PodcastsHear the voice. See the shape of the thought.

GitHub COO Kyle Daigle joins swyx to map what the agent era looks like from inside the platform hosting 200 million developers and now processing commits at 14x last year's pace. Across 84 minutes they cover how Kyle runs GitHub with AI-driven micro-skills and WorkIQ MCP, why former developers in leadership have an unusual edge right now, the full arc of GitHub's platform history from webhooks to Actions to Copilot, and where trust in agent-generated code ultimately has to come from. The conversation is grounded throughout in Kyle's own weekend and executive workflows: building AI-generated revenue presentations, running 15 simultaneous agents on a Saturday, and describing what "ambient AI" would actually need to do before it becomes genuinely useful. ## [00:00] Hook Kyle opens mid-sentence, already deep in his argument: people who detoured into other careers before coding, and came back armed with cross-domain knowledge, are uniquely positioned in the AI era. Running 15 agents on a Saturday while his kids are at lacrosse is not just a productivity flex — it recreates the feeling of creation that got him into software in the first place. > *"I can crank up 15 agents on Saturday, you know, while my kids are doing lacrosse. That's like really powerful and I think it gets me back to that feeling of like creation."* ## [01:21] Introduction Kyle's title is COO of GitHub, but he recently took on CMO of Developer for Microsoft as well — meaning every developer-facing product and communication across the broader Microsoft ecosystem now runs through him. He's been at GitHub for 13 years, joined as a developer, personally built webhooks and the platform/API layer, ran engineering until 2018, then moved into the operational and business side. The dual COO/CMO role is unusual; Kyle frames it as the same job with a larger surface area: tell the truth, be authentic, let the products speak. > *"I built webhooks and worked with teams building the API, built the platform layer, anything that integrated with GitHub, up until really 2018 I built or ran the engineering teams."* ## [04:57] Why AI Got Kyle Coding Again Swyx points out that Kyle's commit graph shows a clear dip through his leadership years and a sharp uptick recently — entirely driven by AI. Kyle is not writing features for GitHub's product; he's building internal agents and workflow tools that stitch together disparate data sources. His primary use case is retrospective: using WorkIQ, MCP servers, Slack, Teams transcripts, and Obsidian notes to ask "what actually happened last week, what worked, and what should I tweak for the next few days." He finds LLMs are exceptionally good at pattern-finding across a week of context, far more so than generating forward-looking plans from scratch. > *"I find AI in like what most of this launch here is actually like less building forward. It's actually like a recursive loop backwards. I'm always looking at what had happened first."* ## [08:25] Running GitHub with AI: WorkIQ, MCP, Slack, Teams, and Skills GitHub rolled out AI internally by meeting people where they already work — Slack, Teams, email — rather than forcing them onto a new tool. Every employee, technical or not, gets the Copilot CLI plus a shared set of atomic micro-skills deposited into repos. The era of the "mega-skill" that handles an entire workflow end-to-end is over; what works are tiny, single-purpose skills that do one thing well and compose cleanly. Kyle uses Postel's Law as a design principle: liberal in what each skill accepts, strict in what it outputs. WorkIQ, the M365 MCP server, lets anyone ask backward-facing questions across every meeting, email, and chat — critical for a fully remote, globally distributed team. > *"We're ending the era of these like massive beautiful perfect skills. What we found is these incredibly micro skills that are just doing one thing for us very very well versus a skill that's going to do that full report that doesn't really exist on our side anymore."* ## [17:00] The Golden Age for Former Developers in Leadership Swyx asks whether people like Kyle — technical backgrounds, now in exec roles — have a structural advantage in the AI era. Kyle's answer: pattern-finding and problem-solving are the durable skills from his developer years, and AI has given him back the ability to apply them directly in code. The more interesting case isn't developers going back to update old side projects; it's people who spent ten-plus years accumulating business knowledge now using that context as leverage when wielding AI tools. The cross-domain background, once a liability in pure engineering orgs, is now a multiplier. > *"I just find that the folks that came from a different career, went to school for something else, went off and did this random thing and then became a software dev — now having the power of an AI where I can crank up 15 agents on Saturday."* ## [18:52] 15 Agents on Saturday and AI-Generated Executive Work Kyle built GitHub's annual revenue planning presentation entirely with AI — a SQLite app to view the data, skills pulling from Obsidian notes and work context, and a deliberate skill that made the output look "humanly bad" so it wouldn't read as AI-generated. He presented it to the CRO and CFO teams without disclosing the process; nobody asked. His point isn't to hide AI from colleagues but to demonstrate that value is in crafting and judgment, not slide assembly. The ability to build a small data-manipulation app and control the final output is, specifically, the advantage that developers carry into leadership. > *"I ultimately built this entire presentation without touching any of it. And I was like, okay, I'm just going to present this to our CRO, the CFO, their teams without mentioning I built it with AI. Never came up once."* ## [21:41] How AI Changes the Chief of Staff Role Kyle still has a chief of staff — but the job has shifted. Slide prep and presentation assembly have moved to AI; what remains irreplaceable is the human connective tissue: knowing which people in which cities should meet, surfacing relationship opportunities across a distributed org, brokering conversations that don't appear in any MCP server. The analogy is email replacing letter-opening: nobody expects the chief of staff to open physical mail anymore, and soon nobody will expect them to build decks either. The judgment about *who* should talk to *whom* is what stays. > *"I still have a chief of staff because the difference is the human connection aspects — I should be meeting with this group and this team and they have an opportunity and I'm going to be in San Francisco today."* ## [23:06] GitHub's History: Actions, npm, Webhooks, and Open Source Kyle walked the platform's architectural history: GitHub Services (pre-2014 arbitrary Ruby execution with no real containerization), webhooks, Pages, and then Actions — launched by Kyle personally at GitHub Universe in October 2018. Actions went from "we should not be running arbitrary Ruby on people's behalf" to a fully containerized compute layer now using Azure Dev Compute for fast, small-VM agent spin-ups. The npm acquisition came from a simple premise: npm was powering the internet and having scaling problems; GitHub's job was to keep it running and raise its security posture. Every security improvement — 2FA enforcement, token invalidation on exposure — breaks something downstream, and that balance between hardening a 15-year-old ecosystem and not causing developer snow days remains the central tension. > *"We have changed the 2FA policies, we've changed the way the tokens work. When we find tokens that have been exposed or potentially exposed, we invalidate them. That creates issues. But we're trying to push the community forward."* ## [30:06] Slop Forks, Vendoring, and AI Dependency Management Swyx raises the "slop fork" pattern — AI-assisted vendoring where you pull in only the source you need rather than importing a whole package — and asks whether it sidesteps npm's vulnerability surface. Kyle: vendoring was how everyone worked in 2013, and there's something true about pulling in only what you need, but it doesn't fix the fundamental problem. An agent evaluating code can be convinced it's secure just as easily as a human can. Static analysis and runtime testing still need investment regardless of package scope. GitHub's historical stance — wait for community RFC and social consensus before cementing a practice — means they won't push a single vendoring standard, but will build tools for maintainers to enforce their own trust rules. > *"The vulnerabilities — in an agent looking at them there's time and time again a million different ways in which we can convince an agent that this thing is like secure or not."* ## [35:18] Pull Requests, Prompt Requests, and Trust in Agent-Generated Code GitHub invented the pull request as a social trust mechanism, and now agents are generating the majority of PRs on many projects. Kyle assessed various alternatives — Peter Coppola's "prompt request" model, Thomas Dohmke's contribution-asset approach — but argues that none fully solve the underlying problem: trust is social, not technical. Even if a PR is 100% verified by static analysis, humans still reach for human signals (does Mitchell approve it?) before merging. GitHub's current direction centers on giving maintainers malleable tools to define their own trust heuristics rather than imposing a universal standard, because any single standard immediately becomes a gamification target. The endgame is something closer to human digital identity. > *"The reason why there's not a single answer is ultimately we're trying to codify trust. Right now when an agent writes code and another agent reviews code and then Kyle goes and looks at it, the trust is kind of diffuse."* ## [42:42] GitHub Stars, 200M+ Developers, and the New AI Builder Wave GitHub crossed 200 million accounts — up from 80 million not long ago. The rapid star accumulation on new AI projects is mostly genuine: an entire new cohort who built their first app in the AI era is swarming the zeitgeist. Kyle refuses to split hairs about who "counts" as a developer, drawing on his own experience being called a fraud for having a GitHub account before he knew what git was. The gamification problem is real (whack-a-mole anti-abuse, now AI-powered), but the majority of the star velocity is new builders who want to participate in the moment the way Kyle wanted to participate in the Ruby era. > *"It's not just developers. It's folks that have maybe started coding or only joined in since the AI era. And those projects are going up because you want to be a part of this moment."* ## [46:36] GitHub Spark, Low-Code, and Why GitHub Still Shows the Code GitHub experimented with Spark as an easy app-build-and-run experience. The lesson: for developers, the value was always simple runtime, not a UI veneer hiding the code. GitHub's architectural principle is non-negotiable — they will always show you the code. The broader goal Kyle articulates is lowering the barrier to that first "I had an idea and I built it" moment: anyone should be able to swap a light switch without needing to open the breaker box. > *"Anytime we try to put a veneer on top of something, we still always show you the code. That's kind of like a tenant. We're never gonna hide the code from you ever."* ## [48:59] GitHub's Hardest Era: 14x Growth, Reliability, and Scale GitHub went from 1 billion commits in all of 2025 to 275 million per week in April 2026 — a 14x year-on-year rate still accelerating. This broke things in new ways: not the old webhooks reliability problems (those were fixed and rewrote), but novel permission-layer failures only visible at cross-object scale. The core pain point is MySQL 1, a monolithic permissions database GitHub has been decomposing for years; permissioning is where most cross-cutting outages originate. Simultaneously, the industry is shifting back toward monorepos, which carry unique git infrastructure performance characteristics. Kyle frames the scaling problem as "diagonal" — vertical and horizontal both stop working, so you crack open services running unchanged for 10-15 years and rewrite them. > *"We're doing more in a month than we did in a year last year. By roughly every measure, there's growth that is much much bigger. And that is breaking our system in new ways, not old ways."* ## [60:42] Actions as the Compute Layer for CI/CD and Automation Actions has evolved well beyond CI/CD into a general-purpose automation compute layer — the root of significant availability pressure because every agent task and agentic workflow translates into more builds and more CPU. GitHub is expanding compute through both its own data centers and Azure cloud, and is using Azure Dev Compute (fast small-VM spin-up) under the hood for containerized agent execution. The path to fewer outages is a step-change model: large foundational infrastructure fixes that take time, then visible plateau improvements in availability rather than incremental noise reduction. > *"Actions is the core compute layer for either CI or side project. More tools, more agents, more PRs mean more builds. More builds need more CPUs and we simply need more CPUs."* ## [63:25] The State and Future of GitHub Copilot Copilot's history: launched as code completion, then shifted energy toward fine-tuning as the industry demanded better accuracy, and then next-gen models arrived and made fine-tuning less critical — creating confusion about where Copilot was going. The current architecture unifies a single SDK and agent harness across code completion, the new CLI, the new desktop app, and cloud agents. The future Kyle describes covers the full SDLC: security remediation, issue triage, documentation drift detection — not just writing code. The remaining hard problem is context and memory: getting GitHub to "act like Kyle wants it to act" across all his dependencies, preferences, and team context. > *"What we think is that it's not solely about the code generation. It's really about having the ability to use these coding agent brained harnesses across not just the coding experience but also security remediation, every GitHub issue that comes in."* ## [69:45] Ambient AI, Background Agents, and the Future of the SDLC Kyle argues the industry is still stuck in a "hyper-myopic" frame where coding agents only know about code. What he actually wants is ambient AI that carries every spec doc, every email thread, every conversation, every Obsidian note into its decision-making as a developer — not as a recall tool you query, but as persistent background context that shapes implementation choices in real time. OpenClaw interests him precisely because it connects personal context to agent action; but the missing piece is making that context available *during* software development. The extreme version — AI that proactively directs you rather than waiting to be asked — is the inversion of control that both excites and slightly alarms him. > *"The most interesting thing to me in AI is actual ambient AI. I'm looking to be implementing a new feature and for it to know every spec doc, every email, the conversations that I've had online, everything about how this could be implemented and be able to use that as part of its decision-making."* ## [74:30] OpenClaw, Enterprise Security, and the New OS for Agents Microsoft has a CVP dedicated to OpenClaw — unusual given Microsoft doesn't own Anthropic. Kyle explains: OpenClaw demonstrated what a valuable personal agent actually looks like (full personal context, computer use, not just chat), and Microsoft's job is to make that work in enterprise — OS-level sandboxing on Windows so you can run an agent on a work device without it becoming a security incident. The framing Kyle reaches for: Microsoft is the original operating systems company, and agents need a new OS layer. Workloads have changed so fundamentally that the right question is no longer "do we need more inference?" but "what type of compute do we need to run these agentic flows?" — all the way down to silicon. > *"Microsoft is the original operating systems company and here's the new operating system for AI. Operating systems need to look different than they looked five years ago because it's not just you using them anymore."* ## [79:24] Build Announcements, WorkIQ, FoundryIQ, and Microsoft Context Kyle previews what GitHub and Microsoft are announcing at Build: WorkIQ (M365 context engine via MCP, powerful for retrospective questioning across all work assets) and FoundryIQ (same intelligence layer that connects to existing data stores without requiring migration). The pitch for enterprise developers: "how I build on the weekend should be how I build at work" — but Fortune 500 companies can't just vibe-code and ship; security and compliance gates have to move as fast as development does. WorkIQ and FoundryIQ are the attempt to bring weekend-level agility into the enterprise context layer, with the governance that lets it survive in large organizations. > *"Work IQ, Foundry IQ — these context engines are wild good and we've given them to our developers at GitHub. You can ask questions around everything in your work context and it's surprisingly powerful."* ## [83:02] What Should swyx Ask Satya? swyx is about to interview Satya Nadella at Build and asks Kyle what to ask. Kyle's recommendation: challenge Satya on what he believes is demonstrably true about the AI and inference landscape in two to three years — not as a throwaway futurist question, but as a direct test of the internal bets Microsoft is making right now. Significant external skepticism exists about Microsoft's AI approach, and a straight answer from Satya would be both a genuine stress test and a reassuring signal for the developer community. > *"The best question to ask is what he thinks is true in like two or three years from now. The way that he is looking at this AI problem, the inference problem, the token problem — why is this approach in two years going to pay off?"* ## Entities - **Kyle Daigle** (Person): COO of GitHub and CMO of Developer for Microsoft; 13-year GitHub veteran who built the original webhooks and platform API layer. - **swyx** (Person): Host of Latent Space podcast; developer-advocate-turned-podcaster who conducted this interview at Microsoft Build 2026. - **GitHub Copilot** (Software): GitHub's AI coding assistant, now spanning code completion, CLI, desktop app, and cloud agents under a unified SDK. - **WorkIQ** (Software): Microsoft 365 MCP server that gives employees a context engine over all work assets (Teams, email, calendar, etc.). - **FoundryIQ** (Software): M365 intelligence layer that connects to existing enterprise data stores without requiring migration. - **GitHub Actions** (Software): GitHub's general-purpose compute and CI/CD automation layer; primary source of CPU demand growth from agent workloads. - **OpenClaw** (Software): Anthropic's Claude Code agentic tool; referenced as a model for what a personal AI agent with full context and computer use looks like. - **npm** (Software): JavaScript package registry acquired by GitHub; central to supply-chain security discussions about vendoring, slop forks, and dependency trust. - **Mitch Hashimoto** (Person): Co-founder of HashiCorp, active open-source maintainer; discussed in context of vendoring approaches and GitHub's maintainer relationship model. - **Thomas Dohmke** (Person): CEO of GitHub; referenced in context of PR workflow evolution. - **Microsoft Build** (Organization): Annual Microsoft developer conference; context for this episode's release and Kyle's expanded-role announcements.

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

Alex Rives — director de Ciencia en BioHub y el investigador que lideró ESM-1 hasta ESM-3 en Meta FAIR — se une a Brandon y RJ Honicky para explicar por qué lleva ocho años apostando a que escalar un modelo de lenguaje enmascarado sobre secuencias de proteínas desbloquearía estructura, función y diseño biológico. El episodio cubre el cambio de datos de UniRef a metagenómica que restauró la ley de escala de ESMC, el atlas de características con sparse autoencoders que reproduce un siglo de taxonomía bioquímica sin haber sido entrenado con ella, y el primer éxito reportado en el diseño de anticuerpos de cadena única con afinidades terapéuticas mediante búsqueda en un modelo de mundo. Rives también presenta la Iniciativa de Biología Virtual de $500 millones de BioHub y los principios que, según él, producirán modelos generalistas de la célula. ## [00:00] ESMC diseña anticuerpos — una vista previa Este fragmento inicial está tomado de más adelante en la entrevista, con Rives a media frase describiendo el enfoque de ESMC hacia la biología programable. Describe la búsqueda en un modelo de mundo de proteínas para satisfacer criterios de diseño, y menciona que el equipo ha diseñado mini-binders y, sobre todo, fragmentos de anticuerpos de cadena única (SCFVs) con afinidades de unión relevantes para uso terapéutico. El clip precede a la presentación formal — una señal de lo que el episodio va construyendo. ## [00:33] La Amarga Lección llega a las proteínas Brandon y RJ presentan a Alex como posiblemente "la persona más convencida de la Amarga Lección en biología de proteínas ahora mismo." Rives acepta la etiqueta. Traza su convicción hasta 2018, cuando su equipo en Meta FAIR entrenó el primer modelo de lenguaje transformer sobre secuencias de proteínas con predicción de tokens enmascarados y vio emerger representaciones estructurales y funcionales sin ninguna supervisión explícita. La intuición clave, tomada del artículo de Zellig Harris de 1954 sobre estructura distribucional, es que los contextos en que puede aparecer un aminoácido están determinados por la estructura, función y papel evolutivo de la proteína. Esa presión estadística, aplicada sobre miles de millones de secuencias de toda la vida, debería forzar al modelo a aprender las variables ocultas que gobiernan la biología de proteínas. > *"Creo en las leyes de escala."* ## [06:00] El linaje ESM: de ESM2 a ESMC Rives repasa cuatro generaciones de ESM. ESM2 mostró mejoras con el escalado pero alcanzó rendimientos decrecientes en 10B parámetros — no porque el modelo estuviera saturado, sino porque los datos lo estaban. UniRef, la base de datos de proteínas de referencia, captura organismos cultivados y se sesga fuertemente hacia la biología relevante para el ser humano. La solución para ESMC fue la metagenómica: secuencias extraídas de chimeneas hidrotermales, suelos polares y aguas residuales, ensambladas a partir de lecturas de ADN ambiental sin asignación de organismo, contigs parciales incluidos. Añadir miles de millones de secuencias metagenómicas al entrenamiento restauró una ley de escala log-lineal limpia, con ejecuciones a escala menor que predecían con precisión la fidelidad representacional del modelo insignia de 6B parámetros. > *"Ya no hay rendimientos decrecientes al escalar. ESM2 estaba limitado por los datos, no por el cómputo."* ESMC es esencialmente un transformer estándar con objetivos de enmascaramiento convencionales — sin MSA al estilo AlphaFold, sin sesgos inductivos geométricos. Brandon y Rives debaten brevemente si la arquitectura multi-track de ESM3 fue un desvío productivo; Rives dice que ambos paradigmas tienen su lugar, pero el resultado de ESMC sugiere que los priors no eran fundamentales a esta escala de datos. ## [18:30] Interpretabilidad mecanicista y el atlas de características de proteínas Usando sparse autoencoders entrenados en todas las capas de la familia de modelos ESMC (300M, 600M, 6B), el equipo de BioHub extrajo la geometría intrínseca de características del espacio de representación de proteínas. Lo que emergió se corresponde estrechamente con la jerarquía reductiva que la biología desarrolló experimentalmente a lo largo de un siglo — desde la química básica de aminoácidos hasta motivos estructurales, familias de dominios y grandes temas funcionales — sin que ninguna de esa taxonomía se introdujera durante el entrenamiento. > *"La elección de cualquier aminoácido está completamente entrelazada con la elección de todos los demás aminoácidos de la secuencia. Para hacerlo bien, el modelo empezaría a tener estas variables ocultas que representan la biología."* Un hallazgo concreto: el modelo codifica el codo nucleofílico — un motivo catalítico que se cree evolucionó de forma independiente en varias familias de proteínas no relacionadas — como una sola característica que se activa en todas ellas. El equipo también construyó un atlas estructural de 6.800 millones de proteínas no redundantes con estructuras predichas para 1.100 millones de representantes de clústeres, y usó características de SAE para conectar sistemas de edición genética evolutivamente distantes. Algunas proteínas incluidas en esos clústeres no tienen función conocida; Rives las trata como una cola de descubrimiento. La primera versión del atlas ESM ya fue utilizada por un grupo externo para encontrar un nuevo sistema de edición genética. ## [35:30] Diseño de anticuerpos con ESMC Rives describe el diseño de proteínas como búsqueda en un modelo de mundo: invertir el modelo generativo para encontrar secuencias que satisfagan criterios de unión objetivo. Los mini-binders son ya rutinarios; los nanobodies y SCFVs siguen siendo más difíciles para los métodos basados en predicción de estructura porque la evolución de anticuerpos maximiza la diversidad en lugar de converger hacia un pliegue restringido, lo que hace que los enfoques basados en MSA sean menos útiles. ESMC, entrenado sobre esa diversidad a escala, es precisamente donde la representación debería ser más rica. > *"Los anticuerpos no van a beneficiarse de la información evolutiva probablemente de la misma manera que la predicción de la topología estructural de una molécula."* El equipo reporta diseños de SCFV que alcanzan afinidad de grado terapéutico en un número reducido de ensayos, y señala que los SCFVs pueden reformatearse como IgGs completas. ESMFold 2 — el cabezal de predicción de estructura construido sobre las representaciones de ESMC — se ejecuta en segundos por secuencia sin MSA, haciendo factible el mapeo de multímeros a escala de proteoma completo. Rives afirma que el modelo es actualmente el estado del arte en predicción de multímeros con pesos abiertos. ## [42:00] La visión de BioHub: hacia la biología programable A seis meses de incorporarse a BioHub, Rives articula la estructura de la institución: una filantropía que construye biología experimental de frontera, tecnología de medición de frontera e IA de frontera bajo un mandato de ciencia abierta. Sitúa el destino en modelos predictivos personalizados de la fisiología — no un fármaco, sino un sistema capaz de rastrear eventos moleculares al nivel de proteínas hacia arriba a través de circuitos celulares hasta la manifestación de enfermedades en un genoma humano específico. > *"Estamos construyendo una institución científica para este nuevo paradigma."* Traza los niveles de complejidad biológica que deben modelarse en secuencia: proteínas (generación actual), la célula (la siguiente), tejidos y sistemas, fisiología. Pasar de proteínas a células requiere datos que aún no existen y enfoques de modelado que probablemente no se han inventado. Los modelos de "célula virtual" actuales generalizan mal — representan bien los datos de entrenamiento pero fallan al predecir resultados en contextos de intervención novedosos no observados. > *"Tienen una capacidad muy limitada para predecir qué ocurrirá cuando se haga una intervención novedosa en un contexto novedoso no observado."* ## [57:00] La Iniciativa de Biología Virtual y el escalado de datos celulares BioHub anunció recientemente $400M para generación de datos interna y tecnología de medición, más $100M para catalizar esfuerzos externos — juntos, la Iniciativa de Biología Virtual. Rives enmarca esto como financiación semilla: el volumen real de datos necesario es mucho mayor, y la esperanza es que el compromiso de BioHub desencadene una inversión más amplia de la comunidad científica. Identifica tres principios sobre los datos: velocidad (los datos de proteínas tardaron medio siglo; la célula no puede esperar tanto), generalización (la distribución de entrenamiento debe abarcar una vasta diversidad de intervenciones entre tipos celulares y contextos, análogo a la amplitud metagenómica para proteínas), y retroalimentación (bucles experimentales activos guiados por predicciones del modelo — algo como RLVR aplicado a biología en laboratorio húmedo). La secuenciación por perturbación, la transcriptómica espacial y la medición de célula única cruzando modalidades son las tecnologías escalables listas para funcionar ya. Sobre cómputo: ESMC se entrenó con aproximadamente mil millones de secuencias. Se calcula que existen unos 100.000 millones, y el modelo aún no ha explotado del todo ni siquiera los 6.800 millones del atlas actual. Un aumento de 100x en cómputo ayudaría, pero solo si va acompañado de un escalado proporcional de datos. Rives deja empíricamente abierta la pregunta de cuándo aparecerán los rendimientos decrecientes — la curva de ESM2 parecía saturada justo hasta que los datos metagenómicos la borraron. > *"Necesitamos descubrir cómo hacer esto en un par de años. El ritmo al que avanza la IA general significa que la biología estará fundamentalmente limitada por la ciencia experimental y los datos."* ## Entidades - **Alex Rives** (Persona): Director de Ciencia en BioHub; arquitecto de ESM-1, ESM-2, ESM-3, ESMC y ESMFold 2; anteriormente en Meta FAIR. - **Brandon** (Persona): Co-presentador de la subserie Latent Space AI for Science; vinculado a Atomic AI (terapéutica de ARN). - **RJ Honicky** (Persona): Co-presentador; CTO y fundador de Miro Omix. - **ESMC** (Software): Modelo de lenguaje de proteínas de cuarta generación de BioHub/EvoScale; 300M–6B parámetros; entrenado con ~1.000 millones de secuencias incluyendo datos metagenómicos; código abierto bajo licencia MIT. - **ESMFold 2** (Software): Modelo de predicción de estructura construido sobre las representaciones de ESMC; sin MSA, inferencia en segundos por secuencia; estado del arte en predicción de multímeros con pesos abiertos. - **ESM** (Software): Evolutionary Scale Modeling — el linaje multigeneracional de modelos de lenguaje de proteínas (ESM-1, ESM-2, ESM-3, ESMC) pionero del equipo de Rives. - **Sparse Autoencoders / SAEs** (Concepto): Herramienta de interpretabilidad mecanicista usada para extraer la geometría intrínseca de características del espacio de representación de ESMC; revela jerarquías biológicamente interpretables sin supervisión. - **Bitter Lesson** (Concepto): Argumento de Richard Sutton de que los métodos generales que aprovechan cómputo y datos superan sistemáticamente a los métodos que codifican conocimiento de dominio; aplicado aquí al escalado de biología de proteínas. - **Secuenciación metagenómica** (Concepto): Secuenciación de ADN ambiental que captura la diversidad microbiana y viral sin necesidad de cultivo; la expansión de datos que restauró la ley de escala de ESMC donde UniRef se había saturado. - **BioHub** (Organización): Chan Zuckerberg BioHub; una filantropía que construye herramientas de ciencia abierta en la intersección de biología experimental, tecnología de medición e IA. - **Iniciativa de Biología Virtual** (Concepto): Compromiso de $500M de BioHub ($400M interno, $100M externo) para generar los datos a escala celular necesarios para entrenar modelos generalistas de la célula. - **AlphaFold** (Software): Sistema de predicción de estructura de DeepMind; usa MSAs y sesgos inductivos geométricos; contrastado con el enfoque sin MSA de ESMC. - **UniRef** (Software/Base de datos): Base de datos de secuencias de proteínas curada de referencia; datos de entrenamiento de ESM2, posteriormente identificada como el cuello de botella que causó la meseta de escalado de ESM2. - **Codo nucleofílico** (Concepto): Motivo estructural catalítico presente en múltiples familias de proteínas evolutivamente no relacionadas; codificado como una sola característica de ESMC que se activa en todas ellas. - **Zellig Harris** (Persona): Lingüista; su artículo de 1954 "Distributional Structure" articuló que los contextos de las palabras codifican significado — precursor teórico que Rives cita para justificar por qué las estadísticas de contexto de aminoácidos deberían codificar función biológica.

En este episodio relámpago, swyx conversa con Sunil Pai — responsable de la plataforma para desarrolladores de Cloudflare y, según swyx, creador de Code Mode — para abordar tres hilos: la apuesta de Cloudflare por Durable Objects y Dynamic Workers como sustrato para agentes de IA, el malentendido en Twitter con Vercel que estuvo a punto de costarle la carrera a Sunil, y por qué hacer un fork de código es un acto de respeto y no de agresión. Sunil cierra con un reto directo: deja de construir otro framework de agentes incremental y construye ciencia ficción. ## [00:00] ¿Quién inventó Code Mode? El vídeo abre con una pizarra de tres segundos. Lo que sigue de inmediato — swyx presentando a Sunil como "creador de Code Mode", Sunil aceptando el crédito con grandilocuencia fingida y afirmando que lleva pensando en ello desde la infancia — es el intercambio de bienvenida que este marcador cubre contextualmente. Pura complicidad entre dos viejos amigos, sin teaser tomado de más adelante. ## [00:03] Introducción y trayectoria de Sunil Pai swyx reintroduce a Sunil como un viejo amigo y ponente principal en AIE Europe. El breve reencuentro sitúa lo que viene: el foco actual de Sunil es la plataforma de Cloudflare para agentes de IA, y el reciente lanzamiento de Cloud Managed Agents de Anthropic le da un punto de contraste concreto al que oponerse. > *"Quería simplemente ponerme al día con todo lo que está pasando en el mundo de Cloudflare."* ## [00:30] Los nuevos agentes gestionados en la nube El producto Cloud Managed Agents que Anthropic acaba de lanzar — una plataforma para construir y desplegar agentes de larga duración — es el punto de partida de Sunil. Dice que aprecia al equipo de Anthropic y le parece interesante el producto, pero su reacción al leer la especificación fue competitiva: Cloudflare puede hacerlo mejor. swyx pregunta qué tiene Cloudflare concretamente para que esa afirmación sea creíble. > *"Miré el producto y pensé: creo que quiero competir. Creo que podemos hacer algo mejor con Workers y Durable Objects."* ## [01:10] La infraestructura central de Cloudflare: Durable Objects y Dynamic Workers Sunil nombra dos primitivas que, en su opinión, toda plataforma de agentes acabará necesitando. Durable Objects son unidades serverless con estado — su tesis es que son la primera implementación a nivel de infraestructura del modelo de actores, no en código de usuario. Dynamic Workers es la respuesta de Cloudflare para ejecutar código generado por LLMs de forma segura: un `eval` rediseñado con arranque instantáneo, superficie de API configurable y tráfico de salida bloqueado por defecto. Juntos permiten ejecutar pasos de agente en cómputo aislado sin levantar máquinas virtuales completas. > *"Es la primera implementación del modelo de actores en una capa de infraestructura, no en código de usuario."* ## [02:34] Cómo Cloudflare aborda la arquitectura de agentes de IA El servidor MCP de Cloudflare, construido por su colega Matt Carey, muestra Dynamic Workers en acción. La API de Cloudflare tiene 2600 endpoints — exponer una herramienta por endpoint destruiría cualquier ventana de contexto de un LLM. En su lugar, el servidor lo colapsa todo en dos llamadas: `search` y `execute`, ambas respaldadas por código JavaScript que corre en un isolate. El agente envía código, el isolate lo ejecuta y devuelve el resultado — sin ida y vuelta, con verificación de tipos. > *"En una sola llamada, sin ir y venir con el LLM, y tiene verificación de tipos. Resulta que los LLMs son muy buenos ejecutando código."* ## [03:40] El futuro del software agéntico y la estandarización del "harness" swyx pregunta si el concepto de harness de la especificación de Anthropic podría convertirse en un estándar multiplataforma. La respuesta de Sunil: nadie ha construido aún el React de los agentes de IA. Traza la analogía con React en 2013 deliberadamente — la gente salió de la charla de JSConf diciendo que Facebook odiaba JavaScript, y aun así React definió todos los frameworks de UI que vinieron después. Ahora mismo cada equipo construye su propio harness a su manera, y nada es reproducible entre lenguajes, empresas e infraestructuras. swyx lanza la idea de que las skills — markdown plano — podrían ser ya esa capa unificadora; a Sunil le parece genuinamente atractiva, aunque teme el techo de especificidad. > *"Es muy difícil, pero la forma en que lo tengo en la cabeza es: nadie ha construido el React todavía."* ## [06:11] El fenómeno de los "slop forks" y la cultura open-source swyx plantea los "slop forks" — forks generados por IA de proyectos populares — y Sunil se anima. En su lectura, hacer un fork es un gesto de prestigio y respeto, no de robo. El ecosistema de React creció mediante forks. Anima a quien quiera construir algo que compita con el Cloudflare Agents SDK a que lo haga: todos ganan si lo hacen. > *"El fork es una gran señal de prestigio y respeto en mi cultura."* ## [06:36] El malentendido en redes sociales entre Vercel y Cloudflare En JSConf España, Sunil conoció a Harvey de Vercel y disfrutó mucho del tiempo con él. Descubrió Just Bash de Vercel Labs — una implementación pura en JavaScript de Bash — y quería portarlo a Cloudflare. Apuntó a Opus al código durante el almuerzo, recibió 5000 líneas de vuelta y planeaba limpiarlo antes de enviar un PR en condiciones el lunes. Se fue a dormir y se despertó con mensajes de la dirección de Cloudflare preguntando si había visto Twitter: el CTO de Vercel había criticado públicamente el trabajo, enmarcándolo como una decisión corporativa y no como un proyecto personal. Sunil respondió con claridad, explicó el contexto y luego vio cómo media internet salía a defenderle. > *"Entro en Twitter y el CTO de Vercel está criticando mi trabajo diciendo… 'Esto lo hizo Cloudflare.'"* ## [09:45] La importancia del fork en el desarrollo de software swyx conecta el incidente con Vercel con un patrón más amplio: una base de código filtrada que alguien reescribió en Python para escapar de la licencia (los abogados dictaminaron que era una obra derivada de todas formas). El argumento real de swyx es que merece la pena fomentar los slop forks — haz un fork de una dependencia, véndala, hazla tuya — para evitar la rotura repentina de upstream del problema de LiteLLM o Axios. Sunil coincide: antes de NPM, el software se difundía en Usenet exactamente con ese patrón, y acortar el ciclo del fork no es más que esa tradición continuando. > *"El fork es algo tan fundamental en la forma en que construimos software."* ## [12:04] La naturaleza adversarial de los repositorios open-source actuales El Cloudflare Agents SDK ha tenido que cerrar las contribuciones mediante pull requests; ahora solo se aceptan issues. Sunil habla con mantenedores de open-source en la conferencia que describen lo mismo: los repositorios se han convertido en territorio adversarial, y el peor vector de ataque son los falsos informes de seguridad que parecen completamente legítimos hasta que los lees con detenimiento. swyx lo conecta con una charla matutina de Peter de Claude Code — la principal superficie de ataque actual es una dependencia comprometida que se cuele en Claude Code, lo que daría acceso a todos los desarrolladores que lo usan. > *"Los repositorios open-source se han vuelto tan adversariales que la gente casi tiene miedo de ganar popularidad en ese espacio."* ## [13:04] Reflexiones finales y el llamado a ser original El cierre de Sunil es directo: deja de construir el décimo framework de agentes. Construye ciencia ficción. Construye algo para tu familia. Usa el Agent SDK, pero úsalo en algo donde la infraestructura y los LLMs casi te fallen — porque ahí es donde vive el próximo salto cualitativo. swyx cierra con un guiño al término "alpha thought leading" que Sunil acuñó en React Rally 2018. > *"Construye cosas de ciencia ficción. Cosas para tu familia. Tienes tanto poder para cambiar el mundo y quiero que la gente simplemente sea original."* ## Entidades - **swyx** (Persona): Presentador de Latent Space; amigo de larga data de Sunil; acuñó "alpha thought leading" tras una ocurrencia de Sunil en React Rally 2018. - **Sunil Pai** (Persona): Responsable de la plataforma para desarrolladores en Cloudflare; reconocido por swyx como creador de Code Mode; ponente principal en AIE Europe. - **Cloudflare** (Organización): Empresa de infraestructura cloud; construye infraestructura para agentes sobre Durable Objects y Dynamic Workers. - **Anthropic** (Organización): Empresa de IA; lanzó Cloud Managed Agents, el producto con el que Sunil posiciona a Cloudflare para competir. - **Vercel** (Organización): Empresa de cloud para frontend; Sunil usa su AI SDK; protagonista del malentendido en Twitter. - **Durable Objects** (Software): Primitiva serverless con estado de Cloudflare; Sunil sostiene que es la primera implementación del modelo de actores a nivel de infraestructura. - **Dynamic Workers** (Software): Funcionalidad de Cloudflare para ejecutar JavaScript generado por LLMs o usuarios en un isolate seguro con arranque instantáneo. - **Just Bash** (Software): Proyecto de Vercel Labs — una implementación pura en JavaScript de Bash — que Sunil estaba portando a Cloudflare cuando ocurrió el incidente en Twitter. - **MCP** (Concepto): Model Context Protocol; el servidor MCP de Cloudflare colapsa 2600 endpoints de API en dos llamadas usando Dynamic Workers. - **Slop forks** (Concepto): Forks de proyectos existentes generados por IA; Sunil los enmarca como continuación de la cultura del fork en open-source — una señal de respeto, no de plagio.

Grabado en directo en AI Engineer London, swyx conversa con Omar Sanseviero — director de Developer Experience en Google DeepMind — en una charla de 30 minutos sobre las novedades arquitectónicas de Gemma 4, la estrategia de modelos abiertos de Google y hacia dónde crece el equipo de DevEx. Omar destapa los embeddings por capa, explica por qué se ha enfriado la fiebre del fine-tuning, qué cambia realmente con la llegada de Kaggle a DeepMind y si la "auto-investigación" es real o puro ruido. ## [00:00] Introducción a Gemma 4 y el alcance del equipo El argumento de venta de Omar en una frase: Gemma 4 concentra "el modelo abierto más capaz que hemos lanzado hasta ahora", con la restricción de maximizar la inteligencia por parámetro y soporte multimodal completo, sin que el tamaño del modelo se vuelva inmanejable para la inferencia local. > *"Intentamos comprimir tanta inteligencia por parámetro como fue posible."* ## [00:23] Parámetros efectivos vs. parámetros activos El movimiento arquitectónico clave en los modelos pequeños de Gemma 4 es insertar una tabla de embeddings por capa en cada bloque del transformer. Al ser una búsqueda en tabla y no una multiplicación matricial, los 3.000 millones de parámetros de embeddings no necesitan estar en la memoria de la GPU — pueden vivir en la CPU o en disco, mientras que solo los 2.000 millones activos hacen cómputo real. Omar reconoce sin rodeos que este truco está pensado para dispositivos: a mayor escala, las arquitecturas densas o MoE son más apropiadas. > *"El modelo Gemma 4 es E2B. Eso significa que efectivamente carga 2.000 millones de parámetros en la GPU. En realidad tiene casi 5.000 millones, pero esos 3.000 millones adicionales pueden residir en la CPU o en el disco."* ## [01:43] Casos de uso en dispositivo e integración de Gemini Nano Los teléfonos Pixel y los Samsung de gama alta incluyen Gemini Nano de serie — un modelo entrenado sobre Gemma 3N, la arquitectura que Google diseñó específicamente para las limitaciones del móvil. La misma idea de descargar parámetros se aplica a estas variantes más pequeñas. Cuando swyx pregunta si escala al rango de 29B–31B, Omar responde que "están haciendo muchos experimentos — pronto más novedades." > *"Cuando compras estos teléfonos de gama alta, ya puedes usar Gemini de serie."* ## [03:14] El lanzamiento de un modelo por dentro y el ecosistema de desarrolladores El equipo de Gemma es más pequeño de lo que la gente imagina: dos o tres PMs, un responsable de marketing y el núcleo de ingenieros e investigadores. La complejidad de un lanzamiento viene del grafo externo: 50 partners (llama.cpp, Ollama, MLX, Hugging Face, vLLM, Nvidia, AMD y más) coordinados en paralelo, más la colaboración interna con Google Cloud, Vertex, ADK y Android. El lanzamiento de Gemma 4 también incluyó una integración nativa con el modo agente de Android Studio, que permite ejecutar inferencia local de Gemma 4 para asistencia de código. > *"Tuvimos casi 50 partners externos para el lanzamiento de Gemma 4, que ha sido el lanzamiento más complejo."* ## [04:29] Uso sin conexión vs. API y el crecimiento futuro de los modelos La división privacidad/conectividad es real, pero no cuenta toda la historia. Omar traza una línea más clara: los modelos locales de hoy son excelentes en capacidades — llamadas a funciones, seguimiento de instrucciones, tareas agénticas — pero siguen perdiendo en densidad de conocimiento; necesitas un modelo grande para recuperar datos de nicho con fiabilidad. Su apuesta a 1–2 años: un modelo del nivel de Gemini Pro corriendo completamente en el dispositivo, habilitando experiencias que hoy requieren conexión a una API. > *"Creo que en 1 o 2 años podrás ejecutar un modelo tan potente como Gemini Pro directamente en tu teléfono."* ## [06:26] Capacidades y limitaciones multimodales de Gemma 4 Gemma 4 hereda la pila de investigación de Gemini 3, lo que da incluso al modelo de 2B comprensión de audio (reconocimiento de voz, transcripción con traducción, preguntas sobre clips de audio) y visión (detección de objetos, señalización, subtítulos). Omar nombra explícitamente dos carencias: falta la segmentación de imágenes, y no se puede combinar vídeo y audio en un mismo prompt — deben entrar como flujos separados. La síntesis de voz nativa está en exploración, pero sin anuncios por ahora. > *"Podemos procesar vídeo o audio por separado, pero si quieres pasar en el mismo prompt una parte visual y una parte de audio, aún necesitamos mejorar en eso."* ## [08:08] Claves del tokenizador multilingüe El tokenizador de Gemma es el mismo que impulsa Gemini — una decisión de diseño que le otorga una base multilingüe excepcionalmente sólida en 140 idiomas. El hallazgo concreto de Omar: toma Gemma 3 como base, haz fine-tuning para un idioma del sudeste asiático como el vietnamita, y supera a modelos base que puntúan más alto en benchmarks en inglés. El tokenizador captura tokens propios de cada lengua en lugar de forzar scripts no latinos a través de fragmentos de subpalabras optimizados para el inglés. > *"Si haces fine-tuning de todos estos modelos para un idioma específico del sudeste asiático — vietnamita, por ejemplo — Gemma daría mejores resultados aunque los otros modelos base fueran potencialmente superiores."* ## [09:30] El equipo de Developer Experience de Google en AI Engineer Londres es la sede de DeepMind, así que presentarse con un equipo completo en AI Engineer Europe fue una declaración deliberada. Omar trajo investigadores de Gemma 4, generación de texto por difusión, robótica, ML en dispositivo y Android — no una gira de DevEx a secas. swyx resume el alcance sin rodeos: "Es el laboratorio con mayor scope. Lo hacéis todo, incluidos delfines." > *"Trajimos gente de robótica, de investigación, de Android. Es muy emocionante mostrar todo lo que está construyendo la compañía."* ## [10:42] Introducción a áreas de investigación: modelos de difusión para texto Google anunció Gemini Diffusion en I/O — un transformer de difusión que genera texto (no imágenes) a velocidades muy superiores a la decodificación autoregresiva. La valoración honesta de Omar: la calidad todavía no alcanza a los modelos autoregresivos, y hacer fine-tuning de transformers de difusión es más difícil porque los cambios de distribución afectan al enrutamiento de forma distinta. swyx esboza una arquitectura plausible donde los modelos de difusión actúan como ejecutores rápidos de sistema 1 mientras los autoregresivos gestionan la planificación compleja — Omar lo ve posible, pero prematuro. > *"Por ahora sigue siendo muy experimental. La calidad del modelo aún es algo inferior a la de un modelo autoregresivo normal."* ## [13:37] Estado actual del fine-tuning y tendencias de la comunidad Las comunidades de fine-tuning alcanzaron su auge alrededor de 2023; Omar observa ahora el repliegue de la marea. Varios partners del lanzamiento de Gemma 4 tenían planeado hacer fine-tuning del modelo de visión de 27B y cancelaron a mitad porque el modelo base ya cubría la necesidad. Cambios de comportamiento de uso general que antes requerían fine-tuning ahora se resuelven con prompting. Lo que queda: fine-tuning específico de dominio para sanidad, finanzas y datos de nicho — más el reto organizativo de gestionar la compatibilidad de LoRA cuando el modelo base se actualiza. > *"Vi muchos de esos casos — y ahora veo menos entusiasmo por el fine-tuning en modelos conversacionales de uso general."* ## [16:29] Ventajas y desventajas entre arquitecturas densas y dispersas Gemma 4 lanza dos modelos grandes con recuentos de parámetros similares: un modelo denso de 31B (máxima inteligencia bruta, cabe en una GPU de consumo cuantizado) y un MoE de 27B con 4B parámetros activos (inferencia más rápida dentro del mismo hardware). Las elecciones de tamaño fueron decisiones deliberadas de facilidad para el desarrollador. La advertencia de Omar para quienes hagan fine-tuning: las recetas de entrenamiento MoE y los hiperparámetros no se trasladan bien desde modelos densos — el cambio de distribución impacta el enrutamiento de formas que no se entienden del todo, posiblemente porque los cambios en la distribución de entrada alteran qué expertos se activan. > *"Los MoE son difíciles de ajustar. Funcionan muy bien para inferencia, pero cuando la gente los hace fine-tuning, les cuesta bastante."* ## [18:29] Inteligencia por parámetro e investigación futura A lo largo de Gemma 2, 3 y 4, Google ha mantenido el recuento máximo de parámetros aproximadamente constante en ~30B mientras el techo de capacidades ha subido de forma significativa — una demostración directa de la mejora en inteligencia por parámetro. El problema de comparación más difícil: una vez que se introduce la dispersión MoE y el desplazamiento de parámetros, los recuentos de parámetros dejan de ser una moneda común. El horizonte honesto de Omar: las limitaciones de conocimiento son probablemente estructurales — un modelo de 30B dentro de 3 años seguirá fallando en hechos muy específicos porque la teoría de la información limita cuánto puedes comprimir en pesos fijos. > *"¿Cuál es la inteligencia por parámetro? ¿Cómo maximizamos esta inteligencia por parámetro?"* ## [20:09] Gemma Scope e interpretabilidad mecanicista Google publicó Gemma Scope en diciembre — un conjunto de herramientas para analizar las activaciones por capa en los modelos Gemma 3, respaldado por un dataset de activaciones de varios terabytes (posiblemente a escala de petabytes) que cubre cada capa. Omar presenta la interpretabilidad mecanicista como una rampa de entrada de bajo cómputo a la investigación en ML: no necesitas un clúster de entrenamiento para hacer análisis de activaciones, y los experimentos te dan intuición tangible sobre cómo funcionan los internos del transformer. > *"Es un área donde no necesitas mucho cómputo para empezar. Te permite entender cómo funciona un modelo."* ## [21:12] La intersección entre investigación e ingeniería El catalizador para llevar investigadores a una conferencia de ingeniería: los ingenieros confían más en los modelos cuando entienden cómo fueron construidos, aunque nunca vayan a entrenar uno. Omar y swyx señalan que la frontera entre investigación e ingeniería se ha desdibujado — la mayor parte del trabajo investigador son ablaciones empíricas más cercanas a la ingeniería que a la teoría, y los agentes de código dan a los ingenieros acceso directo a experimentación que antes requería formación investigadora. Omar cita la comunidad de franken-merge y Axolotl como ejemplo de Reddit y Discord redescubriendo técnicas de forma independiente que los laboratorios publicaron después como papers. > *"Hay mucha experimentación empírica — ver qué funciona, qué no, mover cosas — que para mí es mucho más ingeniería que investigación."* ## [23:59] Perspectivas sobre la "auto-investigación" y la automatización agéntica swyx plantea la pregunta de fondo: ¿la auto-investigación es solo "barridos agénticos de parámetros" o puede producir descubrimientos al estilo del movimiento 37 que nadie habría buscado? Omar es cautelosamente escéptico — el historial de AutoML fue en su mayoría búsqueda en rejilla disfrazada, y el trabajo profundo de arquitectura probablemente no sea automatizable en los próximos 1–2 años. Pero sí cree que el fine-tuning en sí pronto será completamente guiado por agentes: los usuarios le pedirán a un agente que lance experimentos en lugar de escribir código de entrenamiento, usando herramientas como AutoTrain de Hugging Face o el CLI de Axolotl. > *"La próxima generación de personas que hagan fine-tuning serán personas que no programen en absoluto. La mayoría lo hará con un par de instrucciones."* ## [26:06] Expansión del equipo, hubs globales e integración con Kaggle El equipo de DevEx está contratando en Singapur y en India — ubicado junto a las oficinas de investigación de DeepMind para que el personal de DevRel pueda hablar directamente con los investigadores en lugar de estar en oficinas satélite de ventas aisladas. La noticia organizativa más relevante: Kaggle se incorporó a DeepMind, y su infraestructura de competiciones y benchmarks conecta directamente con las brechas de capacidad de Gemma/Gemini — los benchmarks creados por la comunidad pueden retroalimentar el proceso de entrenamiento. Omar describe el modelo como orientado al feedback: el equipo interactúa en redes sociales y eventos para entender qué están construyendo los desarrolladores, y lleva esa señal al equipo de modelado. > *"La forma en que hacemos Gemma, Gemini y todas nuestras herramientas se basa realmente en el feedback de las startups, la comunidad y los desarrolladores."* ## Entidades - **Omar Sanseviero** (Persona): Director de Developer Experience en Google DeepMind; anteriormente desarrolló DevRel en Hugging Face; lidera el ecosistema de desarrolladores de Gemma. - **swyx** (Persona): Presentador del podcast Latent Space; entrevistador en AI Engineer London 2026. - **Gemma 4** (Software): Familia de modelos abiertos de Google con arquitectura de embeddings por capa (desplazamiento de parámetros efectivos E2B), variantes de 2B/4B/27B MoE/31B denso, soporte para 140 idiomas y entrada multimodal. - **Gemini Nano** (Software): Modelo en dispositivo construido sobre la arquitectura de Gemma, incluido de serie en teléfonos Pixel y Samsung de gama alta. - **Gemma Scope** (Software): Kit de herramientas de Google para interpretabilidad mecanicista — analiza activaciones por capa en modelos Gemma 3; publicado en diciembre de 2025 con datos de activaciones a escala de petabytes. - **Gemini Diffusion** (Software): Transformer de difusión experimental de Google para generación de texto (no imágenes), anunciado en Google I/O; su principal ventaja es la velocidad de inferencia. - **Kaggle** (Organización): Plataforma de competiciones y benchmarks que se incorporó a Google DeepMind; integra evaluaciones creadas por la comunidad en los bucles de feedback de capacidad de Gemini. - **Google DeepMind** (Organización): Laboratorio consolidado de IA de Google; su scope abarca Gemma, Gemini, robótica, ML en dispositivo e interpretabilidad mecanicista. - **AI Engineer London** (Organización): Conferencia de ingeniería de IA aplicada (edición 2026); sede de esta entrevista y ciudad natal de DeepMind. - **MoE (Mixture of Experts)** (Concepto): Arquitectura dispersa donde solo un subconjunto de parámetros se activa por token; inferencia más rápida que la densa con el mismo recuento de parámetros, pero más difícil de ajustar por el enrutamiento sensible a la distribución. - **Embeddings por capa** (Concepto): Cambio arquitectónico de Gemma 4 — una tabla de búsqueda de embeddings insertada en cada capa del transformer, que permite que 3.000 millones de parámetros vivan fuera de la GPU sin coste de multiplicación matricial. - **Inteligencia por parámetro** (Concepto): La relación capacidad-peso que Gemma 2→3→4 ha mejorado manteniendo el recuento total de parámetros ~constante en 30B.

Ivan Burazin, CEO of Daytona, discusses the massive shift from building developer environments for humans to providing composable computers for AI agents. With 74% month-over-month growth and 850,000 daily runs, Daytona provides the bare-metal infrastructure required for stateful, high-performance agentic workflows. This conversation explores the technical challenges of spiky compute, the $10 trillion computer-use market, and why the future AI cloud will look more like Stripe than AWS. ## [00:00] Hook Ivan Burazin describes the intense, direct demand for Daytona's infrastructure, with potential users calling him personally to request access. This level of interest signaled a massive, untapped market for providing execution environments to every future AI agent. The team realized they had identified a critical missing piece in the AI development stack. > *I've never experienced this that people literally call you if you do not give them access. Like they want access right now.* > *[0, 0]* > * ] }, { * > *title": "Introduction* > *{'start': 72.0, 'summary': "Host swyx introduces Ivan Burazin, noting their shared history in the developer experience and 'end of localhost' movements. Ivan recalls reaching out to swyx years ago for advice on developer experience while working at a previous role. They reflect on how their early interactions and mutual interests in cloud-based development tools eventually led to their current collaboration.", 'quotes': ['I was one of the co-founders of code anywhere... we were thinking a long time of like local host should die.', [1, 36], '\n ]\n },\n {\n ', 'title": "CodeAnywhere', 'Shift', 'and the end of localhost', {'start': 195.0, 'summary': 'Ivan discusses his long history with his co-founder, dating back to early 2000s virtualization and the creation of CodeAnywhere. As the first browser-based IDE, CodeAnywhere predated modern infrastructure like Docker and Kubernetes, which provided the team with deep foundational knowledge. After a successful run with the Shift developer conference, they returned to their infrastructure roots to launch Daytona.', 'quotes': ['We originally started stacking stacking servers doing like virtualization in the early 2000s... and that was a services company which we sold.', [3, 38], '\n ]\n },\n {\n "title": "What Daytona is: composable computers for AI agents",\n "start": 358.0,\n "summary": ', "Ivan defines Daytona as a provider of 'composable computers' for AI agents", "moving beyond the limited industry term 'sandboxes.' He explains that agents require diverse computing environments tailored to specific tasks", 'much like different hardware setups for human professionals. This API-driven infrastructure allows agents to execute code in production-grade environments rather than just temporary test boxes.', {'quotes': ['What Daytona is today is essentially composable computers for AI agents... the market calls them sandboxes which [is] misleading.', [6, 41], '\n ]\n },\n {\n ', 'title": "The pivot from dev environments to AI sandboxes', {'start': 487.0, 'summary': "Ivan explains how observing early agents like Devon and OpenHands led to a realization that AI agents require a dedicated compute runtime. While their initial SaaS offering for human automation saw low traction, it attracted developers who specifically needed sandboxes for their agents. This feedback loop revealed a massive, underserved market for agent-specific infrastructure that standard cloud providers weren't addressing.", 'quotes': ['a lot of people reached out that were building agents and they were like hey my agent needs a compute sandbox runtime', [8, 50], '\n ]\n },\n {\n ', 'title": "The New Year’s Eve MVP and customers begging for API keys', {'start': 617.0, 'summary': "On New Year's Eve, Ivan 'vibe-coded' the first MVP of what would become the new Daytona. Although the CTO initially dismissed the code as 'garbage,' the core idea was strong enough to warrant a two-week professional rebuild. When they demoed this version to previous skeptics, the response was immediate and overwhelming, with users demanding API access before the calls even ended.", 'quotes': ["I've never experienced this that people literally call you if you do not give them access.", [12, 18], '\n ]\n },\n {\n ', 'title": "Bare metal', 'stateful sandboxes', 'and Daytona’s scheduler', {'start': 776.0, 'summary': "The team approached the technical architecture from first principles, deciding to run on bare metal rather than traditional VMs. They aimed to combine the speed of AWS Lambda with the stateful, long-running nature of an EC2 instance. This allows agents to 'pause and come back' to their work, much like a human closing a laptop lid, without losing state or performance.", 'quotes': ["agents will be like humans in the sense of you don't want your laptop to be shut down until you're done with work", [13, 57], '\n ]\n },\n {\n ', 'title": "60ms startup', 50, 0, 'sandboxes', 'and 850K daily runs', {'start': 1048.0, 'summary': "Daytona's infrastructure is optimized for both individual speed and massive concurrency, with a single instance spinning up in just 60 milliseconds. This scale supports high-volume customers who perform nearly 850,000 runs daily, with some requesting capacity for half a million concurrent CPUs. The system utilizes a custom scheduler and local NVMe drives to eliminate network latency and maximize IOPS.", 'quotes': ['Our time to spin up one is 60 milliseconds with network latency... if you want to spin up 50,000 at once, we are now at about 75 seconds.', [17, 40], ',\n ', 'The biggest customer of ours does like about 850', 0, "every single day is sort of where they're where they're just shy of a million.", [18, 17], '\n ]\n },\n {\n ', 'title": "Spiky RL/eval workloads and the new agent infra problem', {'start': 1313.0, 'summary': "The 'spiky' nature of AI workloads presents a major challenge for compute providers, leading to a mean utilization rate of only 15% despite peaks hitting 90%. Workloads are categorized into 'background agents' that follow human cycles and 'evaluations/RL' which fire off massive bursts of activity at unpredictable hours. To manage this, Daytona must use capacity commits to handle sudden bursts of 100,000 or more CPUs.", 'quotes': ["Daytona's mean utilization is 15%... because it's very spiky. But it's very spiky but we get up to 90%.", [23, 1], '\n ]\n },\n {\n ', 'title": "RL workloads', 'Kubernetes pain', 'and dynamic resizing', {'start': 1692.0, 'summary': "Daytona competes primarily against managed Kubernetes services like EKS and GKS, positioning itself as a more ergonomic 'Twilio or Stripe' for compute. Unlike Kubernetes, Daytona offers a seamless API for spinning up sandboxes with significantly faster startup times. A key advantage is the ability to dynamically resize sandboxes on the fly to prevent out-of-memory (OOM) errors, a feature difficult to implement on other platforms.", 'quotes': ["Daytona although it's a compute provider it's more akin to a Twilio and Stripe from a consumption perspective than it is an AWS", [29, 46], '\n ]\n },\n {\n ', 'title": "Why every AI agent needs a computer', {'start': 2011.0, 'summary': "Ivan outlines the massive scale of knowledge work, estimating a $50 trillion global salary pool, much of which is locked in legacy Windows applications. He argues that true automation requires 'human emulators' that can interact with these legacy systems via GUIs when APIs are incomplete. By automating 40% of this work, the market opportunity for agentic computer use reaches approximately $10 trillion annually.", 'quotes': ['If you take 40% of that, you get to essentially like 10 trillion dollars a year.', [35, 20], '\n ]\n },\n {\n ', 'title": "macOS sandboxes and Apple’s licensing problem', {'start': 2328.0, 'summary': "The discussion shifts to the difficulties of hosting Mac OS sandboxes compared to Windows and Linux. Apple's restrictive licensing only allows two parallel VMs per machine and requires a 24-hour lock-in for users, making per-second billing economically unfeasible. Furthermore, security restrictions prevent moving memory snapshots between physical machines, severely limiting the scalability of agentic workloads on Mac hardware.", 'quotes': ['Apple is shooting itself in the foot... if it would just enable a concurrency model similar to what you can get on a Windows.', [40, 52], '\n ]\n },\n {\n ', 'title": "Why CLI may matter more than MCP', {'start': 2668.0, 'summary': "The discussion compares the Model Context Protocol (MCP) to the Command Line Interface (CLI) for agentic action. While MCP acts as an interface for APIs, the CLI allows agents to execute scripts and perform deep data analysis within a sandbox. This layer of indirection enables more complex agentic workflows beyond simple data retrieval, allowing agents to actually 'do things' rather than just integrate.", 'quotes': ['the MCP is an interface against an API whereas the CLI is like you can actually go do things... the difference between integrations and actually running scripts.', [45, 34], '\n ]\n },\n {\n ', 'title": "Open source', 'GitHub stars', 'and agent integration', {'start': 2891.0, 'summary': "Ivan details Daytona's transition to an AGPLv3 license for its sandbox product to balance openness with commercial protection. This 'copyleft' approach allows enterprise use but prevents competitors from building proprietary forks without contributing back. Keeping the core engine transparent builds trust with users and allows large enterprises to bypass lengthy security audits by providing agents with full context.", 'quotes': ["in the new sandbox product we did add a AGPL3... you essentially can't make a competitor without open sourcing your stuff.", [49, 49], '\n ]\n },\n {\n ', 'title": "Git', 'CI/CD', 'and agent collaboration bottlenecks', {'start': 3191.0, 'summary': 'Current versioning systems like GitHub are often too slow for the high-velocity output of AI agents, leading to bottlenecks in CI/CD pipelines. Some developers are creating makeshift solutions like dumping codebases into JSON files on S3 to bypass Git overhead. There is a growing need for an agent collaboration layer that precedes the traditional Git-based pipeline to handle companies generating over 1,000 PRs per day.', 'quotes': ["GitHub as-is was an overhead... it wasn't fast enough what they needed.", [54, 3], '\n ]\n },\n {\n ', 'title": "Founder life and building a 25-person infra company', {'start': 3495.0, 'summary': "Daytona's success stems from a core team of 13 people who have worked together for over seven years, fostering a high-trust culture. Ivan acknowledges the difficulty of the founder journey, including being away from family, but posits that growth requires 'pain.' He views his work as building the spiritual successor to serverless and Kubernetes for the agent era, requiring radical responsiveness as a differentiator.", 'quotes': ['Of the 25 people in Daytona, I think about 13 of them we have worked with seven years plus.', [58, 57], '\n ]\n },\n {\n ', 'title": "AI SaaS', 'token resale', 'and API-first business models', {'start': 3764.0, 'summary': 'Ivan presents a critical take on the SaaS ecosystem, arguing that the market is incorrectly applying a premium to vendors who simply resell AI tokens. He points out that these models have significantly worse margins than traditional SaaS. Instead, he advocates for companies to expose their data via APIs and charge for consumption, allowing for actual revenue acceleration through increased agentic usage.', 'quotes': ["The market is adding premium to SAS vendors that are reselling tokens. And I think that's incorrect.", [62, 54], '\n ]\n },\n {\n "title": ', 'GPU sandboxes', 'data centers', 'and compute growth', {'start': 3970.0, 'summary': 'Daytona plans to introduce GPU sandboxes to support workloads like 3D rendering and reinforcement learning on CAD, rather than focusing on inference. While the company currently runs on bare metal via colocation providers, Ivan notes they are architected to potentially own data centers in the future. He currently avoids the high capital risk of building data centers for single-digit margin gains.', 'quotes': ['We will [offer GPUs], but not for inference. Like essentially what we think about is like the GPU sandbox.', [66, 21], '\n ]\n },\n {\n ', 'title": "Why the AI cloud may look more like Stripe than AWS', {'start': 4188.0, 'summary': "The conversation concludes by imagining the 'AWS for AI Agents,' which Ivan suggests might look more like Stripe than a traditional cloud provider. This future 'AI Cloud' will integrate sandboxes, web search, and databases as fundamental primitives. While companies like Cloudflare and OpenAI are competing for this space, Ivan hints that many more infrastructure primitives for agents are yet to be developed.", 'quotes': ["There will be a cloud built out specifically for agents and so that cloud will have sandboxes and it will have web search and it'll have databases.", [70, 47], '\n ]\n },\n {\n ', 'title": "Closing thoughts', {'start': 4286.0, 'summary': 'The discussion ends with the observation that the AI infrastructure market is growing at an unprecedented baseline of 40-75% month-over-month. Ivan and swyx reflect on the race to secure hardware and the shift toward specialized agent clouds that will define the next decade of computing.', 'quotes': ["The entire infrastructure market is growing 40% plus or minus month over month... if you're not growing 40%ish... you don't have to come to work.", [68, 23], '\n ]\n }\n ],\n ', 'entities": [\n {\n "name": "Ivan Burazin', {'type': 'person', 'description': 'CEO of Daytona and co-founder of CodeAnywhere.'}, {'name': 'swyx', 'type': 'person', 'description': 'Host of Latent Space and early investor in Daytona.'}, {'name': 'Daytona', 'type': 'organization', 'description': 'A company providing composable computers and sandboxes for AI agents.'}, {'name': 'CodeAnywhere', 'type': 'organization', 'description': 'The first browser-based IDE, co-founded by Ivan Burazin.'}, {'name': 'Devon', 'type': 'product', 'description': 'An early AI software engineer agent.'}, {'name': 'OpenHands', 'type': 'product', 'description': 'An open-source AI agent project formerly known as OpenDevin.'}, {'name': 'Kubernetes', 'type': 'technology', 'description': "Orchestration technology mentioned as a competitor to Daytona's ergonomic API."}, {'name': 'Apple', 'type': 'organization', 'description': 'Mentioned regarding restrictive Mac OS virtualization licensing.'}, {'name': 'Salesforce', 'type': 'organization', 'description': 'Cloud-based software company mentioned for its API-first strategy.'}, {'name': 'GitHub', 'type': 'organization', 'description': 'Developer platform noted for being a bottleneck in agentic CI/CD workflows.'}, {'name': 'Nvidia', 'type': 'organization', 'description': 'The primary provider of GPUs whose supply constraints dictate market growth.'}, {'name': 'Stripe', 'type': 'organization', 'description': 'Used as a comparison for the consumption-based model of the future AI cloud.'}], 'tags': ['ai-agents', 'infrastructure', 'sandboxing', 'bare-metal', 'cloud-computing', 'developer-tools', 'computer-use', 'saas-growth'], 'seo_title': "AI Agents Need Computers: Ivan Burazin on Daytona's Pivot", 'seo_description': 'Ivan Burazin explains why AI agents need composable computers and how Daytona pivoted from dev environments to 850K daily agent runs.', 'confidence': {'score': 0.98, 'rationale': 'The summary synthesizes multiple detailed chunks covering technical metrics, business strategy, and market philosophy with high fidelity to the source.'}}]}]}]}]}]}]}]}]}]}]}]}]}]}]}]}]}]}]}* ## [01:12] Introduction ## [03:15] CodeAnywhere, Shift, and the end of localhost ## [05:58] What Daytona is: composable computers for AI agents ## [08:07] The pivot from dev environments to AI sandboxes ## [10:17] The New Year’s Eve MVP and customers begging for API keys ## [12:56] Bare metal, stateful sandboxes, and Daytona’s scheduler ## [17:28] 60ms startup, 50,000 sandboxes, and 850K daily runs ## [21:53] Spiky RL/eval workloads and the new agent infra problem ## [28:12] RL workloads, Kubernetes pain, and dynamic resizing ## [33:31] Why every AI agent needs a computer ## [38:48] macOS sandboxes and Apple’s licensing problem ## [44:28] Why CLI may matter more than MCP ## [48:11] Open source, GitHub stars, and agent integration ## [53:11] Git, CI/CD, and agent collaboration bottlenecks ## [58:15] Founder life and building a 25-person infra company ## [1:02:44] AI SaaS, token resale, and API-first business models ## [1:06:10] GPU sandboxes, data centers, and compute growth ## [1:09:48] Why the AI cloud may look more like Stripe than AWS ## [1:11:26] Closing thoughts

Jake Cooper, CEO of Railway, details the platform's evolution from a high-burn startup to a sustainable, bare-metal cloud infrastructure powering 3 million users. He argues that the rise of AI agents necessitates a fundamental rebuild of the cloud, moving away from human-centric tools like Kubernetes and pull requests toward high-density CLI handles and production forking. This conversation provides a roadmap for building modular, high-scale systems capable of supporting the next generation of automated software development. ## [00:00] Intro Jake Cooper argues that developers should stop writing code by hand and instead focus on reviewing agent-generated code to maintain architectural integrity. He emphasizes that while AI tools have improved significantly, underlying architectural patterns matter more than ever in an automated workflow. The hosts introduce Jake as the 'Conductor' of Railway, setting the stage for a discussion on the future of cloud platforms and developer experience. > *you should be reviewing the code that you are writing instead of trying to go and write it by hand.* > *[0, 10]* ## [01:19] What Is Railway? Railway is described as a platform that allows users to deploy applications and databases instantly via a canvas or AI prompts like Claude. Jake explains that the goal is to manage software versioning and environment cloning to reduce the complexity of traditional tools like Docker and Kubernetes. By tracking all changes, Railway enables developers to fork production environments into parallel universes for safe validation without reproducing staging environments manually. > *railway is the easiest way to ship anything.* > *[2, 29]* > *we want to make it really easy for not just to like deploy things, but for you to almost like evolve applications over time.* > *[2, 49]* ## [03:26] Jake’s Path to Railway Jake details his professional journey from front-end work at Wolfram to building distributed systems for Jump bikes at Uber using Cadence. He describes his engineering philosophy as a willingness to 'swim to the bottom of the pool,' which includes writing kernel patches to ensure the best possible user experience. Additionally, he critiques GitHub's architecture, specifically the 'broken pointers' created by cloning, which complicates upstream contributions. > *we will swim to the bottom of the swimming pool to go and get the experience* > *[4, 35]* > *GitHub's original sin is that it's like almost a series of broken pointers.* > *[6, 2]* ## [07:32] Railway’s Six-Year Growth Story Jake presents a growth chart illustrating the rapid increase in daily signups for the Railway platform, which has transitioned from a 'slow grind' to adding 100,000 users weekly. Early growth was driven by high-touch interaction on Discord and a determination to acquire the first 100 core users manually. This visual data serves as a transition into the company's history of scaling and its move toward becoming a primary cloud provider. > *so I just wanted to like pull up this glorious chart you say which is basically your usage or number of daily signups* > *[7, 34]* > *Trying to get those initial like first 100 users to like actually kind of come back to it.* > *[8, 21]* ## [10:11] Rebuilding the Business After the Free Tier At one point, Railway was losing $500,000 a month while only generating $50,000 in revenue, despite having $20 million in the bank. Cooper realized this was an unsustainable business model and chose to prioritize long-term viability over vanity metrics, temporarily closing the free tier to rebuild. The company now maintains a lean team of 35 people, preferring to build automated systems rather than throwing headcount at problems. > *We basically had to kind of close off the the free kind of users for a little while, rebuild the business.* > *[11, 47]* > *We're 35 people right now... we don't want to just like add headcount for the sake of headcount.* > *[10, 52]* ## [12:36] Agents as the Next Software Platform Over the last six months, Railway has prioritized 'agentic' development as the primary mechanism for building and deploying software. Cooper believes the industry is moving from assembly and high-level languages to 'words' as the primary interface. He envisions a future where thousands of agents run in parallel, requiring new tools for coordination and version control to manage the super-exponential growth of workloads. > *We've moved from assembly to C to C++ to JavaScript to now like words.* > *[13, 23]* ## [14:48] Railway’s Infrastructure Philosophy Jake Cooper explains that Railway prioritizes control over low-level primitives like network, compute, and storage to optimize for AI agent workloads. By avoiding Kubernetes in favor of custom orchestration, the team can place workloads with high precision to ensure memory efficiency. This level of control is necessary to prevent cost structures from ballooning as agent usage increases and requires thousands of parallel instances. > *you have to be very very efficient with these agents... or you're going to massively massively blow up your cost structure* > *[15, 10]* > *How do you get agents to coordinate? How do you go and get them to be able to like safely version changes?* > *[14, 28]* ## [17:01] Bare Metal, Cloud Economics, and the Compute Crunch Cooper describes the transition to bare metal as highly lucrative, reporting a payback period of just three months compared to cloud rental costs. This strategy allows the company to achieve 70% margins while leveraging hardware that remains viable for several years. He also notes the surprising appreciation of hardware assets, such as RAM, due to the global compute shortage and supply chain constraints. > *our payback period when we go to to metal... if we rent it in the cloud, our payback period is about 3 months.* > *[17, 2]* > *hardware and all of this stuff is... appreciated in value because RAM has gone up* > *[17, 50]* ## [18:41] Cloud Bursting and Five-Cloud Networking To maintain growth without being compute-constrained, Railway utilizes a hybrid cloud strategy for bursting capacity across AWS, GCP, and Oracle. This required building a custom network overlay capable of straddling five different cloud environments simultaneously. While this complexity led to past reliability challenges, it now allows Railway to scale rapidly regardless of individual provider quotas or hardware availability. > *I spent a weekend rebuilding our entire like network like overlay essentially so that we could straddle uh five different clouds* > *[19, 41]* > *we still maintain like cloud presence for like bursting essentially* > *[18, 52]* ## [21:39] Data Center Debt and Infra Financing Cooper highlights the strategic use of data center debt, secured against hardware, as a more efficient alternative to venture capital for infrastructure expansion. By treating compute capacity as a linear driver of revenue, Railway can scale as quickly as they can deploy new hardware. He encourages infrastructure startups to explore diverse financing tools rather than relying solely on expensive venture equity for physical assets. > *we can scale revenue as basically as quickly as we can scale compute* > *[21, 20]* > *our margins on metal are like quite high for the like 70%.* > *[20, 46]* ## [24:50] Data Centers in Space Jake Cooper and the hosts explore the technical challenges of placing data centers in space, specifically the issue of heat dissipation in a vacuum. Cooper expresses skepticism toward current proposals that ignore fundamental thermodynamic laws, comparing the 'figure it out later' mentality to science fiction. He highlights the difficulty VCs face in distinguishing between visionary ideas and technical 'grifts' in the space-tech sector. > *I haven't seen anybody like prove how you're going to go and dissipate that much heat in a vacuum* > *[25, 16]* > *how do you know what's like basically not possible and like is a grift versus like uh is possible but like sounds completely insane* > *[26, 16]* ## [26:43] What Agents Need From Infrastructure Cooper outlines the infrastructure needs of AI agents, noting they require versioning, observability, and storage similar to humans but at a 1000x scale. He predicts that current industry standards like Kubernetes and Envoy will become bottlenecks as agentic workloads compress development cycles. To support this growth, infrastructure must be modular enough to allow for the rapid replacement of failing components without human intervention. > *the workload profile doesn't change so much as it gets like massively massively compressed because you need to do thousands of these things* > *[28, 28]* > *you just need at a thousandx scale* > *[29, 13]* ## [29:43] CLIs, Canvas, and Agent-Native UX Cooper explains that while humans prefer simplicity, agents benefit from high-density CLI interfaces with numerous flags that serve as 'handles.' The Railway Canvas is also evolving into an output mechanism and 'context anchor' rather than just an input tool. This hierarchical view of infrastructure prevents critical knowledge from being siloed as teams scale complex 'hyperstructures' using automated agents. > *If you hand it to an agent and you say, 'Hey, that's 40 arguments and 600 flags.' Like, oh yeah, this is excellent.* > *[30, 35]* > *It has to be almost like an anchor for your context. It has to be like a port in the storm.* > *[34, 27]* ## [36:34] Central Station, Incidents, and Responsible Disclosure Railway utilizes an internal tool called Central Station to aggregate feedback and user context, moving away from static communication channels like Slack. The team emphasizes transparency by exposing real-time metrics and detailed incident reports, operating under a core value of 'honor.' This approach involves over-disclosing issues to users rather than providing vague or misleading information during outages. > *We'd rather overdisclose and know that you know that something is wrong versus almost like having your provider gaslight you.* > *[40, 22]* > *If you can dynamically aggregate that information and dynamically route it to the right person... this is no longer a manual process.* > *[37, 10]* ## [41:49] Safe Rollouts, SRE Agents, and Production Forks To mitigate the impact of bugs, Railway employs incremental rollouts and makes it easy to test behaviors in safe, shadowed environments. Cooper argues that production should not be treated as 'sacred' to the point of stagnation; instead, infrastructure should allow for trivial production forks. This is essential for AI agents, which face a 'stacking entropy' problem without safe iteration primitives to prevent system drift. > *We've built so much ceremony around like production is sacred... we need to get to a point where it's just trivially easy to test different behaviors.* > *[41, 33]* > *I think if you don't have the primitives to make iterating in production safe, it becomes very very difficult.* > *[44, 3]* ## [46:19] AI SRE, Specs, Code, and Tests Jake Cooper reflects on his transition from an AI skeptic to a believer, noting that the safety of AI SREs depends on infrastructure primitives. He advocates for the 'Holy Trinity' of software engineering: a clear specification, the code, and the tests. By aligning these three, developers and agents can reconcile discrepancies and maintain system integrity during rapid, automated iteration. > *If you just unleash an AI SRE on your production infrastructure... it's going to nuke your production database.* > *[46, 37]* > *You need three points essentially which is you need a clear spec... you need the code and then you need the tests.* > *[48, 22]* ## [49:43] Self-Replicating Infrastructure and the New Serverless The speakers explore the concept of agents using the Railway CLI to modify their own infrastructure, creating a self-replicating loop. This shift necessitates a move away from expensive, static virtual machines toward cheap, instantaneous 'atomic units of deploy' like isolates or sandboxes. The goal is to make throwaway copies of production as trivial and cost-effective as possible for agentic experimentation. > *The agent can like modify its own infra which I think is... yeah it's nuts.* > *[50, 4]* > *How do you go and make those throwaway copies like as trivial as possible to spin up run super cheap etc.* > *[50, 53]* ## [54:37] Heroku, Temporal, and Workflow Engines Cooper attributes the decline of Heroku to Salesforce's lack of focus on compute as a core business, leading to product stagnation. Railway positions itself as a 'fluid compute' provider, leveraging Cooper's decade of experience with Temporal (and its precursor Cadence) for durable workflows. Railway is a power user of Temporal, using it to manage complex, long-running infrastructure tasks at scale. > *The business of Salesforce is to build a really really good CRM... and then you acquire this business as a compute business that's kind of an offshoot* > *[55, 33]* > *I have used Temporal for almost like 10 years now, right? Because like Cadence, all of us other things.* > *[60, 5]* ## [1:05:26] Railpack, Nixpacks, and Lazy-Loaded Filesystems Railway is developing Railpack, an engine for determining source code dependencies, which evolved from their earlier Nix-based tool, Nixpacks. While Nix offers theoretical benefits for versioning, Railway found it caused significant image bloat and scaling issues for real-world workloads. They are now exploring content-addressable file systems to enable lazy loading of data into memory for faster deployments. > *If you want version X and version Y, you end up bloating a lot of your kind of like package like space.* > *[66, 2]* ## [1:07:20] Coding Agents, Token Spend, and Roadmap Acceleration With a monthly cloud spend reaching $300,000, Railway heavily incentivizes the use of AI coding agents among its employees. Cooper argues that manual code generation is an inefficient use of time, urging developers to focus on architectural patterns and code review. This allows the team to 'speedrun' their product roadmap by automating complex infrastructure tasks and test generation. > *If you are writing code by hand you are doing this wrong... you should be reviewing the code that you are writing.* > *[67, 37]* > *If you're not using the AI systems to almost like speedrun your road map... then you're kind of missing a large point.* > *[69, 12]* ## [1:12:15] The Pull Request Is Dying The traditional SDLC is undergoing a radical transformation where the pull request and manual code review are losing relevance. Impact is increasingly measured by the 'percentage of tokens that end up in production' rather than lines of code. As AI systems handle more reconciliation and validation, the focus shifts from the PR to the initial prompt and final deployment. > *The pull request is dying... it's going to be the prompt... and beyond that code review is also kind of dying.* > *[72, 23]* > *The really naive way to go in and measure this is almost like your percentage of tokens that end up in production.* > *[71, 40]* ## [1:13:47] Feature Flags and the Agent-Era SDLC Jake Cooper discusses the critical role of feature flagging in managing the 1000x compression of the SDLC driven by AI agents. He argues that incremental rollouts and blast radius management through flagging will become even more essential for safety as deployment speed increases. This culture of flagging allows for rapid experimentation without compromising system stability for enterprise customers. > *Everything's just going to get compressed by like a thousandx so that everybody can go and do that.* > *[77, 21]* ## [1:17:34] Cattle, Pets, and Cloning Machines Jake offers a contrarian view on the 'cattle not pets' philosophy, suggesting that snapshotting allows developers to treat infrastructure like 'pets' again. By snapshotting every frame and lazily loading file systems, the overhead of traditional DevOps tools like Dockerfiles is reduced. Railway even modifies the kernel to support persistent connections during these system snapshots. > *I think you can move towards having pets so long as... you have a cloning machine for your pets.* > *[78, 2]* > *If you can snapshot every single thing at every frame, then like it actually doesn't matter if you know that obliterated.* > *[78, 12]* ## [1:20:48] Solo Founder Lessons Jake reflects on his path as a solo founder, contrasting it with the Silicon Valley consensus of finding a co-founder. He emphasizes the need to be obsessed with every layer of the stack, from kernel-level changes to go-to-market strategies. He argues that having two co-founders can often lead to deadlocks without a clear tiebreak, whereas solo leadership allows for singular vision. > *Two is the worst number of co-founders is because you have no tiebreak... you basically are like, well, I disagree on this thing.* > *[82, 49]* ## [1:25:31] Focus, GPUs, and Building a New Cloud Railway is intentionally avoiding the GPU provider market for now to maintain its core mission, though Cooper admits GPUs are an inevitable part of their long-term roadmap. He stresses that companies are defined as much by what they choose not to do as by what they execute. The ultimate goal is full vertical integration to ensure a seamless experience from logic to execution. > *I think you're you're defined almost more by the things that you don't do than the things that you do* > *[86, 8]* > *I can tell you for a fact that we will not be doing GPUs now, but we 100% will be doing GPUs at some point.* > *[86, 50]* ## [1:29:39] Closing Thoughts Cooper reveals that Railway is moving toward 100% ownership of its data centers to avoid copying the infrastructure of legacy hyperscalers. By inventing their own infrastructure from scratch, Railway aims to support 'vibe coding,' where the friction between a thought and a live application is completely removed. This approach empowers a new generation of 'citizen developers' to build at the speed of thought. > *there should be no friction in between what your thought is and reality that kind of comes out.* > *[89, 4]* > *we've been very very deliberate to like invent our own infrastructure from scratch.* > *[88, 30]* ## Entities - **Jake Cooper** (person): CEO and 'Conductor' of Railway. - **Railway** (organization): A cloud platform designed for easy deployment and environment management. - **Uber** (organization): Jake's former employer where he worked on distributed systems for Jump bikes. - **Temporal** (software): A workflow orchestration platform used by Railway for reliable infrastructure tasks. - **Salesforce** (organization): The CRM company that acquired Heroku, leading to its perceived stagnation. - **Heroku** (organization): A pioneer PaaS platform that Railway is often compared to. - **AWS** (organization): Amazon Web Services, used by Railway for hybrid cloud bursting. - **GCP** (organization): Google Cloud Platform, one of the five clouds Railway straddles. - **Claude** (software): An AI model mentioned as an interface for deploying on Railway. - **GitHub** (organization): A code hosting platform discussed regarding its architectural flaws in versioning. - **Kubernetes** (software): An orchestration system Railway chooses to avoid for higher-order control. - **Central Station** (product): Railway's internal tool for aggregating user context and support feedback.

Ukraine produced 4 million FPV drones last year; China could produce 4 billion. That asymmetry frames two hours of unusually concrete conversation between Yaroslav Azhnyuk — serial tech founder turned AI-drone builder at The Fourth Law — and economist Noah Smith, who has been writing about the economics of drone warfare since before most Western policy circles took it seriously. They cover the full tech stack (cameras, autonomy modules, fiber optic links, interceptors, a semiconductor fab under construction), a five-level autonomy taxonomy, an eight-dimension autonomous-battlefield framework, and China's manufacturing edge that has no near-term Western answer. The through-line: the West is still planning to fight the last war, Ukraine is the defense valley where the next war is already live, and the gap is widening faster than most people realize. ## [00:00] Cold Open: China's 4 Billion Drones and the Cameras-to-Explosives Pipeline Yaroslav opens cold with a single arithmetic comparison that structures the rest of the episode. Ukraine, not an industrial powerhouse, built 4 million FPV drones in a year. China, with an order-of-magnitude larger manufacturing base and a consumer electronics supply chain already producing the same cameras, motors, and chips, could produce 4 billion. Noah immediately asks whether that makes China the supreme conventional military power on earth right now. Yaroslav won't claim certainty, but won't rule it out either. > *"I don't think we have all the information to claim that, but we cannot count it out. And that alone should be, you know, a big warning sign."* The cold open also plants the personal pivot that the rest of the episode unpacks: Yaroslav went from making cameras that fling treats to pets to cameras that fling explosives to occupiers. ## [01:04] Introduction: Brandon, Noah Smith, and Yaroslav Azhnyuk Guest host Brandon normally runs a science podcast; this episode is the exception. Noah Smith — Noahpinion Substack, economist focused on industrial policy and geopolitics — is co-host and co-interviewer. Yaroslav sets the personal context: on February 23rd, 2022, he and his then-fiancée landed in Kyiv at 11 p.m. on what turned out to be one of the last flights into the city. Eight hours later, the bombs fell. The 17-hour drive west that followed — empty streets, gas stations out of fuel, pouring diesel into windshield-washer canisters — reads like a scene from an apocalyptic film because, for the people living it, it was exactly that. > *"We basically packed our belongings and got in the car and spent 17 hours riding west. That was exactly like that. I, you know, missiles are falling, like there was smoke in Kyiv."* ## [05:41] From Tech Entrepreneur to Defense: PetCube, Brave One, and the D3 Fund Yaroslav's path from pet-tech to defense wasn't a straight line. In San Francisco from 2014 to 2020 building PetCube (one of the leading pet-camera companies), he had never taken military coursework and considered wars a thing of the past. Day one of the invasion he knew he would fight back with everything he could — but weapons weren't the first instinct. Early efforts included lobbying U.S. Congress on Lend-Lease (passed May 2022, underdelivered), co-founding Brave 1 (Ukraine's defense-innovation cluster, analogous to DIU), and helping seed the D3 Fund co-started by Eric Schmidt. By 2023, two things became undeniable: the war would last, and drones had permanently redefined warfare — the first software-defined weapon platform in history, where a battlefield capability upgrade can be pushed overnight like a software update. > *"It's like if you were able to push a software update and get all of your Roman legionaries a new helmet. That has never been possible before."* ## [10:42] The Ethics of Building Weapons: Dual-Use Technology and the Wolf at the Door Brandon raises the dual-use problem: the technology won't stay in Ukrainian hands. Yaroslav's answer is pragmatic rather than philosophical. Every technology from fire to large language models is dual-use; the question for a maker is whether the marginal risk of their contribution outweighs the immediate need. Ukraine is in a forest with a wolf. You deal with the wolf first, then consult Greenpeace. He's clear-eyed that no technology stays contained — the parallel concern about LLMs freely available in North Korea and Russia applies equally to drone autonomy — but frames his own company's responsibility narrowly: they supply to the Ukrainian government and armed forces, not to arbitrary buyers. > *"When you're in a situation where you're in a forest in front of a wolf, you know, you first going to deal with a wolf that wants to eat you and then you're going to go consult Greenpeace."* ## [14:01] The Tech Stack: Cameras, Autonomy Modules, Interceptors, and a Semiconductor Fab The Fourth Law's structure is three interlocking business units. Cameras (daytime and thermal, sold to 200+ Ukrainian drone manufacturers). Drone autonomy modules (sold to the same ecosystem). And UAV products sold direct to the armed forces: FPV strike drones, bombers, Shahed interceptors, and ISR interceptors — drones that hunt Russian reconnaissance drones before they can relay targeting data. The thermal-camera arm is about to start construction on two semiconductor fabs to manufacture sensor chips in-house, driven by the realization that dependence on foreign sensor supply chains is a strategic vulnerability. > *"We're about to start construction of two semiconductor plants to make sensors for thermal cameras. That's super exciting for me as a computer science guy — doing semiconductor, super cool."* ## [18:47] Fiber Optic vs. AI: The Radio Horizon Problem and $32/km Cable The chapter is really about why radio-only FPV drones fail at long range — not just from jamming, but from the curvature of the Earth. Below roughly 60-100 meters altitude at 30-40 km range, a drone enters a radio shadow behind hills, forests, or the horizon itself. The pilot loses video and control precisely when closing on a target that is, by definition, on the ground. Fiber optic cable ($32/km, spooled from the drone) solves the shadow problem but adds weight, limits range, and reduces maneuverability. AI fills the gap differently: terminal guidance lets the drone complete the last few hundred meters autonomously even after the radio link breaks. The two approaches aren't mutually exclusive — you can run AI on top of a fiber optic link to command hundreds of drones with fewer operators. > *"If your drone goes low — and usually Russian infantry and vehicles, they're on the ground and you want to hit them, you need to go low — lower you go, maybe you'll get behind a hill or behind a forest, and if you're far enough you'll just get behind the curvature of the Earth."* ## [25:32] FPV Drones: The New God of War — 70–80% of Frontline Casualties Artillery was historically called "the god of war" because it caused 80% of battlefield casualties. On the current Ukrainian front line, 70-80% of casualties are inflicted by FPV drones — the same fraction, a different weapon. Tanks, designed to dominate land warfare for decades, are now routinely destroyed by $400 consumer-grade quadcopters because armor was never built to defend against attacks from directly above. The trajectory follows the same curve as calculators becoming irrelevant once smartphones arrived: not a linear substitution but an exponential displacement where the new technology's influence grows nonlinearly. > *"They used to say that artillery is the god of war because artillery used to cause like 80% of casualties, and now on that ranking FPV drones rule."* ## [28:28] The Five Levels of Drone Autonomy: From Terminal Guidance to Full Autonomy Yaroslav lays out five autonomy levels describing where the field stands and where it's heading. Level 1 is terminal guidance — the drone flies under human control and locks onto a target only in the final seconds. Level 2 is bombing — dropping munitions from altitude without directly ramming a target. Levels 3-4 introduce increasing target-selection and navigation independence: the drone can identify radio-emitting equipment, track vehicles, or navigate through GPS-denied environments. Level 5 is full autonomy — launch-and-forget, no human in the loop for any mission phase. Current battlefield deployment sits mostly at Levels 1-3. The jump to higher levels isn't primarily a technical problem anymore; it's a deployment, doctrine, and trust problem. Human confirmation remains in the loop at every stage involving lethal targeting decisions — for now. > *"Technology progresses and its influence grows nonlinearly. It's all exponential."* ## [41:37] The Eight Dimensions of the Autonomous Battlefield The five autonomy levels describe a single drone's capability. The eight dimensions describe the full battlefield context those drones operate in. Dimension 1: level of autonomy (the five-level scale). Dimension 2: platform type (quadcopter, fixed-wing, missile, naval drone). Dimension 3: environment (day/night, urban/forest/open terrain). Dimension 4: target type (moving vehicle, static structure, radio emitter). Dimension 5: swarm size and coordination. Dimension 6: command-and-control architecture. Dimension 7: sensing modality (optical, thermal, RF). Dimension 8: infrastructure (simulation, data pipelines, security, deployment tooling). Each dimension interacts with every other. A Level-4 autonomous drone performing well in open daylight terrain may fail completely in a forest at night. Battlefield AI systems have to be evaluated across all eight dimensions simultaneously, not just on the single axis of autonomy level. > *"I say dimension because each of them works with another. It's crucial to understand how autonomy evolves in a modern battlefield environment."* ## [45:32] AI Safety and the Morality of Autonomous Weapons Yaroslav's position flips the standard AI-safety framing: in five to ten years, it will be *immoral* to use weapons *without* AI, because human-only weapons produce more collateral damage and friendly fire. He draws the analogy to manually driven cars — once autonomous vehicles are the norm, letting a human drive on a public road becomes the dangerous choice. Noah pushes to the logical endpoint: a Level-6 "AI general" — one large model that ingests all battlefield data and agentically selects targets, with humans reduced to repairing drones. Yaroslav says technically it could be done now. The constraint is deployment and trust, not capability. He references what was publicly described about AI-assisted target designation in the Iran operation: AI surfaces 127 targets, human reviews the list and presses okay. That's already close to an AI general with a rubber-stamp layer. > *"I think 5 to 10 years from now it will be immoral to use weapons without AI because weapons without AI will be more likely to cause collateral damage or unwanted damage."* ## [51:31] The End of the Rifleman? Noah's 2013 Prediction vs. Battlefield Reality Noah revisits a prediction he made in 2013: the rifleman is obsolete, replaced by standoff weapons. Ukraine both confirms and complicates it. FPV drones have unquestionably displaced the rifle as the primary instrument of attrition — but infantrymen haven't disappeared. They dig trenches, hold terrain, conduct logistics, and survive for months in dugouts under continuous drone threat by adapting: better camouflage, smaller movement signatures, drone-awareness drills. Yaroslav extends the timeline question to humanoid robots. The world is built for bipedal humans; there's genuine utility in a platform that can operate a rifle, open a door, or crew a vehicle. He puts a Terminator-style scenario — humanoid combat robots — at 10 years out, not science fiction. But modern warfare, they agree, is a multi-dimensional problem — dozens of drone types, land ops, reconnaissance, psychological operations, aviation, tanks, logistics — and the press focus on whichever technology is newest understates how much every layer still matters. > *"Modern warfare is really very complex and the fact that drones are the latest coolest thing doesn't mean that now it's that and only that."* ## [01:05:13] China's Manufacturing Advantage and Western Vulnerabilities This is where Noah Smith's economics background drives the conversation. The U.S.-China drone comparison isn't about unit price or autonomy level — it's about manufacturing throughput at scale. China's consumer electronics supply chain already produces the motors, cameras, chips, and battery cells that go into FPV drones. Switching that capacity to military production requires regulatory will, not retooling. Ukraine builds fixed-wing drones with 10 km range from hobby components; China can build fixed-wing drones with 200-300 km range at the same cost curve. The West's vulnerability isn't just quantity. It's thermal cameras (overwhelmingly sourced from China), semiconductor fabs (two generations behind on drone-relevant sensors), and procurement speed (a Western defense contract takes years to award; Ukraine iterates weekly). Yaroslav is optimistic about Western human capital — the engineers exist — but openly frustrated with European institutional inertia and uncertain about whether the U.S. has fully absorbed the lessons from Ukraine and the Middle East. > *"We don't have all the information to claim that, but we cannot count that out. If we want to keep the resemblance of our good past life, we have to do something about it."* ## [01:24:21] Policy Advice for Western Defense: Defense Valley and the Widening Gap Yaroslav's top policy prescriptions are framed around the William Gibson quote he attributes to Arthur C. Clarke: the future is already here, just not evenly distributed. Kyiv is Defense Valley — the place where the future of war arrived first, with hundreds of specialized companies, battle-tested commanders at every rank, and a government that learned to move at startup speed. Priority 1: deep integration with Ukraine's defense ecosystem, not just procurement but embedded learning. Priority 2: procurement reform — the drone-dominance initiative is the right direction and needs to scale 10x. Priority 3: long-range drone readiness for contested maritime environments (Shahed-class drones with 2,000 km range cover the entire Pacific island chain). He worries that the U.S. learned less from Ukraine than it should have and may be repeating the pattern with Iran. > *"Kyiv and Ukraine is sort of the defense valley. It's the point where the future of defense has already arrived, and there's a ton of things to learn from that."* ## [01:32:54] The Drone Race: Who's Ahead, Category by Category Russia was at parity or ahead in drone capability 18 months ago; Ukraine has since pulled ahead on FPV and autonomy. But Russia has a 4x population advantage and significantly more industrial capacity than Ukraine alone — scale disparity is why Western supply matters. The race breaks down by category: FPV strike (Ukraine leads), ISR reconnaissance (contested), glide bombs (Russia leads, dropping from bomber aircraft at scale), deep-strike drones (Russia leads on volume), and interceptors (Ukraine innovating rapidly, Russia catching up). Russia uses helicopters to intercept Ukrainian deep-strike drones — a costly but effective countermeasure revealing how each new offense spawns a tailored defense, at weekly iteration cycles. > *"Everyone says Russia's behind right now in the drone war. But that wasn't true a year ago."* ## [01:41:57] Countermeasures: Shotguns, Jammers, Lasers, and Fishnets Shotguns work — they're the primary kinetic countermeasure against incoming FPV drones — but only for a trained soldier who can hit a 20 cm target moving at 100 km/h under combat stress. Electronic jammers are the most widespread defense: block the radio or GPS link and the drone loses guidance. The catch is that the same spectrum the jammer blankets is often used by your own forces, and jammers are being defeated by frequency-hopping and fiber optic links. Russian tanks now look like porcupines — improvised metal cages and electronic-warfare antennas bolted on top to defeat top-attack drones. Ukraine's answer is shaped charges specifically tuned for the gap between the cage and the hull. Lasers are effective but expensive ($10M+ per system to kill a $400 drone) and slow to slew onto fast-moving targets. Fishnets — literally mesh nets — are being deployed around static positions because they're cheap, snag rotors, and require no power. > *"Then the tanks — if you look at Russian tanks and sometimes Ukrainian tanks or equipment — they all look like porcupines."* ## [01:58:19] The Wedding and Final Takeaway: Be Prepared for War Brandon closes with two questions. First: did Yaroslav actually get married in that chapel on February 23rd? They got legally married, but postponed the reception until the war is over. Second: one takeaway for the audience. Yaroslav's answer is a restatement of the Roman proverb: *si vis pacem, para bellum*. > *"You want peace, be prepared for war. Got to invest in defense and security."* ## Entities - **Yaroslav Azhnyuk** (Person): Founder of The Fourth Law (AI drone autonomy + thermal cameras, Ukraine); previously co-founder of PetCube; co-founder of Brave 1 and D3 Fund; born and raised in Kyiv. - **Noah Smith** (Person): Economist; author of the Noahpinion Substack; co-host for this episode; focus on industrial policy, manufacturing economics, and geopolitics. - **Brandon** (Person): Regular Latent Space host (science podcast background); guest host for this episode. - **The Fourth Law** (Organization): Yaroslav's AI-guided drone company; three business units — thermal cameras, drone autonomy modules, UAV products (FPV strike, bombers, interceptors). Leading drone-AI team in Ukraine. - **PetCube** (Organization): Consumer pet-camera company Yaroslav co-founded in San Francisco (2014–2020); the origin of the "cameras that fling treats / cameras that fling explosives" pivot. - **Brave 1** (Organization): Ukraine's defense-innovation cluster; analogous to DIU (Defense Innovation Unit) in the U.S.; co-founded with Yaroslav's involvement. - **D3 Fund** (Organization): Defense-tech investment fund co-founded with Eric Schmidt (ex-Google CEO) to accelerate Ukraine's drone ecosystem. - **FPV Drone** (Concept): First-Person-View drone — pilot sees through onboard camera in real time; currently responsible for 70-80% of frontline casualties; dominant tactical weapon of the Ukraine conflict. - **Five Levels of Drone Autonomy** (Concept): Yaroslav's taxonomy from terminal guidance (Level 1) to full autonomous operation (Level 5); most current battlefield deployment is Levels 1-3. - **Eight Dimensions of the Autonomous Battlefield** (Concept): Yaroslav's framework for evaluating drone systems across platform type, environment, target class, swarm scale, C2 architecture, sensing modality, and infrastructure. - **Defense Valley** (Concept): Yaroslav's term for Kyiv/Ukraine as the global hub where the future of defense tech is already live — analogous to Silicon Valley for consumer tech. - **Radio Horizon** (Concept): Earth-curvature effect that cuts radio/video links to low-flying FPV drones at 30-40 km range; primary technical driver for fiber optic drone adoption. - **Shahed** (Concept): Iranian-designed loitering munition used by Russia; fixed-wing, up to 2,000 km range; archetype for long-range drone threats to Western bases and Pacific-scenario planning.

Janie Lee y Chai Asawa de Abridge se suman a swyx y Jacob Effron de Redpoint en un cruce entre Latent Space × Unsupervised Learning para contar cómo un transcriptor AI se convirtió en la "capa de inteligencia clínica" de la sanidad. Recorren la filosofía de producto inspirada en el aire acondicionado, el caso de uso de la autorización previa, un stack de evals construido alrededor de científicos clínicos y jueces LLM, por qué HIPAA remodela el volante de datos, y qué se necesita para funcionar de forma fiable en más de 100 millones de conversaciones médicas. ## [00:00] Introducción El episodio abre con el pitch de Janie Lee: el contexto lo es todo, las alertas deben pasar de reactivas a proactivas, y el producto debe desvanecerse en el fondo como el aire acondicionado hasta que un riesgo clínico justifique la interrupción. swyx aprovecha para hacer un breve llamado a los oyentes a suscribirse. > *"Una cosa que nos gusta decir es que queremos que nuestro producto se sienta como el aire acondicionado. Debería estar de fondo, simplemente mejorando las cosas."* — Janie Lee ## [01:17] Qué hace Abridge swyx presenta el episodio como el cruce anual Latent Space × Unsupervised Learning, con Jacob Effron como invitado dado que Redpoint es inversor de Abridge. Janie describe Abridge como una capa de inteligencia clínica para sistemas de salud, que comenzó con la documentación: los clínicos dedican entre 10 y 20 horas semanales a redactar notas, y la conversación entre paciente y médico está aguas arriba de casi todos los artefactos posteriores: el expediente, el pago, el diagnóstico. Chai añade que todo lo que ocurre antes, durante y después de la visita se vuelve abordable cuando se tiene contexto completo sobre pacientes, pagadores, guías clínicas y literatura médica. > *"Abridge es una capa de inteligencia clínica para sistemas de salud. Realmente empezamos con la documentación y construyendo para los clínicos."* — Janie Lee ## [03:22] De la documentación ambiental a la inteligencia clínica Janie describe los tres "actos" de Abridge: ahorrar tiempo (el producto de transcripción original que devolvió las tardes a los médicos, el "pajama time"), ahorrar y generar dinero para sistemas sanitarios que operan con márgenes históricamente bajos, y en última instancia salvar vidas. Que el producto se abra millones de veces a la semana, antes, durante y después de cada visita, es lo que hace viable la expansión. > *"Le llaman 'pajama time'... los médicos después del trabajo, en pijama en casa, escribiendo y al día con sus notas cada día."* — Janie Lee ## [05:21] Apoyo a la decisión clínica y el contexto como eje central Jacob le pregunta a Chai cómo el apoyo a la decisión clínica de Abridge se compara con su trabajo anterior en Glean. Chai traza la diferencia: en Glean una respuesta incorrecta es molesta; en sanidad es de alto riesgo y la superficie de usuario es mucho más estrecha, menos personas, pero cada resultado tiene que acertar. Eso moldea todo, desde la evaluación offline hasta el despliegue progresivo, y enlaza con la visión tipo Jarvis de un asistente que realmente te conoce, la misma que cada hackathon de la última década intentó construir. > *"La visión Jarvis, esa que en cada hackathon al que fui durante la última década siempre tenía un competidor al estilo Jarvis, pero creo que Abridge empezó exactamente desde esa oportunidad y sigue avanzando en esa dirección."* — Chai Asawa ## [08:14] Fatiga por alertas, inteligencia proactiva y autorización previa Jacob plantea el clásico problema de la fatiga por alertas: cómo decidir cuándo romper el silencio del aire acondicionado e interrumpir. El ejemplo concreto de Janie es la autorización previa: un rechazo de una MRI que hoy llega semanas después puede convertirse en un aviso en tiempo real mientras el paciente sigue en consulta, condicionado a las políticas del pagador, los datos del EHR, diagnósticos anteriores y protocolos específicos del centro. La dificultad está en la fontanería de datos: la autorización previa solo funciona si el asistente puede unir cada señal relevante en el momento preciso. > *"Para que ese ejemplo de autorización previa sea posible, piensa en todos los datos que necesitas tener."* — Janie Lee ## [13:53] Formatos del AI ambiental y clientes en sanidad swyx pregunta por los formatos. Hoy la superficie principal es el móvil, pero Abridge también funciona en escritorio, plugins de navegador dentro del EHR, dispositivos en sala para pacientes ingresados, flujos de trabajo de enfermería, y empieza a explorar la realidad aumentada. El cliente es múltiple: CMIOs, CFOs, CIOs, clínicos, pacientes, pagadores y farmacéuticas están en algún punto del ciclo, con las interacciones de los pagadores a través de intercambios estructurados en lugar de acceso directo a los datos brutos de Abridge. > *"Habláis mucho de AI ambiental. ¿Es principalmente en el teléfono?"* — swyx ## [18:16] Los problemas de IA más difíciles en sanidad Ante la pregunta de cuál es el problema de IA más difícil en Abridge, Chai elige: soporte en tiempo real de alta calidad, baja latencia y bajo coste en un entorno clínico de alto riesgo. Modelar la larga cola de políticas de pagadores en representaciones intermedias sobre las que el sistema pueda razonar es un ejemplo concreto. La frontera de Pareto no deja de moverse, y tienen que empujarla ellos mismos en lugar de esperar a mejoras genéricas. > *"Y por supuesto la frontera de Pareto siempre está cambiando, pero también intentamos hacerlo ahora."* — Chai Asawa ## [19:43] Modelos de frontera, datos propietarios y estrategia de modelos Jacob pregunta qué usan tal cual frente a lo que construyen. El enfoque de Chai: los modelos de frontera siguen absorbiendo conocimiento médico general, así que la ventaja de Abridge reside en los datos propietarios de conversaciones médicas y en el comportamiento específico por especialidad construido encima. Son explícitamente agnósticos al modelo cuando pueden serlo: lo que importa es la experiencia final del producto, y combinan lo mejor según el flujo de trabajo. > *"Podemos usar algo para esto y aquello, y solo nos importa, al final del día, la mejor experiencia de producto."* — Chai Asawa ## [22:24] El EHR como sistema de archivos para agentes El enfoque de Chai para el próximo año: en el fondo, todo agente es un agente de código, y dentro de la sanidad el EHR funciona como el sistema de archivos: un gran almacén de información estructurada que no cabe en la ventana de contexto de ningún modelo actual. Janie añade que el objetivo sigue siendo mantener al clínico centrado en el paciente: tener el contexto adecuado listo en el momento preciso, sin revivir la conversación. > *"Casi todo agente es en el fondo un agente de código. Le das un sistema de archivos, puede escribir su propio código... puedes pensar en el EHR exactamente como un sistema de archivos."* — Chai Asawa ## [25:20] Personalización, memoria y preferencias del clínico Jacob pregunta cómo Abridge gestiona la personalización por médico. La respuesta de Janie es en capas: las ediciones individuales se convierten en señal, los valores predeterminados por especialidad se superponen, y las políticas del sistema sanitario lo envuelven todo. Chai habla de la memoria como un nuevo tipo de sistema de registro: tareas en segundo plano que consolidan señales entre visitas, similar a cómo el sueño consolida la memoria en las personas, de modo que el modelo "aprende" de cada edición y de cada no-edición. > *"Una de las cosas interesantes que también producimos es que la memoria es como uno de esos nuevos sistemas de registro."* — Chai Asawa ## [31:57] Evals, jueces LLM y despliegue progresivo Janie describe el stack de evals: clínicos internos realizan la revisión de primer paso LFD, los jueces LLM se calibran con esos datos anotados, evaluadores externos aportan una lectura independiente, y los evals por especialidad capturan lo que los genéricos pasan por alto. Chai añade una analogía con los coches autónomos: quieren contacto con la realidad lo antes posible, pero solo mediante despliegue progresivo, para que la distribución offline refleje realmente la distribución de producción. > *"Quiero entrar en contacto con la realidad lo antes posible, pero quiero un despliegue progresivo, porque por mucho que tenga un conjunto de evals offline, quiero que su distribución coincida realmente con la distribución de la vida real."* — Chai Asawa ## [38:04] HIPAA, desidentificación y privacidad La privacidad se trata como una restricción dura sobre el volante de datos. Chai explica que cualquier dato usado como base de evals online o de aprendizaje tiene que estar desidentificado de forma irreversible, y han diseñado procesos en torno a eso. Janie añade que los contratos con clientes también condicionan quién dentro de Abridge puede acceder a PHI, por lo que el listón para lo que vuelve a los datos de entrenamiento es contractualmente alto, no solo por política interna. > *"Cualquier dato que usemos necesita estar desidentificado; cualquier dato real que usemos como base de conjuntos de evals online o de aprendizaje, y por eso hay que..."* — Chai Asawa ## [40:38] 100 millones de conversaciones y operación a escala Con más de 100 millones de conversaciones, la superficie de trabajo cambia: el enrutamiento de modelos, el post-entrenamiento, los presupuestos de fiabilidad y el coste por llamada se convierten en preocupaciones de primer orden. Chai habla de los insights que pueden ofrecerse a los clínicos, y extiende la mirada hacia adelante: eventualmente esa misma conversación podría generar señales para pacientes y consumidores directamente, no solo para proveedores. > *"Hay tantísimo en nuestro conjunto de datos de cien millones de conversaciones. Puedes imaginar cosas como insights que puedes dar al clínico."* — Chai Asawa ## [45:27] Integración con el EHR y la capa de inteligencia clínica swyx pregunta sobre la relación con el EHR. Abridge invierte mucho en interoperabilidad profunda: la asociación con el EHR es el mínimo necesario para la adopción por parte de los clínicos, pero el valor que Abridge añade encima opera en otro alcance: contexto entre visitas, razonamiento consciente del pagador y el tipo de inteligencia clínica que el propio EHR no está estructurado para producir. > *"Uno de los socios clave es el EHR, y me pregunto cómo es esa relación."* — swyx ## [47:56] Regulación sanitaria, latencia e IA de alto riesgo Jacob pregunta qué ha aprendido Abridge de la regulación. Janie rebate la narrativa habitual: la IA sanitaria tiene en realidad vientos regulatorios a favor, porque el listón es tan alto que los problemas más difíciles acaban resolviéndose aquí primero. Chai explica los "trucos inteligentes" que lanzan hoy sabiendo que la frontera seguirá avanzando, y aceptando que algunos de esos trucos no sobrevivirán cinco años. > *"Creo que es donde algunos de los problemas de IA más difíciles se resolverán primero, precisamente porque el listón es tan alto."* — Janie Lee ## [51:28] Científicos clínicos y calidad en la larga cola Janie describe un rol interno de Abridge llamado el científico clínico: médicos que también son técnicos, desde ingenieros full-stack hasta "prompters increíblemente hábiles". Tenerlos integrados en los equipos de producto y evaluación eleva el listón de lo que se lanza, porque las personas que escriben los criterios LFD son las que realmente entienden qué significa ser clínicamente útil. swyx lo conecta con el aprendizaje activo sobre puntos débiles conocidos, el tipo de pulido que se ha perdido en la mayoría de los equipos de IA. > *"Tenemos un rol llamado el científico clínico, y creo que escuché a uno de nuestros líderes referirse a ellos recientemente como mutantes."* — Janie Lee ## [54:21] Lecciones de Glean e infraestructura de IA duradera Jacob pregunta a Chai qué se traslada de Glean. La respuesta gira principalmente en torno a lo que resiste el paso del tiempo: capas de contexto, sistemas orientados a eventos, Kafka, Temporal, sockets, CRDTs del manual de colaboración de Google Docs. Los sistemas multiagente heredan los mismos problemas de resolución de conflictos que tienen los humanos, y los patrones de infraestructura de la última década no se están descartando, sino reutilizando. > *"Hay mucha tecnología orientada a eventos... ya sea Kafka, Temporal, sockets y demás; cómo integrar todo eso es algo que también creo que es duradero."* — Chai Asawa ## [58:20] El futuro de los flujos de trabajo agentivos en sanidad Un breve intercambio sobre cómo sería un Abridge más agentivo: sigue anclado en el papel del clínico en la relación con el paciente, pero con más trabajo en segundo plano, reaccionando a resultados de laboratorio, redactando seguimientos, asumiendo capacidades en nombre del clínico sin apropiarse de la relación. > *"Incluso más capacidades en nombre del clínico, que creemos tiene un papel fundamental en la conexión con el paciente."* — Chai Asawa ## [58:51] PRDs, claridad de producto y construcción de productos AI serios La pregunta rápida de Jacob: en qué has cambiado de opinión sobre IA en el último año. Janie rebate la visión popular: los prototipos no lo son todo, los PRDs no están muertos. A medida que los productos se vuelven más complejos e impulsados por IA, la disciplina de claridad escrita de un PRD real importa más, no menos. El resto de la sección trata sobre cómo construir productos AI serios en sanidad: propiedad, disciplina de especificación escrita y resistencia al desarrollo impulsado por demos. > *"La opinión más provocadora es que los prototipos son el fin último y que los PRDs están muertos."* — Janie Lee (la opinión que cambió) ## [64:28] Herramientas de programación con IA en Abridge La pregunta de cierre habitual de swyx. Abridge usa Claude Code y Cursor internamente, y Jacob lanza una broma: le gustaría ver a Claude dirigir una empresa valorada en mil millones de dólares antes de generar ingresos. > *"Claude va a hacer esto... me gustaría ver a Claude... dirigir una empresa a mil millones de dólares antes de tener ingresos."* — Jacob Effron ## [65:23] Cierre Chai invita a los oyentes a visitar el sitio web de Abridge para acceder a sus artículos técnicos sobre reducción de alucinaciones, evals y el resto del stack de investigación. swyx y Jacob cierran con agradecimientos. > *"En el sitio web de Abridge tenemos muchos de nuestros artículos técnicos donde hemos hecho trabajo muy interesante, como reducir las alucinaciones."* — Chai Asawa ## Entidades - **Janie Lee** (Persona): Operadora de la era fundacional de Abridge; responsable del área de producto y comercial de la capa de inteligencia clínica. - **Chai Asawa** (Persona): Responsable de apoyo a la decisión clínica en Abridge; anteriormente en Glean. - **swyx** (Persona): Presentador de Latent Space. - **Jacob Effron** (Persona): Socio en Redpoint Ventures; presentador del podcast Unsupervised Learning. - **Abridge** (Organización): Empresa de IA sanitaria que construye la capa de inteligencia clínica, comenzando con documentación ambiental y expandiéndose hacia apoyo a la decisión, autorización previa, evals e integración con EHR. - **Glean** (Organización): Empresa de búsqueda empresarial con IA; referenciada como el empleador anterior de Chai y como contraste horizontal frente al enfoque vertical de Abridge. - **Redpoint Ventures** (Organización): Firma de capital riesgo; inversora en Abridge y sede del cruce con Unsupervised Learning. - **EHR (Electronic Health Record)** (Concepto): El sistema de registro central en el que operan los sistemas sanitarios; en palabras de Chai, el EHR funciona como un sistema de archivos para los agentes sanitarios. - **Autorización previa** (Concepto): Un caso de uso central de Abridge: convertir los rechazos de pagadores que hoy tardan semanas en orientación en tiempo real durante la visita. - **Proceso LFD** (Concepto): Revisión interna de Abridge liderada por clínicos como primer paso, usada para calibrar los jueces LLM y definir los criterios de evaluación. - **Científico clínico** (Concepto): Un rol de Abridge: médicos con perfil técnico, integrados en los equipos de producto y evaluación. - **Despliegue progresivo** (Concepto): La disciplina de despliegue de Abridge: lanzar a una fracción del tráfico real para mantener honesta la distribución offline, siguiendo el patrón de lanzamiento de los vehículos autónomos. - **Claude Code** (Software): Herramienta de programación con IA usada internamente en Abridge. - **Cursor** (Software): Editor de código con IA también usado internamente en Abridge.

Matt Pocock joins swyx at AI Engineer Europe to argue that the old software design canon — DDD, deep modules, ubiquitous language — matters more, not less, in the AI coding era. The thesis: code is not just a compile target; a codebase that is easy for humans to change is easy for AI to change. Along the way they cover course-making, why traditional lectures still beat AI-native learning, and TypeScript's quiet takeover of AI engineering. ## [00:04] Opening at AIE Europe and the Cursed Course swyx welcomes Matt to the AI Engineer Europe podcast booth in London. Matt jokes that AIE is "the worst" event he has ever attended (the location is in fact astonishing) before turning to his Claude Code course, which is just wrapping up its two-week cohort. He explains why he runs short cohorts: AI moves so fast that self-paced courses cannot guarantee updates, and the "curse" of releasing into breaking changes — AI SDK v5 dropped on day two of his AI SDK v4 course, and the Claude Code source leaked during this one — is now baked in. The conversation then turns to teaching as a craft. Matt rejects the "pundit" branch of YouTuber identity — he is not trying to predict the future, only to teach durable material — and notes that being a teacher first is what differentiates his content. > *I'm not a guy who's trying to predict the future. I'm just trying to teach.* ## [02:51] Why Engineering Fundamentals Matter More with AI Matt previews his AIE talk. The popular narrative says code no longer matters because English plus an AI compiler can produce applications. Every time he tried to ignore the code, he ended up with "a terrible mess." So he went back to the classics — *Extreme Programming*, *The Pragmatic Programmer*, *A Philosophy of Software Design*, DDD — and discovered they ported directly into prompts. Keeping the architecture in your head, even when you delegate implementation, yields outsized dividends. > *If you have a code base that's easy to change for humans, it's going to be easy for AI to change, too.* ## [04:23] Narrow Waist and Deep Modules swyx introduces the "narrow waist" concept from internet architecture (TCP/IP, HTTP at layers 3–4) as a way to contain AI-generated slop: define rigid interfaces, delegate the inside. He extends it to running AIE as a nine-person business — "model-view-claw" instead of MVC, where coordination across people and AI is the real systems problem. Matt maps this onto John Ousterhout's notion of *deep modules*: a large amount of functionality behind a simple interface, ports and adapters style. This is, in his experience, the best way to use AI for coding — be intentional about the interface as a human, then delegate the implementation. > *Deep modules basically — a large amount of functionality with a simple interface. Kind of ports and adapters, right?* ## [06:37] Domain-Driven Design Meets AI DDD is having a moment, and Matt argues it works *because* the framework has been around long enough to sit in the latent space of these models. You do not have to invent new vocabulary; you can bolt on a system that is composable and that the model already understands. The deeper point: DDD is fundamentally about aligning code with language, which is exactly what you want when speaking to an AI. He makes it concrete with the `mattpocock/skills` repo (≈13k stars) and its "ubiquitous language" skill — a Claude Code skill that scans your codebase, surfaces the arcane jargon, and refines it with you into a markdown file he keeps open while prompting. He references it from `agents.md` but does not paste it wholesale, so the agent finds it when searching for those terms. > *Essentially, you're trying to create a unified domain model so that the AI and you are speaking the same language.* ## [10:05] Teaching as an Overpowered Skill swyx asks how Matt got so good at explaining things. Matt credits six years as a voice coach before becoming a developer — communication felt like an unfair advantage when he started as a junior. He has since narrowed his focus: split time between learning material and finding the right phrases for it. The old texts help because they give him pre-built mental models to explain new ideas through. He walks through his course-making process: an "explore and exploit" phase, a Zettelkasten-style Obsidian vault, a custom planning app, P1/P2/P3 prioritization, and the rule that *each lesson teaches exactly one thing* with dependencies made explicit. Most of what he produces ends up on the cutting room floor. > *The ability to communicate always just felt like a ridiculous overpowered skill that I had in my locker that no one else had.* ## [13:20] How People Actually Learn AI Engineering The conversation turns to whether AI has changed how people learn. Matt distinguishes knowledge (lectures), skills (interactive exercises), and wisdom (small-group discussion — and now, talking to an AI). Counterintuitively, the more he leans into AI-experimental teaching, the more it turns his audience off. Most learners still want traditional lectures; swyx recalls Maven's cohort-based education arc landing in the same place. Matt's compromise is to force the work without forcing the form: in his TypeScript material he throws learners into a problem first and gives them the knowledge afterwards. > *The more I lean into the kind of AI experimental stuff, the more it actually turns people off my materials.* ## [15:04] TypeScript Overtaking Python swyx flags that TypeScript overtook Python in the GitHub survey this year — a shift he did not see coming, particularly in AI engineering where Python's expressiveness has been dominant on the backend. Matt's echo chamber is 100% TypeScript, but his real argument is ecosystem: when you care about UX and shipping chat-style applications, the framework gravity is in TypeScript (Vercel's Next.js, Cloudflare's variants). swyx admits this would meaningfully change which frameworks he promotes. > *If you're concerned about UX, concerned about shipping great stuff, you're mostly doing it in TypeScript.* ## [16:45] Inversion of Control and Composable Skills Matt looks ahead. His TypeScript-evals bet (Everlight) stalled — "no one's excited to do evals." The next frontier is *inversion of control*: as coding agents converge on similar architectures (Firebase-style backends, small tool sets), the interesting axis becomes how much control sits with the developer versus the harness. Claude Code's opacity buys ease of use but loses observability; Pydantic AI ("Pi") swings the other way — total control, total maintenance burden. He closes by pointing past coding agents entirely. Software engineers are a step ahead because AI produces quality output in their domain, but the composable skills he authors — like his three-sentence "grill me" skill that makes the AI interrogate you until you reach a shared understanding — generalize to any domain where you want the AI aligned with you. > *The inversion of control is going to be really important — you put more control in the hands of the developer and less in the harness.* ## Entities - **Matt Pocock** (Person): Creator of Total TypeScript and AI Hero; teaches TypeScript and AI Engineering through two-week cohort courses. - **Shawn Wang / swyx** (Person): Host; founder of AI Engineer and the AIE conference series. - **AI Engineer Europe (AIE)** (Organization): The London conference where this conversation was recorded; Matt's talk hit 1M views in 13 days — fastest in AIE history. - **AI Hero** (Organization): Matt's AI engineering education platform (aihero.dev). - **Claude Code** (Software): Anthropic's coding agent; subject of Matt's just-finished course and a recurring example throughout. - **Domain-Driven Design (DDD)** (Concept): Software methodology centered on aligning code with the language of the business domain; Matt argues it ports cleanly into AI prompting. - **Ubiquitous Language** (Concept): DDD practice of maintaining a shared vocabulary doc; Matt's namesake Claude Code skill scans a repo and refines this with the user. - **Deep Modules / Narrow Waist** (Concept): Architectural pattern (Ousterhout / internet protocols) of large functionality behind a small interface — Matt's preferred shape for AI-assisted codebases. - **mattpocock/skills** (Software): Matt's open-source repository of Claude Code skills; ≈13k stars at recording time. - **Pydantic AI (Pi)** (Software): Python agent framework built from low-level primitives; cited as the high-control counterpoint to Claude Code's opaque harness. - **Obsidian** (Software): Note-taking app reportedly run by a team of four; the example for non-engineering domains where AI leverage compounds.

Alex Lupsasca — 2024 New Horizons Breakthrough Prize winner and OpenAI resident scientist — recounts how GPT-5 resolved a year-long open problem in quantum field theory: proving that single-minus gluon tree amplitudes are non-zero and finding their compact closed form. He then describes how the publicly available GPT Pro, given the gluon paper as a seed, independently generalized the result to graviton amplitudes in under three days of human clock time. Throughout the conversation, Lupsasca reflects on what this trajectory means for how physics is done, how the next generation of physicists will be trained, and where the remaining bottlenecks — verification, creativity, and publishing infrastructure — still lie. ## [00:00] Introduction to AI's impact on physics research Lupsasca opens in medias res, framing the episode's central claim before the formal introduction: AI has crossed a threshold where it can resolve questions that stumped human experts for over a year. He describes this not as a curiosity for theoretical physicists but as a profound, if underappreciated, change in the nature of scientific discovery itself. > *"That's a certain milestone that we've passed, and I think maybe for the average person on the street who doesn't care about theoretical physics, this is not very noticeable, but I think it's a very profound change and we've really passed some kind of a threshold."* ## [00:43] Guest introduction: Alex Luposka The hosts — Brandon (Atomic AI) and RJ Honicky (Miro Omix) — introduce Lupsasca as a Vanderbilt professor and OpenAI fellow who holds both the 2024 New Horizons in Physics Breakthrough Prize (often called the "Oscars for science") and the IUPAP Young Scientist Award. Lupsasca immediately sets the narrative arc: a year ago, AI was useful for email but not for his work; ChatGPT o3 was the first model that genuinely helped with research math; then GPT-5 reproduced one of his hardest published results in 30 minutes. > *"When GPT-5 came out it was able to reproduce one of my best papers that took me a very long time to come up with in like 30 minutes. And that's when I really became AI pilled."* ## [02:49] Alex joining OpenAI and the shift in physics research After GPT-5's release, Lupsasca began evangelizing the shift to colleagues who were skeptical. Finding OpenAI equally excited, and being on sabbatical, he joined as resident scientist — the person physicists around the world now email when something astonishing happens. He describes receiving an inbound that week about Codex simulating the Sachdev-Ye-Kitaev (SYK) model in 10 minutes, a feat that many research groups had struggled to achieve due to the narrow Venn diagram of physicists with strong coding skills. > *"I talked to OpenAI. They were also really excited and I thought I have to get in on this and to understand that this is happening and not be a part of it is a huge mistake so I have to go to OpenAI."* ## [04:08] The release of GPT-5 and the shift in capabilities Lupsasca contrasts the lukewarm Twitter reception of GPT-5 (complaints that it was not better at writing email) with what he observed at the science frontier. He notes GPT-5.4 is another significant jump, and describes how AI capabilities for physics have been accelerating rapidly since o3, the first reasoning model strong enough for research-grade mathematics. He uses this as a bridge to the central technical story of the episode: a pair of new papers on gluon and graviton scattering amplitudes. > *"At the science frontier the capabilities were really taking off."* ## [10:05] Explaining Quantum Field Theory and amplitude calculations Lupsasca gives an accessible primer on quantum field theory (QFT), the framework that reconciles special relativity and quantum mechanics. The key objects in QFT are scattering amplitudes — complex-valued functions that encode the quantum probability for a set of incoming particles (with given energies, momenta, and polarizations) to scatter into a set of outgoing particles. These amplitudes are computed at particle colliders like the LHC, and knowing the n-point amplitude (for any number n of particles) encodes essentially the full content of the theory. > *"If you have a particular force and you're able to compute the n-point amplitudes... you know everything about the theory."* ## [14:20] Overview of gluons and the strong force Gluons are the force-carrying particles of the strong nuclear force — the force that, despite like-charge repulsion between protons, holds the atomic nucleus together. They are the QFT analog of photons for electromagnetism and gravitons for gravity. Like photons, gluons carry a polarization (helicity): positive (right-handed) or negative (left-handed). This helicity structure is central to the paper discussed next. > *"The strong force is mediated by the exchange of the particles of the strong force, which are called gluons, because they're what glues together the nucleus of the atom."* ## [14:38] Discussing the first research paper on single-minus gluon tree amplitudes Lupsasca unpacks the paper's title — "Single-Minus Gluon Tree Amplitudes Are Non-Zero" — piece by piece. Tree amplitudes are the leading-order (no-loop) contributions to scattering. All-plus-helicity amplitudes are exactly zero by a symmetry argument. Single-minus amplitudes — where all but one gluon have positive helicity — were assumed in textbooks to also be zero by the same argument. The paper proves they are not. The result involves collaboration with Alfredo Guevara (IAS), David Skinner (Cambridge), Andrew Strominger (Harvard), and Kevin Wheel. > *"If you look at the lecture notes and textbooks that have been written on this, the same argument that rules out the all-plus amplitudes also appears to rule out the single-minus amplitudes."* ## [20:56] How ChatGPT helped solve a year-long physics puzzle Strominger, Guevara, and Skinner had understood for about a year that the textbook argument has a loophole: when particles are collinear (exactly aligned in momentum), the standard dimensional-analysis reasoning fails, and single-minus amplitudes can be non-zero. But computing what those non-zero amplitudes equal had eluded them. Lupsasca invited Strominger to visit OpenAI and work on it with AI. The week before Strominger's flight, Lupsasca began using ChatGPT Pro. By the time Strominger landed, they had the answer. > *"Using ChatGPT we solved the problem before he even got off the plane."* ## [23:02] Complexity of manual calculations in physics Lupsasca shows the audience a concrete illustration of the difficulty: the six-point single-minus amplitude, worked out by hand by Alfredo Guevara, is a sum of 32 terms each of which is itself a product of four complicated factors. The number of terms grows factorially with the number of particles n — super-exponential growth. This is the messy representation that the group had been staring at for a year, seeking the analog of the elegant Parke-Taylor formula. > *"By the time you get to six terms, it explodes in your face."* ## [26:12] The history and mechanics of Feynman diagrams Feynman diagrams are a visual language introduced by Richard Feynman to organize perturbative QFT calculations: diagrams represent possible intermediate histories of a scattering process, and the full amplitude is a sum over all of them. Diagrams are organized by number of vertices (interaction points); each additional vertex is suppressed by the coupling constant, so tree diagrams (fewest vertices) dominate. Loop diagrams — where intermediate particles are created and annihilated — contribute smaller corrections. The combinatorial explosion of tree diagrams is the root cause of factorial growth. > *"In principle, there are infinitely many pictures to sum over."* ## [27:44] The Parke-Taylor formula and the quest for simplification In the 1980s, Parke and Taylor computed the "maximally helicity violating" (MHV, or double-minus) gluon amplitudes through a heroic Feynman diagram expansion. Despite the factorial number of terms, everything canceled to leave a single compact formula — the Parke-Taylor formula — that fits in half a line. Strominger, Guevara, and Skinner spent a year looking for the analogous compact formula for the single-minus case. Their search stalled at the level of the messy Feynman representation. > *"Andy, Alfredo and David spent the last year chasing the analog of the Parke-Taylor formula, the very simple answer that was obtained in the '80s for the double minus amplitudes."* ## [31:26] Using ChatGPT to find the simplification in the special phase space region When the five-point single-minus amplitude was fed to ChatGPT Pro, the model identified a special subregion of phase space (where one particle's frequency has opposite sign) in which the amplitude simplifies from eight terms to a product of just three. This appears not to have been a known fact; the model wrote Python code and tested thousands of possibilities to deduce it. Moving to the six-point amplitude (Guevara's hand calculation), ChatGPT simplified 32 terms to a product of 4. It then conjectured the general n-point formula — with only linear growth in the number of terms, the best possible behavior. GPT-5.2 Pro guessed the formula but could not prove it. > *"The formula that it proposed, instead of having this factorial growth... here it's actually linear. So if you double the number of particles, you only double the number of terms."* ## [38:07] Proving the formula from scratch to ensure validity To obtain a proof, Lupsasca used an internal OpenAI model with extended reasoning. He gave it the problem cold — without the conjectured formula — and asked it to find the general answer in the special phase-space region. After 12 hours of computation, the model independently rediscovered the same formula and produced a complete three-step proof. The proof constitutes the bulk of the published paper. The team kept the AI attribution to one paragraph, framing the paper as a physics result that stands on its own merits. > *"We gave it the whole problem from scratch... and it came back with the same formula which we had not given it. So it rediscovered the correct formula. But this time it also found the proof."* ## [41:00] Determining the scientific impact and future research Asked to compare the result to the Parke-Taylor formula, Lupsasca is candid that scientific impact is only assessable decades later, but argues the result is genuinely surprising and should open a line of attack toward deeper questions in quantum gravity. The conversation pivots naturally to the second paper. > *"I think the true value of a paper can only be assessed decades into the future based on how much future work it leads to and what developments it opens up."* ## [42:27] Introduction to the second paper on graviton amplitudes Gravitons are the hypothetical quanta of gravity — the spin-2 force carrier analogous to the spin-1 photon (electromagnetism) and gluon (strong force). Unlike gluons, gravitons have never been directly detected, but they are central to quantum gravity theory. The second paper, "Single-Minus Graviton Tree Amplitudes Are Non-Zero," shows the same loophole applies to gravity and that a compact formula extends there too — despite gravitons being mathematically more complex than gluons. > *"We wrote this paper which is called single minus graviton tree amplitudes are non-zero. So it's the same title almost, except with graviton instead of gluon."* ## [45:41] Defining particles, irreducible representations, and symmetry Lupsasca sketches the modern QFT definition of a particle (an irreducible representation of the Poincaré group, classified by Wigner according to mass, spin, and charge) and explains why gravitons are spin-2 while gluons and photons are spin-1, making graviton polarization data twice as rich. Crucially, the second paper was complete within three days of the first going public — most elapsed time was spent verifying correctness, not computing. > *"Most of the time was spent verifying the answer, not writing, which is insane, actually, if you take a step back."* ## [47:46] How GPT Pro generalized the research to gravity For the graviton paper, no internal model was needed — the publicly available ChatGPT GPT-5.2 Pro sufficed. Lupsasca provided the gluon paper as context plus two paragraphs describing the key mathematical changes, then said "Good luck. You're a brilliant theoretical physicist." Over a 110-page exchange, the model worked through the graviton calculation — applying the directed matrix tree theorem, a piece of known combinatorics that neither Lupsasca nor collaborators had thought to invoke — produced correct intermediate results, and wrote a draft paper very close to the final arXiv version from section 3 onward. > *"It's a real solid result in quantum gravity that was done pretty much completely by an AI with human steering it and asking kind of the right questions."* ## [53:57] The epistemological shift: Is this a new way of doing physics? The hosts raise the central epistemological question: if an undergraduate with domain knowledge and good prompting could have done this, what does graduate training mean now? Lupsasca agrees this is the hardest open question facing academia. He notes that arduous calculation trains not just skill but self-confidence, that the gap between coursework and the research frontier is growing, and that many "easy" problems professors once assigned to students are now solvable by AI in minutes. He offers two concrete ways AI has already changed his own workflow: dramatically reducing time spent confused between steps, and enabling parallel AI scouts that explore multiple research directions simultaneously. > *"With AI, actually, you can launch 10 instances of chat and have each one try a different route and send it as a scout that moves very fast into the unknown."* ## [59:27] The use of AI as a 'scout' for research directions Lupsasca elaborates on the scout metaphor: rather than carefully mapping a route from A to C before committing to it, a researcher can now dispatch many AI "scouts" in parallel, get rapid feedback on which directions are promising, and redirect human attention accordingly. Even when a scout makes errors, its signposts reduce orientation cost for the human following. This constitutes a qualitatively new mode of research — one where the bottleneck shifts from calculation to judgment about which directions matter. > *"Even if ChatGPT doesn't always get everything right, just kind of having a scout that signposts some key steps along the way that you can use to anchor your own movement is extremely helpful."* ## [61:44] The role of 'taste' and collaboration with AI The hosts push on the problem of "taste" — the ability to identify which questions are at the productive edge of knowledge. Lupsasca argues that working effectively with ChatGPT requires the same skill a professor develops advising students: knowing what question to give, at what level of detail. "Taste" — knowing where the frontier is and which questions there are tractable — is the last skill to develop and the one AI currently lacks. AI is, he says, like an extremely technically skilled graduate student: given a sharp, well-posed question, it can do incredibly hard computations correctly, but it does not yet know which question to ask. > *"The difference between a good physicist and a great physicist is knowing what is the right question to ask — that is actually the hardest part of being a scientist."* ## [70:23] Personal evolution from AI skeptic to resident scientist Lupsasca recapitulates his personal arc: skeptic → converted by o3 (which solved in 11 minutes a calculation that would have taken him days) → "AI-pilled" by GPT-5 (which reproduced, in 30 minutes, his best published result on black hole Love numbers and tidal symmetries — a paper whose training cutoff predated its arXiv release) → now resident scientist at OpenAI. He notes that no competing model at the time could match GPT Pro on that calculation. > *"In under 30 minutes, with one hint... it completely solved this problem, which is one of the nicest calculations that I've ever done."* ## [72:46] Solving a black hole perturbation problem with GPT-5 Lupsasca details the "Move 37" moment that converted him: his paper "Why Is There No Love in Black Holes?" establishes new symmetry generators for perturbations of a Kerr black hole (explaining why black hole Love numbers — tidal response coefficients, named after mathematician Augustus Love — are exactly zero). When GPT-5 Pro was first given the full problem cold, it failed. But after being primed with the simpler flat-space warm-up (a 200-year-old known result), it then solved the full Kerr black hole problem in 18 minutes. > *"GPT-5 was able to reproduce one of my hardest calculations, which I think the number of people in the world that could do that you could count on your hands."* ## [76:34] Discussing whether AI can make original, conceptual leaps The hosts ask whether AI is doing genuine recombination versus true creative leaps. Lupsasca cites Terry Tao, who has not yet seen an AI proof that cannot be traced to an obscure reference. But Lupsasca has been impressed and frames the distinction as one of degree rather than kind — humans may also be recombination machines. He believes continued scaling will produce feats of insight that look like creativity, and notes OpenAI is actively working on enabling models to take bigger, more out-of-distribution leaps suited to scientific discovery. > *"I'm not sure there's a qualitative difference. I think it's just a matter of degree — as we continue scaling the capabilities, I don't see why it's going to stop."* ## [80:09] Challenges of 'AI slop' and the future of academic publishing With models now capable of turning out a physics paper in 30 minutes when properly steered, the arXiv preprint server is being flooded with submissions. Lupsasca distinguishes legitimate use (expert steering + careful verification) from "AI slop" — poorly prompted outputs submitted without adequate checking. His proposed response: raise the bar rather than increase volume. The single-minus amplitude papers open a clear line of attack toward genuine quantum gravity questions; the goal should be to pursue harder problems, not to publish incrementally. > *"Instead, I think now that we have this new tool that gives us AI superpowers, I think we should just raise the bar for what it means to write a good paper."* ## [83:13] The bottleneck of writing academic papers Asked what single bottleneck he would remove, Lupsasca nominates the paper-writing process itself — finding it increasingly strange that researchers use AI to do calculations, compress results into a static paper, and then readers feed that paper back into AI to understand it. He envisions interactive, LLM-embedded papers as a plausible future. He also identifies two missing capabilities in current models: (1) the spark of creativity to identify the next important question, and (2) reliable self-verification, so that the onus of checking long AI-generated proofs does not fall entirely on humans. > *"Maybe some kind of interactive paper which lives in some LLM. Maybe your whole paper is some ChatGPT page... I think we're going to head in that direction."* ## [90:19] Final takeaways and looking ahead to the next year Lupsasca's closing message: pay attention. The trajectory from "useful for email" to "solves open problems in quantum gravity" has taken roughly 18 months. Models are solving open problems that expert communities spent years on. Extrapolating forward, with more scaling already in the pipeline, the next 6 to 12 months should bring further surprises. The right posture is excitement, careful verification, and a commitment to pursuing harder problems. > *"If you just extrapolate that into the future, imagine where we're going to be in 6 months or a year — I think it's kind of surreal to live through this time, but it's really happening."* ## Entities - **Alex Lupsasca** (Person): Theoretical physicist, Vanderbilt University professor and OpenAI resident scientist; 2024 New Horizons Breakthrough Prize and IUPAP Young Scientist Award winner; expert in black hole physics and scattering amplitudes. - **Andrew Strominger** (Person): Harvard professor and Lupsasca's former PhD advisor; pioneer of celestial holography; co-author of both single-minus amplitude papers. - **Alfredo Guevara** (Person): Postdoctoral researcher at the Institute for Advanced Study (IAS); performed the foundational hand calculations underpinning the AI-assisted breakthrough. - **David Skinner** (Person): Professor at Cambridge University; co-author of the single-minus gluon amplitude paper. - **Terry Tao** (Person): Fields Medal-winning mathematician at UCLA; referenced regarding the question of whether AI proofs involve genuine creativity. - **Scattering Amplitudes** (Concept): Complex-valued functions in quantum field theory encoding probabilities for particles to scatter; the central mathematical objects of both papers discussed. - **Single-Minus Gluon/Graviton Amplitudes** (Concept): Tree-level scattering amplitudes where all but one particle have positive helicity; previously assumed zero in textbooks but shown non-zero in a collinear phase-space region. - **Parke-Taylor Formula** (Concept): Compact closed-form result for maximally helicity violating (MHV, double-minus) gluon amplitudes derived in the 1980s; the template whose analog was sought for single-minus amplitudes. - **Feynman Diagrams** (Concept): Diagrammatic technique to organize perturbative QFT calculations; individual diagrams represent distinct intermediate-particle histories whose amplitudes are summed. - **Love Numbers** (Concept): Coefficients encoding tidal deformability; famously vanish for black holes, a fact connected to hidden symmetries studied in Lupsasca's "Why Is There No Love in Black Holes?" paper. - **Celestial Holography** (Concept): Research program exploring symmetries of quantum gravity via scattering amplitude structure; motivates studying graviton amplitudes. - **OpenAI** (Organization): AI research company where Lupsasca serves as resident scientist; developer of GPT-5 and the internal extended-reasoning model used for the amplitude proof. - **arXiv** (Organization): Open-access physics and mathematics preprint server; mentioned in the context of AI-generated "slop" flooding submissions. - **GPT-5 / ChatGPT Pro** (Software): OpenAI's frontier language model used as the primary AI tool in both amplitude papers; capable of extended reasoning steps of 20-34 minutes per prompt.