PodcastsHear the voice. See the shape of the thought.

Joon Sung Park, founder and CEO of Simile and creator of Stanford's Smallville generative-agents study, walks Sonya Huang through the arc from a 25-agent game town that spontaneously threw a Valentine's party to a company that simulated 1,000 Americans and predicted their answers 85% as accurately as the people reproduced their own. His core argument: today's frontier labs are building the "CPU of intelligence" — rational machines superhuman at problems with right answers — while simulating real human society needs the opposite, a model that encodes people's irrational values, preferences, and taste. CVS uses it for concept testing; some customers simulate their own earnings calls; and Joon's longer bet is a "CERN of human society" that could one day model bank runs, climate cooperation, or the early signals of a collapsing democracy. ## [00:00] Inside Smallville: 25 agents throw a Valentine's party The conversation opens on Joon's conviction — that science fiction's advanced societies always rest on two pillars, "some version of AGI and some version of simulations that really help guide the society" — before Sonya takes him back to Smallville, the April 2023 Stanford project that made his name. The setup was 25 generative agents, each given a persona and equipped with memory, planning, and reflection, then left to live in a small game town: wake up, do routines, go to work, form relationships. What surprised the team was emergent coordination. Isabella, a café owner, decided to throw a Valentine's Day party, spent the day before gathering materials and inviting customers, and on the day itself the party actually formed. > *some of the agents did not explicitly get invited, but we had one agent who got the invite, Claus, who decided to ask his crush out on a date* ## [03:34] From a foundation-models paper to simulating a subreddit Joon traces the origin back to 2020, the year GPT-3 was about to land. As a Stanford researcher he co-wrote the "Opportunities and Risks of Foundation Models" paper, and the part that gripped him was not that the models could classify or generate — interaction researchers had done that for years — but that they could encode human behavior. Coming out of the social-computing tradition, he saw a long-standing hole: there was no way to test how millions of people would behave on a platform short of shipping it and watching what happens, sometimes at real cost. That led to the 2022 Social Simulacra paper, the precursor to generative agents, which populated a simulated subreddit with thousands of personas to let a designer see community dynamics before launch. > *The only way we test it today is you basically field test it. You release your prototype, see what happens.* ## [07:57] The CPU of intelligence can't model irrational humans Asked when models got good enough for a faithful representation of society, Joon marks the path from GPT-3 — janky, no instruction tuning, needing prompt tricks just to follow orders — to today's foundation level where these applications become imaginable. But he draws a sharp limit. The frontier labs' north star is a rational, superhuman machine optimized for objective problems, and that is the wrong target for simulating people. As accuracy on objective benchmarks climbs, the ability to predict and simulate human behavior diverges, because people are not rational. > *We have a lot of subjective values, preferences, and taste.* ## [10:04] Why this became a company, not another paper Joon distinguishes the two vehicles bluntly: research is built for breadth, where each researcher owns a slice of thesis and is "not necessarily known for finishing our job," while a company is built for depth on a single conviction. The pull toward a company came roughly half a year after the generative-agents paper, first from social scientists wanting to run RCTs on the platform, then from Fortune 500 boards and CEOs who saw the demo at Stanford and asked whether the surveys and market questions they could never answer might run in simulation. Before committing, the team validated accuracy: simulations of 1,000 people across the US population. > *we can actually predict people's behaviors 85% as accurately as people replicate their own* ## [12:43] How a Simile engagement works — and the say-do gap Simile's first major customer is CVS, brought in by a senior VP of human insights who had read the validation paper and felt bottlenecked by how few questions he could field-test. The workflow mirrors how firms already use polling and panel companies: a customer names a population they want to understand, and Simile — through a strategic partnership with Gallup — reaches real humans, asks the magical 15-minute questions, and turns that data into agents that answer far beyond the original survey. Sonya pushes on why an LLM alone can't just role-play a 34-year-old woman from a coastal metro. Joon's answer is the say-do gap: models are trained on what people said online, not what they actually do, and closing that gap requires behavioral data — RCTs, pricing studies, and life-story interviews that surface the long-tail of a person. > *There are things that people say and then there are people there are things that people actually do and the gap there is real* ## [20:27] The GPU of intelligence: from concept tests to earnings calls Here Joon gives the framing that anchors the company. Today's models are the CPU of intelligence — one model trained on rational data, superb at objective questions. Simile is building something closer to the GPU: not superhuman, but as human as possible, where individual subunits represent the real viewpoints of different populations. Customers usually enter through a concrete door — concept testing, where instead of testing 5 to 10 ideas they imagine testing a thousand ideas across a thousand sub-populations — then move toward product testing with a temporal dimension and multi-agent simulation. One recurring and initially surprising ask: simulate the company's own earnings call to see how the audience reacts. > *imagine the current today's model are akin to the CPU of intelligence unit* ## [26:32] How accurate is it? Convergence versus divergence On evaluation, Joon starts from the theoretical limit — humans answer the same question slightly differently each time, so perfect prediction is impossible — then describes the metric: total variation distance between the ground-truth and simulated response distributions, with a TVD under 0.15 treated as strong enough for decisions. The deeper idea is two categories of simulation. Convergent ones tolerate compounding error because the pull toward an outcome is strong — like a network always forming a hub, the scale-free structure that powered PageRank. Divergent ones — was World War I inevitable, who wins an election — can't be expected to repeat, so the evaluation shifts to confidence: run it 100 times, see how often outcome X appears, and show the diversity of possible futures. He likens the work to the early days of inferential statistics setting the p < 0.05 threshold. > *was World War I inevitable or was it not?* ## [31:56] A CERN for human society Sonya raises the grander possibility — that fields like macroeconomics, which she sees as human behavior at scale, might one day be partly solved by simulation, including the venture question of where value accrues across the AI stack. Joon agrees there is "a Nobel Prize to be won there," recalling how Thomas Schelling's deliberately crude agent-based segregation models revealed something deep about macro behavior. The augmented version replaces red-dot/blue-dot agents with agents that replicate the full richness of individuals, opening questions economists actually asked him: when does a bank run happen, can nations be modeled solving climate's collective-action problem, what are the early signals of a democracy about to collapse. He imagines a simulation that costs $100 million and months to run once but answers a fundamental question — a Hubble telescope for human society. > *building simulator that's akin to the CERN of human society* ## Entities - **Joon Sung Park** (Person): Founder and CEO of Simile; created Stanford's Smallville generative-agents study and co-authored Social Simulacra. - **Sonya Huang** (Person): Partner at Sequoia Capital, AI investing; host of the conversation. - **Simile** (Organization): Applied AI lab building models that simulate human behavior and societies for concept testing, product testing, and multi-agent scenarios. - **Smallville** (Concept): 2023 Stanford experiment with 25 generative agents living in a game town, known for emergent behavior like a self-organized Valentine's party. - **Social Simulacra** (Concept): 2022 paper simulating a subreddit with thousands of personas; precursor to generative agents. - **Say-do gap** (Concept): The difference between what people say (the basis of LLM training data) and what they actually do, which behavioral data is collected to close. - **CPU vs GPU of intelligence** (Concept): Joon's framing — frontier labs build a rational "CPU" superhuman at objective problems; Simile builds a "GPU" encoding the diversity of human values and taste. - **Total variation distance** (Concept): Simile's accuracy metric comparing ground-truth and simulated response distributions; TVD < 0.15 treated as decision-grade. - **CVS** (Organization): Simile's first major customer, using it for concept testing via its human-insights team. - **Gallup** (Organization): Polling and panel partner Simile uses to reach real humans and ground simulations in real data.

David Senra has spent a decade reading 400+ founder biographies and recently started interviewing the living ones face to face. His single-word answer to what they all share is focus — what he calls "mute the world and build your own" — and he walks Brian Halligan through why that trait, combined with a near-compulsive drive rooted in early experiences, explains more about founder success than any Silicon Valley pattern-matching checklist. The conversation covers childhood origins, founder archetypes, the danger of selling your best company, and how the AI era is making extreme craft more valuable than ever — while the fundamental human wiring of great founders stays the same. ## [00:00] Introduction Brian Halligan opens by framing what he wants from David: a distillation of what the very best founders — from Jesus of Nazareth to Jensen Huang — actually share, and how to use that knowledge to pick and coach them. The episode starts mid-thought with David on Tony Xu of DoorDash, who, by the end of dinner celebrating a milestone, was already cataloguing the seventeen things still going wrong. That restlessness, David argues, is the tell. > *"By the time the dinner before the dinner is over, I'm thinking of the 17 things that are not going right. That's why it's great."* ## [01:11] Focus Above All David's one-word answer is focus. Not hustle, not resilience, not intelligence — focus. He describes it as something qualitatively different from what other high performers do, almost a separate species: they are not looking around at what competitors are doing, they genuinely do not care. His shorthand is "mute the world and build your own." > *"If I had to distill every single thing down to one word, it just be like focus. They're just unbelievably focused compared to not only the average person. It's almost like they're a different species."* ## [01:50] Dana White UFC Focus Dana White is David's freshest example of missionary focus. White grew up a self-described loser working as a bellman in Boston, moved to Vegas to be near the fight industry with nothing to lose, and eventually talked the Fertitta brothers into buying the UFC for $2 million. For six years they lost money. Then they lost another $40 million before turning profitable. Twenty-six years later White closed a TV deal worth nearly $8 billion — and his explanation for how it happened is that he never once read a business book or listened to a business podcast. He just made what he wanted to see. > *"His entire world is his business and then anything doing outside he doesn't care about. He's just unbelievably focused."* ## [04:19] Focus vs Obsession Brian asks whether focus and obsession are the same thing. David says they're closely related but different: focus is the act of saying no to good ideas so you can pursue a great one. He cites Jony Ive recounting Steve Jobs's distinction — focus is saying no to a good idea you really want to do because it distracts you from a great idea — and notes that anyone intensely focused on something will look obsessed from the outside, but the mechanism is active exclusion rather than passive fixation. > *"Focus is saying no to a good idea that you really want to do in because it distracts you from a great idea."* ## [05:05] Origins in Childhood Brian asks where the obsession comes from: normal upbringings, or something broken early? David says it's not one thing, but nearly all of the founders he's studied are not what you'd call well-adjusted. He brings in the Francis Ford Coppola biography as the source of the line that crystallized a pattern he'd been seeing repeatedly — that the son's drive is always embedded in the story of the father — and describes how he thinks of filmmakers, podcast hosts, and startup founders as the same entrepreneurial type. > *"The answer is it's not one thing."* ## [06:07] Coppola and His Father The pattern David keeps finding is that the father's story is embedded in the son. Coppola's father was a brilliant but failed musician who told his young son "there can only be one genius in the family — it's me," then spent years putting him down. Coppola internalized that and built one of the most relentless work ethics in Hollywood, eventually winning the Academy Award and letting his father write the score, which also won an Oscar. David applies this through Charlie Munger's framework: to truly understand an idea you have to tie it to the personality that developed it, which is why biography outperforms strategy books. > *"You can always understand the son by the story of his father. The story of the father is embedded in the son."* ## [08:48] Assholes and Archetypes Brian raises the cliché that great founders are assholes. David rejects it flatly. He's working with Daniel Ek of Spotify on a project to map founder archetypes — the hypothesis being that founder-problem fit matters more than product-market fit. Ek spent years trying to imitate Steve Jobs and wasted that time wearing a personality that wasn't his. He's more of a coach archetype. David's point: there is no single archetype, there are probably six to eight, and understanding which one you are is more valuable than imitating whichever founder happens to be famous right now. > *"The most important is founder problem fit. Like think about Demis from DeepMind. There's one great company he had in him. It was DeepMind. He was put on this planet to do what he is doing."* ## [11:14] Autism and Originality Brian raises the high prevalence of autism spectrum traits among the modern trillion-dollar CEOs — Jobs, Gates, Bezos, Zuckerberg, Jensen, Ellison. David reads Peter Thiel's take: the founders who seem mildly Asperger's are missing the imitation-socialization gene, which means no one talks them out of their strange original ideas before those ideas are fully formed. David's caveat: the Bay Area is now full of people performing anti-imitativeness, which makes them the most mimetic of all. Rockefeller probably didn't fit the spectrum pattern — but he had advanced social skills and still built the most dominant company in history. > *"We need to ask what it is about our society where those of us who do not suffer from Asperger's are at some massive disadvantage because we will be talked out of our interesting, original, creative ideas before they're even fully formed."* ## [14:55] Immigrant Drive and Grit David speaks from personal experience as the son of a Cuban immigrant: people who risked their lives on rafts to cross 90 miles of ocean give their children a different baseline for what risk and opportunity mean. Brian notes that only three of the ten largest American tech founders were immigrants — Jensen, Elon, Sergey — while most were suburban upper-middle-class. David's rejoinder: those three account for a disproportionate fraction of total market cap, and many of the others had immigrant fathers. The advantage may transmit across a generation. > *"Think about how much you love your son and how bad Cuba had to be and communism had to be to put your 14-year-old or nine-year-old son on a raft and hope to make that 90-mile journey to South Florida."* ## [16:38] Bet on the Founder David says if he were a VC he wouldn't run any rubric — he'd just bet on the person. Ed Catmull told him the clearest version of this: give a great idea to a mediocre team and they'll ruin it; give a mediocre idea to a great team and they'll either fix it or throw it out and build something better. Ideas come from people, so people matter more than ideas. David's test: does this person have the quality that Travis Kalanick had at Uber, which is that they will make it work or die trying? > *"If you give a great idea to a mediocre team, they'll mess it up. If you give a mediocre idea to a great team, they either fix it or throw it out and create something new."* ## [17:52] Solo vs Partners The conventional wisdom — co-founders are better, optimal number is three — doesn't match what David sees across history. Most great companies had one dominant driving force, and the "co-founder" either left (Wozniak), was essentially an operator the founder acquired (Frick at Carnegie Steel), or was a complementary personality who consciously subjugated himself to a once-in-a-century talent (Munger to Buffett). When David met Munger, Munger admitted he always thought he was smarter than everyone else, but recognized Buffett's singular focus and made a deliberate calculation to subordinate his own ego to it. > *"If I could do life again, I'd still think I was smarter than everybody else, but I would do a better job of hiding it."* ## [23:20] Negative Self Talk Fuel Jensen Huang says he looks in the mirror every morning and asks himself why he sucks so much. Elon describes his mind as a storm and seems genuinely unsettled when things are going well. Most of the founders David has studied run on negative self-talk as a fuel source — but David recently changed this about himself. Brad Jacobs, who built eight separate billion-dollar companies over 45 years, told him: the negative drive got you here, but it's not serving you anymore. Now you love the work. Make your inner drive generative. David says something clicked and he hasn't gone back. > *"Your inner drive should be generative. It should be like, 'Hey, I'm trying to make something that's good for the world that I love to do that I'm very proud of.'"* ## [26:39] Platform Shifts and Founder Mode Brian asks whether major platform shifts — the industrial revolution, assembly line, now AI — change the profile of who succeeds and how they run companies. Brian describes the Paul Graham founder-mode vs. manager-mode distinction and his own "Dorsey mode" framing: flat org chart, titles eliminated, an AI system at the center making an increasing percentage of decisions while humans feed it context and apply judgment. He sees this as structurally different from any previous platform shift. > *"Over time, the AI system makes very few of the decisions today, but maybe 5%, 10% — the percentage of decisions the AI system makes versus the humans starts to flip."* ## [28:07] Dell Versus IBM David asked Michael Dell directly whether this moment feels like anything he's been through before. Dell said no — this is categorically different. David is ordinarily skeptical of "this time is different" claims, but agrees with Dell, Toby Lütke, and Jack Dorsey that the amount of leverage now available to a small team changes the math of company-building fundamentally. IBM once had 80% market share of the entire technology industry and was the first company ever to hit a $100 billion market cap. Dell took them on from a University of Texas dorm room with $1,000 — and was profitable every single quarter for his first twenty years. > *"I actually think the way to run a company — I do think the way to do it and how you could do it and what's available to you is completely different."* ## [30:02] Infinite Leverage Edge Naval Ravikant's line — "in the age of infinite leverage, being at the extreme of your craft is very important" — was written before AI. David thinks AI just amplifies that truth by another order of magnitude. His example is Jordi from TBN: he wasn't 2x better at podcast marketing than the next person, he was 100x better, and the economic rewards available to someone at that frontier are not 100x bigger, they're potentially 1,000x bigger. The premium on focus and mastery is going up, not down. > *"In the age of infinite leverage, being at the extreme of your craft is very important."* ## [31:38] Focus Versus Speed Brian pushes back: the AI-native founders he knows — Harvey, Lovable, ElevenLabs — are moving fast on many fronts simultaneously. Is focus still the rule? David's answer: they haven't built durable businesses yet, so it's too early to know. His deeper concern is what happens after you sell. He's spent time with founders in their 70s and 80s who sold their best company and spent decades trying to recapture the magic on second and third bets — almost none succeeded. If you truly have a generational company, don't sell it. You're either all in or all out. > *"You're all in or all out — but why would you be all in on your second, third, fourth, fifth best idea?"* ## [34:20] Taste And Listening Brian asks whether great taste is a genuine founder trait or a fashionable concept. David says taste is very real, and his clearest example is Rick Rubin — still doing at 62 what he started at 18 in his dorm room. But David's more specific claim is that Rubin's edge isn't just taste, it's that he's a professional listener. Most people in conversation are waiting to respond. Rubin is actually interested. That quality of attention, transferred from music production to podcasting, is what makes him exceptional. David also addresses founder authenticity: not everyone should be unfiltered — it depends on who you are, what industry you're in, and what you're trying to build. > *"He took a skill from music and applied it to podcasts. You're a professional listener."* ## [40:52] Founder Traits And Balance The core shared traits David has identified across 400+ biographies: obsession, high disagreeableness, cost control obsession, and micromanagement — what Paul Graham called "founder mode," which David notes is not new at all. Rockefeller was actually an exception on disagreeableness, never raised his voice, but was a force of nature in other ways. On the work-life balance question: David can name exactly three founders across four centuries who had genuinely well-rounded personal lives. Sam Walton, writing his autobiography while dying of cancer, said he'd do it all exactly the same way. Phil Knight at 75 still can't fully reconcile his absence from his sons' lives. What motivates the great ones isn't money — it's control. > *"I don't think small egos build big companies — I think all of these people have giant egos. I think some of them are just better at hiding it. And what motivates most founders is not money, it's control."* ## [54:22] Closing Takeaways Brian distills three takeaways: deep founder-market obsession is the real common thread; having good work-life balance while building a great company is genuinely rare (three out of 400); and impostor syndrome is worth working on — Brian references Brian Chesky's shift from leading from fear to leading from love as the model. The episode closes with Dana White's formula: understand deeply who you are, understand deeply what you want to do in the world, then wake up every day and execute. Stay in the game long enough to get lucky. > *"Stay in the game long enough to get lucky."* ## Entities - **David Senra** (Person): Host of the Founders podcast; has read 400+ founder biographies and now interviews living founders face to face - **Brian Halligan** (Person): Co-founder and executive chairman of HubSpot; hosts this Sequoia Capital series - **Dana White** (Person): Founder/CEO of UFC; bought it for $2M in 2001, recently closed a ~$8B TV rights deal - **Daniel Ek** (Person): Founder of Spotify; working with David on a founder archetypes framework; advocates founder-problem fit over product-market fit - **Demis Hassabis** (Person): Co-founder of DeepMind; cited as the clearest example of perfect founder-problem fit - **Charlie Munger** (Person): Partner at Berkshire Hathaway; consciously subjugated his ego to Buffett's once-in-a-century talent - **Ed Catmull** (Person): Co-founder of Pixar; Steve Jobs's longest consecutive collaborator; source of the "give a great idea to a mediocre team" principle - **Brad Jacobs** (Person): Entrepreneur who built eight separate billion-dollar companies; advised David on switching from punishing to generative drive - **Rick Rubin** (Person): Music producer; David's example of taste combined with professional listening as a compounding edge - **Founders** (Media): David Senra's podcast covering 400+ biographies of founders from history to present day - **founder-problem fit** (Concept): Daniel Ek's framework — the match between a founder's identity and the specific problem they're solving is the most important form of fit - **infinite leverage** (Concept): Naval Ravikant's idea that in an age of software and AI, being at the extreme of your craft produces disproportionately large rewards - **Sequoia Capital** (Organization): Venture capital firm; Brian Halligan's current base and the host of this podcast series

Alfred Wahlforss built Listen Labs after scratching his own itch: when his viral AI-avatar app hit 20,000 users overnight and churn spiked, he needed to know why—fast. The answer was an AI agent that runs voice interviews at scale, drawing from a panel of 30 million people. A year in, Listen serves 20% of the Fortune 500 and has completed over a million interviews. The deeper finding is counterintuitive: respondents are often more honest with an AI interviewer than a human one, and voice transcripts turn out to be richer training signal than credit card data or behavioral logs. Wahlforss and Sequoia's Konstantine Buhler work through why audience selection consumes 80% of Listen's engineering, how back-tested simulation beats vanilla ChatGPT at message testing, and why—as AGI makes building trivially cheap—knowing *what* to build becomes the scarce resource Listen wants to own. ## [00:00] Introduction Alfred opens in the middle of a thought about audience depth: Listen's long-term goal is to reach a billion people and build rich profiles that reveal each person's genuine areas of expertise—not just demographic boxes, but things like whether someone is a true sneaker influencer versus a casual buyer. Konstantine then formally introduces him: Listen launched roughly a year ago, already counts Microsoft, Anthropic, Sweet Green, NBC, and others as customers, and runs thousands of voice interviews simultaneously. The brief cold-open framing gives the episode its throughline—the value of talking to the *right* person, not just any person. > *"Our goal is to get to a billion people in our audience and then to be able to stratify and know what exactly is this person an expert on."* ## [01:20] How Listen Works The product works in three stages: a researcher types a question (say, "how can we improve Cursor's onboarding?"), Listen's AI agent generates an interview guide, then routes those interviews to matched participants from its 30-million-person panel. Hundreds of conversations run in parallel, the results are synthesized, and recommendations surface. The next stage, launching in a few months, is simulation: after tens of thousands of interviews accumulate on a topic, can Listen predict how customers will answer *future* questions without running a new interview? > *"As we get closer to AGI, it will be easier to build things, but the hard part will be knowing what to build—and that's what we're building at Listen."* ## [02:23] Customer Wins Chubbies discovered that chest hair caught uncomfortably on one of their shirt materials; Listen surfaced the feedback, Chubbies redesigned the garment, and comfort scores jumped. Manscaped used Listen insights to reshape a Super Bowl ad. Skims uses it for ongoing product testing. The through-line Alfred draws: Listen handles both small product details and high-stakes campaign decisions with the same workflow—talk to real people, fast. > *"They discovered that chest hair interface really poorly with one of the materials they have. So it's really uncomfortable to wear one of their shirts, and they changed the shirt and it became radically more comfortable."* ## [03:28] Surveys Versus Reality Konstantine presses on the classic critique: survey respondents lie, or at least contradict themselves. Alfred's evidence: Listen ran the same multiple-choice survey questions back to the same people and found radical inconsistency—but when those same people had to reason through an open-ended voice answer, consistency improved sharply. On sales-data back-testing, Alfred agrees AB tests are the gold standard but notes they require large user bases that most companies don't have. Interview data, properly designed, beats no data. > *"If you go back to the same person and ask them a survey question in a multiple choice fashion, they're much more inconsistent. But when you actually have to think and reason through your answer, you're much more consistent."* ## [05:13] Zoom Like AI Interviews The participant experience is a video call with an AI agent—not a text form. The agent watches facial expressions and vocal tone, giving Listen a second signal layer beyond what people say. Alfred cites advertising testing as the clearest win: respondents might rate an ad highly on a Likert scale but show genuine enthusiasm in video, and that enthusiasm predicts Meta and LinkedIn performance marketing results significantly better than the numeric score. Every data point links back to the actual video clip, so researchers can verify the AI isn't hallucinating sources. > *"For every data point you can always click and then look at the video or see the quote—so you know that AI is not just hallucinating where it's coming from."* ## [07:14] Origin Story Alfred and his co-founder shipped a consumer app called "Be Fake"—an early stable-diffusion fine-tuning tool for creating AI avatars of yourself—which went viral overnight and hit 20,000 users. Churn spiked immediately and they had no idea why. They built an AI interview tool to ask their own users, found it genuinely useful, and pivoted. The market-research product they built for themselves became Listen Labs. > *"We built this AI interview for ourselves because we had a ton of churn and we wanted to understand why—and that's how we got started."* ## [08:01] Old World Research The pre-Listen world had two speeds: slow online survey tools like Qualtrics, or expensive services firms that charge tens of millions to recruit participants, design question methodology, moderate focus groups, and synthesize hundreds of transcripts. Question design alone is an academic discipline—ask "how much would you pay for this?" and you get junk data. The sourcing problem is equally hard: incidence rates of 10% mean nine out of ten recruited panelists get screened out, burning trust and causing churn on the databases themselves. > *"In traditional industries like CPG or even Microsoft, they spend tens of millions of dollars on focus groups to bring people in a room and interview them—and we can help speed that up much faster."* ## [09:50] AI First Benefits Three compounding advantages: speed (results from real people in five minutes), cost (asynchronous interviews pay participants less than synchronous ones, and participants accept that willingly), and honesty (people open up more to a non-judgmental AI than to a human interviewer who might silently judge them). Alfred mentions sensitive use cases—interviewing children about products, with parental consent—as an area where the AI's non-threatening presence produces data that focus groups can't. > *"People are more honest talking to an AI. It's a very therapeutic experience because it's a non-judgmental entity that's really interested in you."* ## [11:32] Finding The Right People Listen spends 80% of its engineering resources on audience quality, not the interview agent itself. The reason: power-law customer segmentation means talking to the wrong 100 people gives you wrong insights. Sweet Green's most valuable customer is urban, high-income, mostly female, and—Alfred's specific example—knows what seed oils are (roughly 1% of the population). Listen builds rich profiles across every interview a panelist ever participates in, so an offhand comment ("I'm a total sneaker head") in an unrelated interview can resurface that person when Nike needs launch feedback. Traditional email-list panels couldn't do cross-topic profiling. > *"Even a product like Sweet Green, which you would think is for everyone, the right audience is typically urban, high household income, mostly female—and they need to know what seed oils are, which only like 1% of the population does."* ## [14:30] CRM And Prospecting Sweet Green already has a CRM full of its most loyal customers—so why use Listen? Three reasons: researching *prospective* customers who aren't yet in the CRM requires an external panel; CRMs are typically disorganized and legally constrained (Google can't spam Gmail users, even its own); and direct outbound email risks getting flagged as spam, which can permanently damage a domain's deliverability. Listen provides clean, third-party panel access that sidesteps all three problems while still supporting CRM-connected campaigns when brands want them. > *"What we found is that the CRM is typically really unorganized, and sometimes there are regulatory issues—if you're at Google, you can't just send emails to people who use Gmail."* ## [15:35] Consulting In The AI Era Konstantine—a former buyer of McKinsey-style consulting—asks whether firms like Bain still have a role. Alfred's view: yes, but margins compress. Bain already uses Listen to accelerate existing workflows. The more optimistic scenario: AI doesn't just replace a research project, it makes research cheap enough to run five simultaneous strategic explorations that a company never would have commissioned before. Alfred predicts consulting expands in scope even as price-per-project falls. On economic surplus, Listen has charged hundreds of thousands of dollars to interview 20 doctors across eight countries—fast—a project that previously would have taken months. The surplus is currently staying with the supplier. Alfred also flags an emerging agentic loop: churn interviews surface bugs, which connect directly to a coding agent that opens a PR and ships the fix. Listen as the customer-intelligence "left side" of an autonomous product development cycle. > *"Because you're able to do it faster, I would argue you should be able to charge more for it—and we have charged hundreds of thousands of dollars to speak to 20 doctors across eight countries."* ## [20:05] Market Research Simulation This is the episode's technical core. Konstantine frames the evolution as 1.0 (call 100 people manually), 2.0 (AI-native simultaneous interviews), and 3.0 (generative simulation). Alfred explains how Listen's simulation works: interview a single person deeply, build a persona model, then scale to a thousand statistically representative agents. Back-testing removes a held-out question and measures prediction accuracy—they reach 95% on stable preference domains and deliberately expose the model to nonsensical queries (dog names) to calibrate what it *can't* predict. Alfred ran a personal live test: 100 title variants for a conference talk, run through Listen's panel simulation. The top-ranked title performed twice as well as the second. He then ran the same test in ChatGPT—which picked the wrong title when shown a past successful talk versus a less successful one. Listen's domain-specific panel data beat the general model. The gap: interview transcripts outperform credit card spend, behavioral logs, or ChatGPT persona prompting because voice conversations capture how a specific *type* of person actually reasons, not just what the average person does. Looking ahead, Alfred sees simulation handling "billboard tagline" decisions while real interviews remain the standard for Super Bowl ad buys. The product's proprietary eval climbed from 20% to 85% on avoiding repetitive questions, then Listen raised the bar with a harder eval (screen-state awareness, skipping irrelevant questions) and is back at 20%—which Alfred frames as the vertical AI flywheel: a proprietary benchmark that only you can keep climbing. > *"We were able to get 95% accuracy to predict how they will answer certain questions. The tricky part is knowing what things you can answer and what you can't."* ## [35:33] Closing Thoughts Alfred's conviction: human input will always be necessary because humans are inherently irrational—TikTok trends can overturn a marketing strategy overnight, and no AGI will preempt that. His uncertainty: the ceiling for simulation quality. His moat argument: network effects on the panel (supply-demand flywheel), data network effects (more interviews → better simulation), and product stickiness (interview history compounds inside the platform). But the simplest advantage he cites is opinionated defaults—early customers using vanilla LLMs to design their own interview guides got bad data and blamed Listen; now the agent enforces question-design best practices and data quality is consistent. Konstantine ends with the "Tide Pods moment" question: can Listen's AI start *generating* product ideas mid-interview rather than just testing them? Alfred says customers already feed AI-generated images into interviews manually; the MCP integration means Claude can loop Listen calls autonomously. The vision is live brainstorming between the AI interviewer and the respondent—ideas surfacing as the customer articulates a pain, not after. > *"Founders want to build something that's complex X, but customers want something that's stupid simple and it just works. And that's the advantage you have as a vertical AI company—you can train the agent to follow best practices in the work that you do."* ## Entities - **Alfred Wahlforss** (Person): Co-founder and CEO of Listen Labs; previously built "Be Fake," a viral AI-avatar consumer app. - **Konstantine Buhler** (Person): Partner at Sequoia Capital; host of the Training Data podcast; former consultant and operator. - **Listen Labs** (Organization): AI-first customer research platform; runs voice interviews with a 30-million-person panel; building generative simulation. - **Market Research Simulation** (Concept): Building persona models from accumulated interview data to predict future customer responses without running new interviews; back-tested against held-out questions. - **Audience Quality** (Concept): Listen's thesis that 80% of research value comes from recruiting the right respondents—power-law customer segments—not just any panelists. - **Be Fake** (Software): Alfred's earlier consumer app (AI avatar fine-tuning via stable diffusion); the origin of Listen's interview tooling. - **Bain** (Organization): Management consulting firm; cited as an active Listen customer using the platform to accelerate traditional research workflows. - **Procter & Gamble** (Organization): Cited as the historical archetype of market-research-driven brand management; Tide Pods and M&M's given as canonical examples. - **Qualtrics** (Software): Legacy survey platform representing the "old world" of market research tooling.

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

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

Brian Halligan interviews Notion co-founder Ivan Zhao on his journey as a 'refounder' who navigated the company through its 2015 Kyoto restart and the 2023 generative AI pivot. Zhao details Notion's transition from a traditional SaaS structure to an AI-native 'jazz band' model that prioritizes technical versatility, taste, and agency over rigid hierarchies. The discussion explores how AI acts as the 'steel' for modern organizations, enabling flatter structures and faster, more reversible decision-making. ## [00:00] Introduction Brian Halligan introduces Ivan Zhao as the 'refounder' of Notion, highlighting his unique ability to restart the company during critical junctures in 2015 and 2023. The conversation sets the stage for Zhao's transition from a traditional SaaS management model to an AI-native organization. Halligan compares Zhao's approach to other tech visionaries like Jack Dorsey, emphasizing the importance of personal style and 'taste' in building a lasting brand. > *I like to think of him as the refounder... he's the canonical example of how a SAS company can move and become an AI company. [00:52]* > *We want to be a jazz band, not a marching band. [00:02]* ## [02:22] From Founder Mode to AI Org Ivan Zhao discusses his detour into traditional delegation and professional management before returning to a hands-on 'founder mode' necessitated by the AI shift. He explains that building with language models is less like predictable bridge engineering and more like 'brewing beer,' where the underlying technology dictates the development path. Zhao emphasizes hiring 'jazz band' people—versatile individuals like designers who code—to navigate the experimental nature of AI integration. > *Building with language model... is like brewing beer. You can't truly predict the things the underlying thing. [06:33]* > *The spirit is technology first-driven development rather than customer-driven first development. [07:01]* ## [11:00] Hiring for Taste and Agency Notion utilizes a 'barbell' hiring strategy that targets both super-junior and super-senior talent while avoiding the 'middle' of traditional SaaS experience. Zhao defines talent as the product of capability, taste, and agency, noting that AI has democratized basic capabilities like coding and writing. Consequently, the company now optimizes for 'agency' and 'taste,' qualities that remain difficult to automate and serve as the primary differentiators for the brand. > *capability got normalized democratized and taste becomes still important [11:53]* > *So the shape it's not it's more like the barbell barbell shape, right? [12:35]* ## [24:28] Refounding Notion in Kyoto In 2015, facing potential failure and low morale, Zhao and co-founder Simon Last laid off their entire staff and relocated to Kyoto, Japan, to rebuild Notion from scratch. This 'Kyoto Reset' allowed them to focus entirely on craft and coding while living a minimalist lifestyle. Zhao chose Kyoto specifically for its status as the 'craft capital of Asia,' which provided the spiritual inspiration needed to view software as a fundamental human tool. > *So my co-founder and I said let's just lay off everybody just go by the two of us. That's the Japan story. [25:41]* > *The story we tell ourselves is like Kyoto is a special place. If you can pull off anywhere, you can pull off from Reborn in Kyoto. [28:05]* ## [30:27] Craft Versus Commerce Zhao views Notion as part of a historical lineage of 'tools for thought,' tracing back to pioneers like Douglas Engelbart and Alan Kay. He criticizes modern Silicon Valley 'tinker culture' for ignoring the history and humanity behind technology. For Zhao, the goal is to find an equilibrium between the pure craft of an artist and the commercial viability of a business, ensuring the product has a 'soul' that resonates with users. > *Tech is like industry doesn't know its past. If you don't know his past you don't know history which is humanity. [31:52]* > *I need to be in equilibrium with my own value of what this company I want to build... [51:33]* ## [32:26] When to Refound For founders whose companies are stagnating, Zhao suggests listening to the 'inner urge' to take drastic action rather than wasting years on ventures without momentum. He argues that refounding is often harder than starting fresh because it involves taking a significant step back to pivot toward a new growth engine. Zhao believes the current AI-driven market is wide open, making it an ideal time for founders to be risk-seeking and follow their intuition. > *For me it's like there's you just feel you have to do something drastic... then you feel liberated once you land in Japan. [32:56]* > *The refounding is harder than it looks. It typically involves like a big step back and two steps forward. [59:57]* ## [34:07] GPT-4 Refounding Shock Zhao describes gaining early access to GPT-4 as a 'full body religious experience' that signaled a fundamental shift in the world. This realization forced a second refounding of Notion, as Zhao felt any work not involving this technology would soon become meaningless. The transition included a grueling 18-month period of low morale while the team waited for the underlying AI models to catch up with their ambitious product vision. > *GBD4 is a religious experience for me. It's like holy [ __ ]... anything you do if you don't do this it will be meaningless. [34:27]* > *that was like a year and a half just go with no error and morale is definitely low [35:50]* ## [45:35] Leadership and Founder Energy Despite being naturally introverted, Zhao explains how he forced himself to master one-to-many communication to build trust within Notion. He maintains a disciplined daily routine, starting at 7 AM and often working until midnight, while using 'guilty pleasure' reading to recharge. To prevent organizational calcification, Notion aggressively acquires startups to bring in 'founder energy,' currently employing over 50 former founders who lead critical domains. > *To lead the group of human you need to do one to many communications otherwise people don't trust you. [46:17]* > *founders are are kind of this kind of like little decalcified meatthead machinery just trying to break things [39:10]* ## [53:17] Sales Culture and Closing Thoughts Notion's transition to enterprise sales involved moving away from 'first-principle' experimentation toward established playbooks, pairing system thinkers with high-energy sales leaders. The conversation concludes with a vision of the 'AI-native' CEO playbook, which replaces traditional 'triangle' hierarchies with a 'circular' model. In this structure, a centralized AI system saturated with company context enables smaller teams to move at breakneck speed with reversible decision-making. > *You should only have each company should only preserve your innovation point to few places... [54:54]* > *All of those kind of one-way doors that Bezos used to talk about are really two-way doors... [62:39]* ## Entities - **Ivan Zhao** (person): Co-founder and CEO of Notion, known for his 'refounder' mindset. - **Brian Halligan** (person): Co-founder of HubSpot and interviewer. - **Notion** (organization): A productivity software company that pivoted to an AI-native model. - **Simon Last** (person): Co-founder of Notion who helped rebuild the company in Kyoto. - **Kyoto** (location): The Japanese city where Notion was restarted in 2015. - **GPT-4** (technology): The AI model that triggered Notion's second refounding. - **Steve Jobs** (person): Former Apple CEO cited as an inspiration for refounding and craft. - **Jack Dorsey** (person): Tech leader mentioned for his AI-centric organizational redesign. - **Douglas Engelbart** (person): Computing pioneer in the 'tools for thought' lineage. - **Erica** (person): CRO of Notion and former CRO of GitHub. - **SaaS** (concept): Software as a Service, the industry context for Notion's evolution. - **Jazz Band** (concept): Metaphor for a flexible, high-agency organizational structure.

Mikey Shulman, co-founder of Suno, discusses the platform's evolution from a physics-based startup to a leader in generative AI music. By modeling music as raw sound waves rather than traditional theory, Suno empowers users to transition from passive listeners to active creators in the era of 'creative entertainment.' ## [00:00] Physics, Raw Sound, and Technical Philosophy Mikey Shulman explains how his background in quantum physics at Harvard influenced Suno's interdisciplinary approach to music technology. By modeling audio as raw sound waves sampled 48,000 times per second rather than using traditional music theory, Suno avoids creative constraints and allows for the emergence of new, microtonal genres. > *I think what I mostly learned is playing at the nexus of two things that don't usually play together is just a massive opportunity. [02:00]* ## [02:15] The Pivot to Consumer Music Generation Initially focused on audio analysis, the Suno team pivoted to generation after breakthroughs in audio compression made high-quality output computationally feasible. They validated the product's 'fun factor' through a Discord bot, discovering that the addictive nature of creation was a stronger signal than traditional business use cases. > *When you are staying up late playing with the thing, and you don't want to go to sleep, it's like a really good sign. [04:49]* ## [11:41] Why Music AI is a Research Problem, Not a Scale Problem Unlike Large Language Models, music generation lacks objective benchmarks, making raw compute scale less effective than targeted research. Shulman emphasizes using human preference data and reinforcement learning to align models with creative tastes, favoring a steady release cadence over long-term isolated development. > *In music there are no right answers. There are no benchmarks. Um, and so scale is somewhat less helpful in solving it. [12:28]* ## [16:22] From Passive Consumption to Creative Entertainment Shulman introduces the concept of 'creative entertainment,' where the act of building provides more fulfillment than the final product itself. He notes that 90% of Suno users are active creators, drawing parallels to the 'bedroom producer' era where accessible tools led to the discovery of new genres. > *People are creating music for the fun and enjoyment and fulfillment that comes with being creative. [17:05]* ## [22:52] Industry Partnerships and Professional Integration Addressing industry concerns, Shulman highlights Suno's partnership with Warner Music Group and its role in augmenting professional workflows. He argues that AI will raise the quality ceiling for artists and predicts that interactive live performances, such as audience participation at Coachella, are the next frontier. > *I think people incorrectly assume that we hate the existing music industry and especially we hate the record labels. [23:17]* ## [25:53] Product Strategy and the Application Moat Suno prioritizes the application layer and user experience as its primary competitive moat, viewing itself as a music company rather than just a technology firm. By focusing on storytelling through full-length lyrical songs and social co-creation features, the company aims to revive the cultural impact of music as a social medium. > *It's unclear how much moat exists in only a model... it's just really undervalued to invest in the product and the UI and the UX. [26:50]* ## Entities - **Mikey Shulman** (person): CEO and co-founder of Suno with a PhD in physics from Harvard. - **Suno** (organization): An AI-powered creative entertainment platform for music generation. - **Sonya Huang** (person): Partner at Sequoia Capital and host of the interview. - **Warner Music Group** (organization): A major global record label that partnered with Suno. - **Discord** (organization): The platform where Suno initially launched its music generation bot. - **Harvard** (organization): The university where Mikey Shulman studied quantum computing. - **Iamona** (person): A poet and artist who uses Suno to create music, illustrating the tool's professional potential. - **Coachella** (event): A major music festival cited as a future venue for interactive AI music experiences.

Boris Cherny, the creator of Claude Code, sits down with Sequoia's Lauren Reeder at AI Ascent 2026 and makes a blunt claim: for the code he writes, coding is already solved. He hasn't typed a line by hand in 2026, runs dozens of agent "loops" at once, and ships most of his work from his phone. The throughline is a bet that writing code is becoming so cheap that the interesting questions move up a level — to what teams look like, what happens to software products, and whether the printing press is the right way to think about what's coming. ## [00:00] Introduction A Sequoia emcee opens the AI Ascent session by asking the room for a show of hands: who uses Claude Code, and who has "Claude Code psychosis." She introduces Boris Cherny as the creator of the tool and hands the interview to Sequoia's Lauren Reeder. > *"We know that the entirety of software development kind of rests on your shoulders."* ## [00:55] Claude Code Crowd Check Reeder frames the conversation for a room full of builders and fills in Boris's background: a career writing code, a TypeScript textbook, an engineer's engineer. The detail that lands hardest is recent — as of early 2026 he hasn't written a single line of code by hand, a sharp reversal for someone who built his reputation on craft. > *"Last time we chatted you hadn't written a single line of code in the last year, or at least so far in 2026, which is quite the change."* ## [02:39] Origin Story of Claude Code Boris explains that Claude Code started almost by accident inside Anthropic Labs, a small incubator he joined in late 2024 that also produced MCP and the desktop app. The team built what it wanted, disbanded, and has since reunited under Mike Krieger for a second round. The motivation was a sense of "product overhang" — capability sitting unused because no product had caught up to it yet. > *"The reason that I started to work on coding is we felt like there was this product overhang."* ## [03:35] From Typeahead to Agents In late 2024 the state of the art was typeahead — press tab, complete one line — which Sonnet 3.5 had just made viable. Boris bet the model was nearly ready to skip that step and write all the code as an agent. It didn't work for the first six months; even after release, Claude Code wasn't a hit. The exponential growth only arrived with Opus 4. > *"I built it, and it just really didn't work for the first 6 months. It was barely usable."* ## [05:07] Is Coding Solved Reeder presses on Boris's on-the-record claim that coding is solved. He polls the room — hand-written code versus fully agent-written — and lands the audience around "50% solved," then says for him it's 100%. He points out the Claude Code codebase itself is unglamorous TypeScript and React, chosen deliberately because that stack is heavily represented in the model's training distribution. > *"For me it's just solved."* ## [06:50] Boris Personal Workflow Boris walks through a setup he first shared on Twitter six months ago and didn't expect to surprise anyone. It has since changed: most of his work now happens from his phone, through the Claude app's code tab, where he keeps five to ten sessions each running a few hundred agents. The tool he reaches for most is the loop — fire-and-forget agents that grind on a task and report back. > *"I sort of feel like loops are the future at this point."* ## [08:51] Future Teams and Generalists Asked what teams will look like, Boris predicts a shift toward generalists. Today a generalist still means an engineer who spans iOS, web, and server; tomorrow it means people who are cross-disciplinary, blending engineering with product and design rather than staying in a single lane. He notes the Claude Code team already skews this way. > *"There's going to be a lot more generalists... generalists that are cross-disciplinary."* ## [10:26] SaaS Apocalypse Predictions Reeder asks the question Boris calls his favorite: if AI makes writing code 10 to 100x cheaper, does the value of software products collapse — a SaaS apocalypse? Boris argues the two things that will actually happen aren't the ones people keep predicting, and uses his guest spot on the Acquired podcast as a detour into why he thinks the conventional framing misses the point. > *"I think there's two things that are going to happen and I don't think either of them is the thing that people have been talking about."* ## [12:57] Audience Q&A Deep Dive The floor opens to the room. An audience member, Dan, asks how much of Claude Code's success Boris attributes to the model versus product decisions — Boris says a mix, roughly 50/50, and won't forecast two years out because the team plans a week at a time. The richest answer is his printing-press analogy: before the press, about 10% of Europe was literate; in the 50 years after, more was published than in the prior thousand, and literacy eventually climbed toward 70%. He uses it to argue that building software is on track to become a near-universal skill. Later questions probe the engineering-versus-world gap, local versus cloud models, and how to parallelize agents with loops, batches, and sub-teams. > *"In the 50 years after the first printing press, there was more literature published in Europe than in the thousand years before."* ## [23:35] Closing and Whats Next For the last question, Boris is asked what kind of product he'd build today that gets more interesting as models improve. He points to Claude Design as a good example — decent now, much better soon — and teases features landing for Claude Code in the coming weeks, plus more work on loops, batch, and massively parallel agents, with computer use in the mix. > *"I think loop and batch and things like this around like massively parallelizing agents, that's going to get better."* ## Entities - **Boris Cherny** (Person): Creator of Claude Code at Anthropic; former Anthropic Labs member, now back on the team under Mike Krieger. - **Lauren Reeder** (Person): Sequoia Capital partner; interviewer for this AI Ascent session. - **Mike Krieger** (Person): Chief Product Officer at Anthropic and Instagram co-founder; leads the reunited Claude Code team. - **Anthropic** (Organization): AI lab behind Claude and Claude Code. - **Claude Code** (Software): Anthropic's agentic coding tool, originated in Anthropic Labs alongside MCP and the desktop app. - **Anthropic Labs** (Organization): Internal incubator where Claude Code, MCP, and the desktop app were built. - **Product overhang** (Concept): Model capability that outpaces the products built on it — the gap Boris set out to close. - **The loop** (Concept): Fire-and-forget agent runs that work a task continuously and report back; Boris's most-used workflow. - **SaaS apocalypse** (Concept): The thesis that cheap AI-written code collapses the value of software products — which Boris pushes back on. - **Printing press analogy** (Concept): Boris's frame for AI coding — literacy went from ~10% to ~70% over centuries; software-building may follow.

Jim Fan, lead of Nvidia's embodied AI research, outlines the transition from language-centric models to World Action Models (WAM) that simulate physical reality. He details a roadmap toward the 'Physical Turing Test' and autonomous factories by 2040, driven by video pre-training and human egocentric data scaling. ## [00:00] Introduction Host Sonya Huang introduces Jim Fan, who leads Nvidia's embodied autonomous research group. Fan reflects on his early days as an intern and the excitement surrounding the future of robotics. > *robots are just one of the most thrilling things that's going to happen.* > *[0, 12]* ## [00:30] DGX One Origin Story Jim Fan recounts the 2016 delivery of the first DGX-1 by Jensen Huang to Elon Musk and the OpenAI team. He highlights how this moment catalyzed the deep learning revolution that led to current AI breakthroughs. > *If you believe in deep learning, deep learning will believe in you.* > *[1, 26]* ## [01:42] The Great Parallel Fan proposes 'The Great Parallel,' applying the successful LLM scaling playbook to robotics. Instead of predicting the next token in a string, the goal is to predict the next physical world state through simulation and alignment. > *instead of simulating strings can we simulate next physical world state?* > *[2, 56]* ## [03:31] Robotics Endgame Setup The strategy for achieving the robotics end game is divided into two primary pillars: model strategy and data strategy. Fan notes that while LLMs are in their final 'boss fight,' robotics is just beginning its scaling journey. > *It boils down to two things, model strategy and data strategy.* > *[3, 32]* ## [03:39] Why VLA Falls Short Visual Language Action (VLA) models are criticized for being 'head-heavy' on language while lacking a fundamental grasp of physics and verbs. Fan argues they are better at encoding static knowledge than dynamic physical interaction. > *VLAs are great at encoding knowledge and nouns, but not so much at physics and verbs.* > *[4, 8]* ## [04:32] Video World Models Fan explains how video models like VEO3 learn internal physics—such as gravity and buoyancy—simply by predicting pixels at scale. These models act as simulators that can solve mazes and plan visual sequences internally. > *Physics emerge by predicting the next blob of pixels at scale.* > *[5, 15]* ## [06:09] DreamZero World Action Nvidia introduces 'Dreamer' and World Action Models (WAM), which jointly decode future world states and motor actions. This allows robots to perform zero-shot tasks by 'dreaming' the correct motion sequence before executing it. > *Dreamer jointly decodes the next world states and next actions.* > *[6, 29]* ## [07:46] Scaling Data Collection To overcome the physical limits of teleoperation, Fan discusses Universal Manipulation Interfaces (UMI) and exoskeletons like Dex-UMI. These tools allow humans to collect high-dexterity data directly without the robot being in the loop. > *we're able to break the curse of 24 hours per robot per day* > *[10, 6]* ## [11:06] EgoScale And Scaling Laws Fan introduces Ego-Exo, a policy trained on 21,000 hours of human egocentric video. This research uncovered a neural scaling law for dexterity, showing a mathematical relationship between pre-training volume and robot performance. > *we discovered this neural scaling law for dexterity.* > *[12, 39]* ## [15:39] DreamDojo And The Roadmap Fan outlines the roadmap to 2040, including the Physical Turing Test and 'lights-out' factories. He introduces Dream Dojo, a neural simulator that replaces classical physics engines with data-driven world models. > *I can say with 95% certainty that we'll get to the end of the end game... by 2040.* > *[19, 19]* ## Entities - **Jim Fan** (person): Lead of the embodied autonomous research group at Nvidia. - **Nvidia** (organization): The technology company developing the hardware and software for the robotics end game. - **Jensen Huang** (person): CEO of Nvidia, mentioned for delivering the first DGX-1 to OpenAI. - **OpenAI** (organization): The research lab that received the first DGX-1 for deep learning development. - **DGX-1** (product): The world's first deep learning supercomputer delivered in 2016. - **VEO3** (model): A video world model capable of simulating physics and visual planning. - **Dreamer** (model): A policy model that predicts future world states and actions simultaneously. - **Ego-Exo** (project): A robotics pre-training framework using large-scale human egocentric video data.

Andrej Karpathy explores the paradigm shift from traditional programming to Software 3.0, where LLMs act as programmable computers driven by context. He details the transition from 'vibe coding' to 'agentic engineering,' emphasizing that while AI handles execution, human taste and understanding remain the ultimate bottlenecks. ## [00:00] Introduction Stephanie Zhan introduces Andrej Karpathy, highlighting his foundational work at OpenAI and Tesla. She notes his unique ability to simplify complex AI shifts and introduces the concept of vibe coding. > *He has a rare gift of making the most complex technical shifts feel both accessible and inevitable. [00:22]* ## [00:44] Feeling Behind as a Coder Karpathy describes a turning point in December 2023 when agentic tools began producing perfect code without manual intervention. This shift led him to adopt vibe coding, trusting the AI to handle complex workflows autonomously. > *I just start to notice that with the latest models the chunks just came out fine. [01:29]* ## [02:28] Software 3.0 Explained Karpathy defines Software 3.0 as a paradigm where the LLM acts as a programmable computer and the context window serves as the primary programming lever. This follows Software 1.0's manual rules and Software 2.0's data-driven weight training. > *Software 3.0 is kind of about your programming now turns to prompting and what's in the context window is your lever. [03:20]* ## [03:44] Agents as the Installer Using the installation of OpenClaw as an example, Karpathy explains how agents replace rigid bash scripts with intelligent, environment-aware execution. This approach allows the AI to debug and adapt to specific system requirements autonomously. > *The agent has its own intelligence that it packages up and then it kind of like follows the instructions. [04:29]* ## [04:49] Menu Gen vs Raw Prompts Karpathy contrasts his custom-coded MenuGen app with raw prompts to models like Gemini, concluding that many traditional software layers are now redundant. He emphasizes that AI can now perform general information processing that was previously impossible with structured code. > *The software 3.0 paradigm is a lot more kind of raw. It just your neural network is doing more and more of the work. [06:11]* ## [07:37] What’s Obvious by 2026 Looking toward 2026, Karpathy envisions neural computers that process raw video and audio directly. These systems would use diffusion models to generate dynamic user interfaces, potentially making traditional UI code obsolete. > *You could imagine completely neural computers... a device that takes raw videos or audio into basically what's a neural net. [08:22]* ## [09:41] Verifiability and Jagged Skills AI models develop 'jagged' capabilities, peaking in verifiable domains like math and code due to reinforcement learning rewards. Karpathy notes the paradox where a model can refactor a massive codebase yet fail simple logic. > *state-of-the-art models today will tell you to walk [to a car wash] because it's so close... This is insane. [11:36]* ## [13:39] Founder Advice and Automation Model performance is heavily dictated by the specific data distributions chosen by frontier labs. Karpathy advises founders to explore the 'circuits' of these models to understand their strengths or use fine-tuning to fill gaps. > *we are slightly at the mercy of whatever the labs are doing, whatever they happen to put into the mix. [12:57]* ## [15:46] From Vibe Coding to Agent Engineering While 'vibe coding' raises the accessibility floor, 'agentic engineering' focuses on maintaining professional quality. This discipline involves coordinating powerful but stochastic agents to accelerate development without sacrificing the engineering bar. > *agentic engineering is about preserving the quality bar of what existed before in professional software. [16:07]* ## [25:17] Agents Everywhere and Learning Karpathy advocates for agent-native infrastructure, expressing frustration with human-centric documentation. He argues that while thinking can be outsourced to AI, human understanding remains a critical bottleneck for directing agents. > *You can outsource your thinking, but you can't outsource your understanding. [28:10]* ## Entities - **Andrej Karpathy** (person): AI researcher and former Director of AI at Tesla and founding member of OpenAI. - **Stephanie Zhan** (person): Partner at Sequoia Capital and host of the discussion. - **Software 3.0** (concept): A paradigm where LLMs act as programmable computers via prompting and context. - **Agentic Engineering** (concept): The professional discipline of coordinating AI agents to maintain software quality. - **MenuGen** (project): An app Karpathy built to OCR and visualize restaurant menus, used as a case study. - **OpenAI** (organization): AI research company co-founded by Karpathy. - **Gemini** (ai-model): Google's LLM used in Karpathy's software comparison. - **Vercel** (organization): A cloud platform used by Karpathy to deploy projects.