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 3,312 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 43 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
Haas: AI Won't Push-Button Design Frontier Chips Outperforming Leading Architectures
“Now, if you go into the tool and say, design me something that's 10% faster than Vera Rubin, 20% cheaper, and 30% more efficient on this model, You're not gonna get press a button and have it happen right away.”
Arm CEO Rene Haas: AI chip supply constraints will last 3–5+ years
“That's not going to stop in 12 months. It's not going to stop in 24 months. I think we're going to be in this constrained environment for Three to five years at least.”
Haas: AI compute capacity is not even close to oversupplying demand
“The bubble in terms of, are we over oversupply to demand? Not even close. And I think again, that's because the demand is insatiable.”
Haas: Edge AI Will Be a Sweet Spot for Arm Architecture
“And in fact, as you get to the smaller footprints, where more and more AI is going to take place, that's going to be a sweet spot for Arm, because the CPU's table stakes anyway, you have to have it to do all the things that are required in the edge device. But…”
Ehrlich says forward-deployed engineers are essential for fast enterprise AI adoption
“Now it seems that you're coming to a large legacy enterprise. They really want to adopt the ad because they have to, the problem is they don't know how to do it. They understand that their processes are very long. There's something that takes a year or two or …”
Hodak: Science is the first to restore form vision in blind patients
“Nobody had previously ever been able to restore a form vision image in the mind's eye of a blind patient in this way.”
Hodak: Vision, hearing, balance, and 1 kbps motor control unlock radical neurotech
“If you can, if you get vision, hearing, balance, and a kilobit per second of motor control, you're halfway to the matrix. And this takes you into some like really trippy interpret reinterpretations of medicine.”
Hodak: Science aligns animal brain recordings with AI model representations
“We use that constructively at science like that. So we know that that is. That is true because we can get alignments between animal brain neural recordings and AI model internal representations.”
Hodak: Cardiovascular disease and non-brain cancer are attackable via organ replacement
“The two leading causes of death though, are cardiovascular disease and cancer is not metastasized to the brain. And I think both of those are going to be really attackable through this type of work.”
Taylor: Valar Atomics Will Scale Fission to Make Energy 10x Cheaper
“We have nuclear reactors that are more manufactured than constructed and they're extremely safe. And we think the combination of these two things is going to allow us to scale nuclear fission and make energy 10 times cheaper for humanity.”
Taylor: Rebooting Nuclear Requires Shifting from Civil Infrastructure to Advanced Manufacturing
“If you're going to reboot nuclear and you're actually believe in, in the atomic energy century, which I absolutely do, it's probably not going to look like what it did in the sixties. It's not going to look like civil infrastructure. Because we're not that goo…”
Taylor: Manufactured SMRs Require Development Driven by Hardware Iteration
“What was missing is that people did not realize that you're going to have to pursue that model through hardware iteration. And that's where really where Valor is unique is that Yes, we believe in SMRs. We believe in reactors that are manufactured. But we also …”
Taylor: Most Nuclear Startups Are Just Modeling and Simulation Companies
“So if you look at the nuclear industry outside of Valor, it's mostly a modeling and simulation industry. Like, when you think of a nuclear company, right, there's, like, nuclear companies out there that everyone knows the name of, and you kind of look under th…”
Taylor: Valar aims to build hundreds of thousands of nuclear reactors
“We want to make thousands of reactors, tens of thousands, hundreds of thousands.”
Taylor: Valar will shut off all power and pumps after reactor scram
“Immediately after we scram, we're actually going to turn off the entire electrical supply to the plant. We're going to turn off the circulator. We're going to turn off the RCCS pump. Every safety system in this plant, we're going to shut off, and we're going t…”
Taylor: Nuclear energy causes fewer deaths per unit of power than solar
“Nuclear energy is the safest form of energy, empirically. If you look at the amount of power generated versus the amount of deaths it's even safer than solar, which is very strange, because how does someone die from solar? But actually, most solar is installed…”
Taylor: TRISO helium-cooled reactor geometry is humanity's safest reactor design
“A triso-fueled, graphite-moderated, helium-cooled reactor of the geometry that we've built it here is just about as safe as humanity knows how to build a nuclear reactor.”
Valar Built a $5M Nuclear Reactor Component In-House For $400K
“And we were quoted about five million dollars and we were told it would take about two and a half years to build. And so we said, well, all right, maybe I'd be willing to pay five million dollars, but I'm certainly not going to wait two and a half years. And s…”
Taylor: Valar Atomics will reach fifth reactor while competitors seek debt financing
“We are willing to go build reactors on equity balance sheet because that's going to allow us to prove it years ahead of anybody else, right? Where, where our competitors are still trying to convince, you know fundamentally risk averse you know, financers to fi…”
Taylor: AI and robotics will make energy the sole cost of goods
“When we figure out AI and robotics that allows us to do semi-autonomous manufacturing, energy will become the cost of all things, right?
The cost of buying a thing will become the cost of energy used to make it.”
Sarah Guo: AI making human skill irrelevant is complete nonsense
“I think the idea that human skill is no longer relevant in the world of superhuman AI, just as a product of the technology itself is nonsense.”
Gil: Investors are conflating market size with speed to scale
“And people Are, at least in my experience collectively, at least investing against the fact that they believe the speed is there, which is different from the market size. People are conflating the two things right now, in my opinion.”
Gil: Late-stage AI investors wrongly assume startups have trillion-dollar trajectories
“One is for the mid to late stage technology markets, people are continuing to invest as if there's velocity to get to a trillion for many companies where I don't think there's a velocity. Again, I think the, Some of them will get to 20, some will get to a hund…”
Gil: Fear of frontier labs drives founders into derivative niche markets
“And I think that's why you're seeing a flight to hardware companies. Oh, the labs will never do this hardware thing. And so we'll do that. And American dynamism and niche applications of AI and something that should be provided by an inference cloud. And et ce…”
Gil: Anthropic and OpenAI Should Never Sell in the Near Term
“There's a handful of companies that should never, ever sell, at least any time in the near term. If you're anthropic, you shouldn't sell. If you're opening, you shouldn't sell. You know, there's a handful of these things that should never sell.”
Guo: Researchers have predicted ASI is 18 months away for five years
“The data I have is that a number of very smart and even very self-aware research scientists have felt that, you know, there was some knee in the curve on recursive self-improvement or ASI, 18 months away, every 18 months for the last five years.”
Guo: Compute and complex data are bigger AI limiters than algorithms
“The question of, like, how you are going to go gather that data for less verifiable, more complex domains, or do you run into the actual constraints on the, like, physical compute accessibility side? Like, I think that's probably more of a limiter than this be…”
Gil: Physical compute constraints enforce an oligopoly among leading AI labs
“Yeah, I mean, the physical compute basically reinforces an oligopoly market because what it does is it creates a ceiling On the rate of progress any single lab can get, if effectively you assume the compute is roughly pro rata across the ecosystem to the big l…”
Gil: AI labs are slowing researcher hiring due to compute costs
“I know some labs have slowed down on their hiring of researchers unless they're above a very, very high bar because the cost isn't the research or it's the compute associated with the person.”
Gil: The 'death of SaaS' narrative is overstated due to token budgeting
“That's why I think the death of SAS is a little bit overstated because why would you use tokens on a bunch of SAS stuff that you're not actually paying that much for per year relative to the outcome of those same tokens being invested against a core product or…”
Gil: 1970s Safety Lobby Killed Abundant Clean Nuclear Energy in the US
“We had a safety lobby in the seventies basically kill abundant clean energy for us, right?”
Tokmak: Gen Z job candidates obsess over a permanent underclass mentality
“I see, especially Gen Z in these days when I'm, you know, looking at hiring, obsessed with this permanent underclass mentality. If I don't make my money in the next 18 months, or if I don't learn everything in the world in the next six months, then I am foreve…”
Tang: Suburban delivery requires 300-pound autonomous vehicles, not passenger cars
“If you're trying to solve that three to five mile delivery, In dense suburbs, which is where most of the deliveries happen, the right metaphor is probably a autonomous motorcycle or a scooter or a bike profile vehicle. And you know, it doesn't even be 4000 pou…”
Fang: Frontier AI models underperform tested on messy enterprise data
“It's like, okay, if you like dumb down the problem, maybe the models do well, but like for some reason, and when we actually have it, With the enterprise data and all the real stuff, it's not performing as well.”
Fogel: Booking has grown faster than Airbnb in alternative rentals for 5 years
“For the last five years, we've grown faster than Airbnb in the alternative accommodation area. We do like to like. We've been growing faster the last five years.”
Brown: AI benchmarks must control for test-time compute
“And so I think the proper way to, and so my claim is the proper way to evaluate the models now is you either have some kind of budget for the benchmark, whether it's tokens or cost or time or whatever, or you plot the performance as a function of the amount of…”
Brown: Scaffolding Easily Inflates AI Benchmark Scores Without Real Gains
“It's really easy to show you can do much better than previous benchmarks or previous, previous models on benchmarks by just, for example, scaffolding a bunch of models together. So if you say, okay, well, we're going to, instead of just running this model once…”
Brown predicts AI will zero-shot his entire PhD thesis within one year
“And I wouldn't be surprised if, you know, six months or a year from now, the model is able to do zero shot an entire poker solver, basically my entire PhD thesis in one go.”
Current AI safety frameworks fail to account for test-time compute scaling
“The preparedness frameworks and responsible scaling policies, they don't really account for the amount of tests I'm computed. They just say, okay, well, what's the capability of the model? The problem is we're in a world now where the capability of the model i…”
OpenAI internal model reportedly disproved the Erdős unit distance conjecture
“We used an internal model at OpenAI a few weeks ago to disprove the unit Erdos unit distance conjecture.”
Complex task compute costs fall 10x to 100x per model release
“The model release cycle is every, every couple months we put out a new model that's even more powerful, and so the cost of disproving the Erdos unit distance gesture drops by, like, 10 or a hundred x with every model release cycle. Probably, in some cases, mor…”
OpenAI discourages researchers from using current models on open math problems
“We are trying to encourage people to not spend all their time just, like, Going through all the mathematical open problems, physics problems, and just seeing, pushing the models to their limits to see what they can prove or disprove. Because we really think th…”
Brown: AI cannot invent novel algorithms better than existing research
“Go ahead and like look at all the published work and synthesize that and then try to come up with something novel and it's not able to do it. And I can give it a lot of time and it's still not able to do it.”
Noam Brown: AI community stuck in bad equilibrium publishing static benchmark grids
“I would talk to researchers about we, it makes sense to show the benchmarks with an x-axis, whether it's tokens or cost or time, there should be an x-axis, and everybody would say, like, yeah, that makes sense, we should do that, but. Well, really, their respo…”
Tan: AI inference is shifting CPU-to-GPU ratios toward 1:4 or 1:1
“Right now, the authentic AI and inference, CPU become, you know, highly in demand. And so, you know, versus one to eight in the training CPU to GPU, now I can see one to four, maybe one to one, and I'm delighted CPU become important.”
Tan: Chip manufacturing will reach 10A and 7A nodes despite escalating costs
“So I think I can see, you know, right now we have A and then now going to production of A, I can see 10 and seven. And so I think that path, I think we can get there, but couldn't it be more and more expensive and more difficult to do?”
Tan: Chip startups should target hyperscalers on day one, even granting warrants
“The next thing is look at, it's very important from day one, you'd have to target the first customer. And usually I like the customer is hyperscale. They have the scale. If they like what you have, they're willing to pay millions of dollars next few years. And…”
Tan: Intel Foundry is very distant from TSMC on performance and yields
“And the Foundry side, we are very distant from TSMC. And then in terms of their performance, so that's we have to be humble looking at building the building block, like I mentioned earlier, the IP, the yield, the defect density, and the cycle time to make it m…”