase someone rewrote in Python to escape the license (lawyers ruled it a derivative work anyway). The real argument swyx makes is that slop forks are worth encouraging — fork a dependency, vendor it, own it — so you avoid the sudden upstream breakage of the LiteLLM or Axios problem. Sunil agrees: before NPM, software spread on Usenet through exactly this pattern, and shortening the fork cycle is just that tradition continuing. > *"Forking is so fundamental to how we build software."* ## [12:04] The adversarial nature of modern open-source repositories The Cloudflare Agents SDK has had to shut down pull-request contributions entirely; only issues are allowed now. Sunil talks to open-source maintainers at the conference who describe the same thing: repos have become adversarial territory, and the worst attack vector is fake security reports that look entirely legitimate until you read them carefully. swyx ties this to a morning talk by Peter from Claude Code — the number one current attack surface is a compromised dependency getting into Claude Code, which would give access to every developer using it. > *"Open source repos have become adversarial to the point that people are almost afraid of gaining popularity in that space."* ## [13:04] Closing thoughts and encouragement to be original Sunil's closing ask is direct: stop building the tenth agent framework. Build science fiction. Build something for your family. Use the Agent SDK, but use it for something where the infrastructure and the LLMs almost fail you — because that's where the next step change lives. swyx closes with a callback to Sunil's 2018 React Rally coinage of "alpha thought leading." > *"Build sci-fi stuff. Build stuff like for your family. You own so much agency in changing the world and I want people to just be original."* ## Entities - **swyx** (Person): Host of Latent Space; long-time friend of Sunil; coined "alpha thought leading" after a Sunil quip at React Rally 2018. - **Sunil Pai** (Person): Developer platform lead at Cloudflare; credited by swyx as creator of Code Mode; keynote speaker at AIE Europe. - **Cloudflare** (Organization): Cloud platform company; building agent infrastructure on Durable Objects and Dynamic Workers. - **Anthropic** (Organization): AI company; launched Cloud Managed Agents, the product Sunil positions Cloudflare to compete with. - **Vercel** (Organization): Frontend cloud company; Sunil uses their AI SDK; subject of the Twitter misunderstanding. - **Durable Objects** (Software): Cloudflare's stateful serverless primitive; Sunil's claim is it is the world's first infrastructure-layer actor-model implementation. - **Dynamic Workers** (Software): Cloudflare feature for running LLM- or user-generated JavaScript in a safe, zero-cold-start isolate. - **Just Bash** (Software): Vercel Labs project — a pure JavaScript implementation of Bash — that Sunil was porting to Cloudflare when the Twitter incident occurred. - **MCP** (Concept): Model Context Protocol; Cloudflare's MCP server collapses 2,600 API endpoints into two tool calls using Dynamic Workers. - **Slop forks** (Concept): AI-generated forks of existing projects; Sunil frames them as continuation of open-source forking culture — a sign of respect, not plagiarism.