Summary

Altman openly disagrees with Tobi Lütke’s call that 2026 is the year “every business is up for grabs” — he thinks disruption comes slower than the technology allows. He admits he got this wrong once already: at GPT-4 in 2023 he expected rapid software disruption, but “the economy just has so much inertia,” people keep buying from the same companies, and “we’ve all been too ambitious on timelines.” He frames the inertia as a positive that makes the transition “smoother and slower” — though many software businesses will still be very up for grabs.
OpenAI is deliberately becoming more of a platform than a product company, and is killing good products to fund the great one. Altman says they killed Sora last year (“used a lot of compute and not as important as Codex”) and killed the Atlas browser despite calling it “the best web browser,” merged ChatGPT and Codex into a single interface, and will “sell great AI at every point on the cost-performance curve” — one interface to your AGI plus an API, rather than competing with customers across product categories.
Altman’s most striking self-criticism: he doesn’t use his own product the way he should, and he blames a missing “iPhone moment.” Despite having Codex, he still clicks around, pastes between messaging apps, and scrolls email “by revealed preference” — “I must secretly like it.” His diagnosis: all the technological pieces exist, but AI is where smartphones were in the Palm Treo era of 2003–04, “mostly missing the product ideas that made the iPhone the iPhone.” He feels more limited by the amount of useful context AI has on him than by model intelligence — an agent that reads tens of thousands of pages in seconds and advises on decisions is “just on the precipice.”
His core management thesis: running a research lab is power-law startup investing. Your best bet outperforms everything else combined; in 2015 the AGI bet got OpenAI “hammered by all of the intellectual giants of the field,” and again on large language models — while thousands of founders started photo-sharing apps. He hunts “non-standard” researchers with unpopular convictions over “a thin veneer on the same idea,” and concedes some chance a lone “monk researcher” finds an angle the big three labs never considered.
On safety, iterative deployment — not ivory-tower caution — is the argument, but he concedes it gets harder from here. After a decade, OpenAI had built something that would have seemed very AGI-like at the outset, while the predicted alignment failure and world-ending outcome had not happened — “that should update people’s predictions about the future.” He invokes the FAA’s accident-reporting culture as the model. The pivot point: “the smartest people in the world are about as smart as the smartest models in the world. And that’s going to flip” — forcing decisions about when to delay development.
His two named AI risks are loss of control and centralized power, and he attacks the industry’s implicit bargain as “a very anti-human sales pitch.” The caricature: “dear peasants, we will bequeath upon you these gifts” of cured disease and cheap goods in exchange for autonomy — driven by “fear and power.” The counter-thesis with investor-relevant implications: “we are about to see the greatest boom in people starting smaller businesses that we have ever seen,” and the field, including OpenAI, has neither talked about it enough nor built products to accelerate it.
The Peter Thiel anecdote is the episode’s best strategy lesson: roughly two months after ChatGPT launched, when insiders wanted to chase five or six other ideas, Thiel said doing anything else was “an obvious mistake.” “The power of this is the power of the Google text box” — no feed, no network effect, just an empty box people had chased Google’s model with for 20 years. “We went super hard on it and it was great.” Paul Graham’s launch-when-embarrassed doctrine was so internalized that Altman doesn’t think he asked Graham before shipping ChatGPT.

Deep dive

1. Tobi Lütke is one of the most interesting CEOs — but Altman rejects his 2026 timeline

Altman’s case for Lütke as one of the most interesting CEOs: “at every moment along the curve” he’s been the most forward-leaning — writing software himself, sending “extremely accurate, detailed, cutting-edge feedback,” declaring “we are not an NPC company” before anyone else, and running “six, eight months ahead of any other CEO” with “no hype.”
Senra relays Lütke’s claim that 2026 will be the year “every business was up for grabs,” that someone will build the AI-native Shopify — “and it’s going to be me.” Altman’s pushback, worth keeping: “I disagree on the timeline. I think it’s going to take a little bit longer,” and not literally every business — anti-AI businesses selling “authentic non-technological experiences” get harder to compete with as AI improves.
The confession behind the disagreement: at GPT-4 in 2023 Altman expected much faster software disruption. “The economy just has so much inertia… I’m grateful for it. But I think it means we’ve all been too ambitious on timelines.” Senra’s parallel: Larry Ellison in the ’80s (“it’s a people problem”) and people still driving to Blockbuster after Netflix shipped DVDs.

2. Altman’s revealed preference: he doesn’t use his own magic

“I have been waiting for this question.” Despite Codex, he’s used computers the same way for 20 years — pasting between apps, “scrolling mindlessly” through email — because “there’s something in my mind that is encoded that doing this kind of stuff is what it means to work… I must secretly like it.”
His diagnosis is mostly product failure, not simply user failure: everyone is “straddling these two worlds.” The analogy — smartphones before the iPhone, his Palm Treo circa 2003–04: “we have all of the technological pieces, but we have not had the iPhone moment of completely changing how someone interfaces with technology.”

3. Compute, research, and the power law — how he actually runs OpenAI

Against the Steve Jobs / patient-zero comparison (via mutual friend Josh Kushner), Altman says most of his effort goes to research and compute: “if we can get that right, I believe that everything else will follow.” Compute suits him — complex supply chains, partnerships, chip design, power systems, and financing “what is probably already or at least rapidly becoming the most expensive infrastructure project in history.”
The closest thing to his old job as startup investor is managing a research program: the power law applies — “your best investment will outperform all of your other investments put together.” In 2015 the AGI bet got OpenAI “hammered by all of the intellectual giants of the field,” then again for betting on large language models; the lesson from investing was “high-risk bets are okay as long as you take the ones where if they work it’s super valuable.”
His filter for researchers mirrors his filter for founders: “non-standard” people “willing to stand by convictions that are very unpopular, may well be wrong, but if right at least they’re going to be really right” — versus “a thin veneer on the same idea that everybody else has.” In 2015 there were DeepMind and “1 or 2 others” pursuing AGI, alongside many thousands of photo-sharing startups; today everyone wants an AI lab while “two, three, whatever” are doing something completely new that only became possible as AI got this good.
Senra recalls Dario putting a small percentage on a “monk researcher” cracking AI from an unconsidered angle. Altman: “There is some chance of that for sure, and I love that. That’s why stuff stays exciting.”

4. “The more something seemed impossible, the more intrigued I was”

The origin: a nerdy kid in St. Louis into robots and sci-fi, who came to college to study AI in ~2005 — where a professor told him “the one thing we know doesn’t work is deep learning… the most guaranteed way to have a bad career.” His self-described distinguishing trait wasn’t the interest but the response: “the weird thing about me was like, okay, let’s try.”
He flags his own memory as unreliable — “how much my current work has colored my memories” — but is now most animated by AI for scientific discovery, “even more important than automation of other tasks,” citing a reread of The Beginning of Infinity.
On his rare investor-first-then-founder path: as an investor “you kind of watch all the crux moments” — firing an executive, high-stakes strategy shifts — across many companies rather than only the limited operating experience of your own five or ten years. “The wealth of the data set that I had there was awesome. I strongly recommend it.” Senra notes Doug Leone could name only one comparable case: the founder of Nubank.

5. Shannon and Turing were right — and humans stay the point

Senra’s history: Shannon and Turing met daily for coffee at Bell Labs in the 1940s, expected machines smarter than humans by ~1955, and wanted them to solve math, write poetry, cure disease. Altman: “I am so sad they are not here to see it, because they were so right about everything… I think they would have said, all right, you’ve done it.”
On whether people will prefer AI podcasts: Senra argues that connection-based work gets more valuable in a post-AI world. Altman agrees that most people will remain fundamentally wired to care about, be around, and interact with other people; those who “only want to communicate with computers” are “a tiny percentage of humanity,” which is why “the world is on the whole not going to be that different even with superintelligence.”

6. The two risks, and the “dear peasants” sales pitch

Altman’s two biggest worries, which he notes are in tension: loss of control (“AI somehow just becomes too powerful in a way that we can’t guarantee the control we want”) and centralization (“one company or model or person with too much power”). Both are anti-human outcomes: “people are the whole point of this all.”
His caricature of some in the field: “we’re going to give the world a cure to all disease and make stuff really cheap in exchange for people giving up their autonomy and impact over the future” — plus “absolutely rampant inequality.” Rendered as: “dear peasants, we will bequeath upon you these gifts… and just trust us.” His verdict: “a very anti-human sales pitch.” Why people make it: “fear and power” — the magnitude of perceived risk justifies trading liberty for safety, which “ends up a way to justify a lot of power-seeking behavior.”

7. Iterative deployment is the safety strategy — and it gets harder now

Where he splits with doomers: agrees on erring toward safety and caution at each level; disagrees “that it’s an unsolvable problem.” At the outset, confident positions were that AGI would not happen within 10 years and that, if it did, alignment would fail and the world would end. A decade on, OpenAI had built something that would have seemed very AGI-like then, while the predicted world-ending outcome had not happened — “that should update people’s predictions about the future.”
The method is straight from startups: ship, get feedback, see where it breaks. A billion weekly users in under four years with broadly-considered-safe deployment — “there is no way we could have done that in an ivory tower.” His template is the FAA: “extremely robust accident reporting, extremely clear-eyed,” clear postmortems, learnings shared with other AI builders.
The hedge, exactly as hedged: “the smartest people in the world are about as smart as the smartest models in the world. And that’s going to flip right now.” Unknown unknowns get harder in absolute terms, forcing “difficult decisions about when we delay development… or wait longer to really study this more.”

8. Everybody uses AI, everybody hates AI — a self-inflicted wound

Altman’s answer to Senra’s puzzle: some resistance is healthy societal inertia, but much is the field’s own messaging — people “saying there’s a 25% chance we’re going to destroy the world and we’re going to race ahead to do it,” or “50% of the jobs are going to go away in the next year and we hope you all are okay.” Missing entirely: any case for “why it’s important that people have more power and personal freedom in the world, not less.”
Senra’s counter-example from the Intel Trinity: Noyce, Grove, and Moore stopped work to educate potential customers, investors, and the country about the microprocessor, at one point running more classes than the local community college’s entire catalog. Altman: “No excuses. We should be doing more. We’ve tried versions of this. We haven’t gotten it quite right.”
The under-told upside: “we are about to see the greatest boom in people starting smaller businesses that we have ever seen” — starting a company has required privilege, luck, and resources, and AI is empowering that — yet “the field, including us, has not talked about that enough… and we have not built enough products to accelerate that.”

9. The bottleneck is shifting from intelligence to context

With the latest models “pretty smart,” Altman says “I feel more limited at this point by the amount of useful context AI has on me” — he wants an agent reading every internal Slack post, every customer story, more research papers than he has energy for, then “bring that context to bear and give me good advice when I have to make a decision.” His framing: “we are just on the precipice” of a new axis where “there is no one that can read tens of thousands of pages of context in some small number of seconds.”
Senra’s live demonstration: his own tool, trained since 2018 on every book note, highlight, and Founders transcript, which he queried while making the recent Claude Shannon episode (“what did Bob Noyce say about this?”). Altman: “That is so cool.”

10. Platform, not product — and the discipline of killing good ideas

The structure: ChatGPT and Codex just merged (Codex “sort of unfortunately named” — it was never just coding); the offering is “a single interface to their own personal or their company’s AGI” plus an API — “we will sell great AI at every point on the cost-performance curve.” Not competing with customers, not subsuming the economy: “100 million new businesses and 8 billion people” using it in all kinds of new ways.
The sacrifices, named specifically: “last year, for example, we killed Sora — good product and fun and cool, but used a lot of compute and not as important as Codex,” and the Atlas browser — “it was the best web browser, but not as important as somewhere else we could put that talent.” The residual focus: “general intelligence for knowledge work and eventually for science,” plus everything upstream — own chip, own data centers, infrastructure software, training.
His candor on the discipline itself: “killing the good ideas to go after the great ideas is the hardest lesson for any entrepreneur… I’m terrible at this. I know I’m bad at this.”

11. Thiel’s text box, Graham’s finger-wag, and who Altman actually calls

Altman describes two explicit counsel categories: long-tenured OpenAI researchers with “shared language, intuition, standards” he can’t replicate outside; and, for non-obvious business problems, Paul Graham and Peter Thiel — “if what LLMs do is predict what word comes next, those are two of the people that I can predict the least what word is going to come next.”
The Thiel specimen: roughly two months after ChatGPT launched, growth felt “unstable, kind of almost low-value,” and insiders had five or six alternative focuses. Thiel: “it’s an obvious mistake to do anything about this besides the fact that it’s growing… the power of this is the power of the Google text box” — no feed, no network effect, none of the Silicon Valley checklist, and that’s precisely why it works. “We went super hard on it and it was great.” Senra’s label: “simple genius.”
Graham’s contribution was a recurring doctrine, though his specific suggestions varied — the “you know what you should do” finger-wag was sometimes great and sometimes terrible. The launch-when-embarrassed principle was so internalized that Altman doesn’t think he asked Graham before launching ChatGPT: “I knew what he would say.” Senra’s Munger story lands here: Munger said he and Buffett “never” talked anymore — “Buffett can just pretend to pick up the phone and he already knows what I’m going to say.”

12. Four and a half years without a product — the whiteboard day

OpenAI’s founding violated the core YC playbook: 4.5 years from founding to first product. His clearest memory — early January 2016, 11 or 12 people at Greg Brockman’s apartment, first-day-of-school excitement, someone fetches a whiteboard, then: “you just feel the energy in the room collapse. None of us know what to do.” Also his dominant memory of the era: “trying and failing to raise money. So much effort, so frustrating.”
They sought advice from Alan Kay and other veterans of the great labs — this was the Bell Labs / Xerox PARC vanity-project moment in Silicon Valley — but “some of it didn’t translate,” and unlike Bell Labs or Polaroid, OpenAI had no monopolistic cash machine.
Substitutes for customer signal that worked: the Dota 2 leaderboard where ideas competed objectively, and “the incredible power of external demos for an eminent person that the researchers really wanted to impress.” What didn’t: “fake deadlines.” Through “mostly chaotic stumbling,” the unsupervised sentiment paper became GPT-1, and the scaling-laws work gave them conviction to scale.

13. YC as an influential band, learning from success, and letters to his son

On YC’s influence — Senra’s uncertain band-that-influenced-every-band analogy: Altman suggests The Band, while Senra thinks it might be the Velvet Underground. Beyond returns, YC’s philosophy (iterative deployment, technical people in charge, betting on young ambitious people without proven resumes) rewired the whole ecosystem. Altman’s categorical claim: run the clock from 2004 with the technology but the old ecosystem, and “I do not think OpenAI would have been possible.”
His contrarian learning claim: “you learn some lessons from failure, but I think you learn way more from success.” Failure lessons are “fairly generic” grit-and-determination or hard to causally attribute (“most things don’t work, so there’s a lot of reasons why things don’t work”); success lessons, applied forward, “were very good and I should have applied those more.” Senra’s frame he “extremely strongly” endorses: founders don’t need new lessons, they need church — the constant reminder to “talk to your users more, ship products earlier, hold a higher bar for who you recruit.”
The closing beat: after his first kid was born, Altman wrote him Sunday letters about his week — “only ever did like eight of them” — and Senra explains the mechanism: “you really can’t hide behind anything… I really care what my kid’s going to think about me, so this thing happened, didn’t feel great about it, better do it differently next week.” Senra’s push, citing Michael Moritz’s The Little Kingdom: keep writing, because founders who don’t journal later say the same thing — “I wish I journaled.”

22-Year-Old Roboparty CEO 黄一 on 5 Rounds and $100M+ in Embodied AI Stealing Reasoning Traces from Proprietary LLM APIs — Ilia Shumailov & Alexander Panfilov