The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 226 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 2 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Kalinowski: OpenAI lacked proper governance around its Department of Defense deal
“And I feel that what happened with the decision-making, the speed of the decision-making, the governance, and the lack of defined guardrails around the announcement of the Department of War deal is not how I thought it should have been done.”
Wu: Blindly following customer requests leads to AI product local maxima
“Because the field is changing so much at any point in time, you know, a lot of people are kind of in this local, local maximum. And if you just blindly listen to your customers, they'll, they'll be like, yeah, I want a better vector store. Like I want a better…”
Godin: Claude delivers kindness and humility, while ChatGPT overpromises
“ChatGPT's reputation with me is not good because it regularly over promises and under delivers, and it does it without kindness or humility. Whereas Claude, I don't know how they did it, at least in my experience, brings kindness and humility.”
Schwartz: AI Startups Cannot Unseat Google Purely on Search Product Quality
“I don't think there's a chance, and this is the biggest takeaway, I don't think there's a chance that ChatGPT or Perplexity or Claude or any of these other LLM startups, because they're startups now, even, even OpenAI as a startup with Bing's part, with Micros…”
Seshan: Starting and finishing writing yourself prevents cognitive deterioration from AI
“Start yourself and yourself with a piece of writing and that doesn't deteriorate your thinking.”
Seshan: AI products fail unless built for models 2-3 months out
“You fail if you build for where the models are now. You fail if you build for where you think the models will be in a year. Like, both outcomes are equally wrong, and I'm sure many people have talked about this, but both outcomes are really equally wrong. If y…”
Seshan: Long docs no longer signal rigor; prototypes and mocks win
“A long dock is not a signal that you thought through something because you can easily produce a long dock That indicates that you haven't, and so actually the point, maybe one of the biggest changes I've experienced personally in my day-to-day, which has been …”
Silber: AI interfaces must eliminate mode switching to reach billions
“And eventually what we want, what we know, and we think about this a lot in design is like, it will true, like, Things will truly become like for billions of users when they don't have to like think about a switch or a mode or anything like that. And so we kno…”
Silber: OpenAI engineers see 10x-100x productivity gains while designers have not
“One of the big things that came up was, you know, we see that You know, engineers have 10 or sometimes like hundred extra productivity. But our design team hasn't, right? Because the design process still takes time.”
Ambrosino: 100% of current OpenAI product code is written by AI
“Cause if you're using the goalposts from last year, it's like, well, a hundred percent of our product right now is AI written code.”
Ambrosino: OpenAI Codex would have failed if released three months earlier
“I like, I am very confident that the codex app that we released in February, if that had been ready. In November, it would have absolutely failed in the market. And the only difference was the models between November and February.”
Evans: Picking AI winners today is like predicting Excite versus Yahoo
“And so then you can kind of get into calling those races where, again, it's like being in 1997 and saying, well, is it going to be Excite or Yahoo? And the answer was no, generally.”
Schoening: Dario Amodei proved his OpenAI success wasn't luck at Anthropic
“Dario is that she wasn't, oh, he wasn't just lucky once at OpenAI. He did the same thing twice and it was successful twice.”
Willison: Coding agents become truly productive only in permissionless 'unsafe' mode
“I think a lot of people who haven't got on board with coding agents yet, haven't tried them in the unsafe mode. They're using coding agent where it's like, oh, can I run this piece of code? Can I edit this file? And that means you have to pay complete attentio…”
Willison: OpenAI and Anthropic didn't build OpenClaw due to security risks
“The reason OpenClaw took off is Anthropic and OpenAI could have built this and they didn't because they didn't know how to build it securely. If you're an independent third party, you don't have that restriction. You can just Build something and put it out the…”
Wu: AI tools widen the productivity spread across engineering teams
“Codex really empowers, like, top performers to get a lot, like, to be a lot more productive, and so it really, like, and I think this may be true for AI more broadly, like, across society, which is, like, the people who really lean in, or, like, the people who…”
Wu: AI startups should build for capabilities that are 80% viable
“My general advice, and I've been giving this to people for a while, and I think it's still true today, is make sure you're building for where the models are going and not where they are today. You know, the, it's clearly a moving target, and I think a lot of t…”
Wu: OpenAI engineers using Codex open 70% more PRs
“So they're actually opening 70% more PRs and than the engineers who aren't using Codex as much and the gap is widening.”
Ubiquitous code generation will increase demand for human software engineering skills
“And as you make code even more ubiquitous, it's actually just going to be used for many more purposes. And so there's just going to be a ton more need for people with this, like humans with this competency.”
The best way for AI models to use computers is writing code
“It turns out the best way for models to use computers is simply to write code.”
Human prompt writing and validation are the main limits to AI productivity
“I think that the current limiting factor, I mean, there's many, but I think a current underappreciated limiting factor is like literally human typing speed or human multitasking speed on like writing prompts. And like, you know, you were talking about, it's li…”
A dedicated AI browser provides superior context compared to desktop screenshots
“A lot of work is done in the web, and if we could build a browser, then we could be contextual for you, but in a much more first class way. We weren't hacking like other desktop software, which have like very varied support for like what content they're render…”
Horowitz estimates OpenAI commands 80% of total AI industry revenue
“Open AI is probably 80% of the revenue in AI or something like that now.”
Balfour: ChatGPT will be the next major product growth distribution platform
“My prediction of the new distribution platform will be chat GPT. In some ways that people probably already think it's happening in some ways that it won't.”
Balfour: OpenAI is about to launch a ChatGPT third-party platform
“There's going to be what they do with like a chat GPT search experience, but I think the bigger thing will be whatever they do with launching a third party platform on top of chat GPT. There's a bunch of signals that they're about to that they're about to laun…”
Turley: AI companies must separate product velocity from rigorous model safety
“I think it's been really important to separate out, you know, the product development velocity, which has to be super high, from, okay, for things like frontier models, there actually needs to be a rigorous process where you red team, you work on the system ca…”
Turley: There is no distinction between the AI model and the product
“The one thing we've learned with ChatGPT is that there really is no distinction between the model and the product, like the model is the product and therefore you need to iterate on it like a product.”
Turley: GPT-5 achieves state-of-the-art results on math and reasoning benchmarks
“One way to look at that is academic benchmarks on many of the standard ones whether or not it's math or reasoning or, you know, just raw intelligence, this model is state of the art.”
Turley: AI features that do not scale with model intelligence should not ship
“Like if, you know, we're shipping a feature and it doesn't get two X better as the model gets two X smarter. It's probably not a feature we should be shipping.”
Turley: ChatGPT is AI's MS-DOS and 'Windows' is not built yet
“I still feel like ChatGPT feels a little bit like MS-DOS. We haven't built Windows yet, and it will be obvious once we do, but, you know, there, there's something that feels a little bit like, imagine MS-DOS, like, gone viral, and you were just trying to, like…”
Mann: Sam Altman managed OpenAI across safety, research, and startup tribes
“One weird thing about OpenAI is that while I was there, Sam talked about having three tribes that needed to be kept in check with each other, which was the safety tribe, the research tribe, and the startup tribe.”
Deng: AGI requires product builders to channel raw intelligence into value
“AGI is just necessary, but not sufficient. A lot of the value is still gonna require a bunch of hustle from a lot of builders to really turn that new source of energy and channel it into something that we humans want to use that solves some of our problems. An…”
Deng: Future generations will not need to write code for work
“I don't actually think he's gonna have to code when he grows up. I think that's gonna be a solved problem, but it's a very, very valuable skill because I think learning to program is learning how to think structure in a structured way, right?”
Rauch: DeepSeek proved AI thinking token visibility is a killer feature
“So the deep seek Stream the thinking tokens moment was a very big moment for industry, I think, because open AI did have the technology, but they decided that for competitive reasons, which, you know, are, it's a reasonable thing to think, no pun intended. The…”
Weil admits OpenAI no longer holds a massive 12-month lead
“It used to be that OpenAI had this, like, massive model lead, you know, 12 months or something ahead of everybody else. That's not true anymore. You know, I like to think we still have a lead. I'd argue that we do, but it's certainly not a massive one.”
Weil: Prompt engineering will become obsolete as AI matures
“I want to kill the idea that you have to be a good prompt engineer. I think if we do our jobs, that stops being true. You know, it's just one of those, like, sharp edges of models that experts can learn, but then you just, over time, you shouldn't need to know…”
Weil: OpenAI will not build industry-specific vertical AI products
“There are immense opportunities in every industry and every vertical in the world to go build AI based products that improve upon the state of the art. And there's just no way we could ever do that ourselves. We don't want to, we couldn't, if we did want to.”
Weil: OpenAI manages 400 million weekly users with 30-40 support staff
“With 400 plus weekly 400 plus million weekly active users, we get, you know, a lot of inbound tickets, right? I don't know how many customer support folks we have, but it's not very many. 3040, I'm not sure. Way, way smaller than you would have at any comparab…”
Weil: OpenAI minimizes scaffolding because rapid advances erase model limits
“We don't spend that much time building scaffolding around the parts that don't match that because our general mindset is in two months, there's going to be a better model and it's going to blow away whatever, you know, the current set of limitations are.”
Osika: Perfect AI software creation and integration will arrive within two years
“The last piece of software, how I see that is that it's almost instant to go from what you want to change in a product or what you, what product you want to build. To having it fully working end to end, integrated with any of your existing systems or integrate…”
Nguyen: Anthropic excels at prioritization, while OpenAI takes more product risks
“I would say, like, Antarctic, I learned from Antarctic that, like, They're much better at, like, focusing and, like, prioritization or, like, very, very hard, like, very hardcore prioritization, I guess, and they need to do it. Like, but I think, like, OpenAI …”
Nguyen: Synthetic Data Outperforms Human Data for AI Product Development
“And the reason why I really love, like, synthetic, like, relying purely on synthetic data instead of, like, collecting Data from humans is because it's, like, much more scalable. It's cheap, less than how, like, you literally sample from the model, and you tea…”
Nguyen: AI is not far from autonomous self-improving product development
“And I don't think, like, we are far away from that kind of, like, self-improvement, models becoming, like, self-improved via, like, then, like, the product development is basically kind of, like, self-improving, like, it's kind of, like, its own, like, organis…”
Weil: Frontier AI increasingly generates answers that evaluators prefer over humans
“The models are getting to the point where they can often beat humans at certain tasks. Like, people prefer the model's answers to a human's answers, and so if you're humans writing your evals, like, you know, so what does that mean?”
Weil: Current AI models are eval-limited rather than intelligence-limited
“I think there's a very real sense in which models today are not intelligence limited. They're eval limited. They can actually do much more and be much more correct on a wider range of things than they are today, and it's really about sort of teaching them.”
GPT-5 will normalize as a useful tool rather than changing everything
“GPT-V is like surely going to be extremely useful and like solve some whole new echelon of Problems. Hopefully they'll be faster. Hopefully it'll be better on all these ways, but like fundamentally the same problems that exist in the world are still going to b…”
OpenAI models will likely never match vertical tools like Harvey
“Our models are probably never going to be as capable as some of the things that Harvey's doing because like our goal and our mission is really to solve this like very general use case.”
Adding AI researchers decreases group productivity when GPU capacity is constrained
“In a world where you're constrained by the amount of GPU capacity that you have as a researcher, which is the case for OpenAI researchers, but also researchers everywhere else, like Each new researcher that you add is actually, like, a net productivity loss fo…”