AI model

“That's where the AI models are really, really good when they're just swapping things over from one thing to the next. It's hard for them to do, like, truly novel new things, but they're pretty good at just, like, translation, transform, transformation, right?”

“I think it's almost certainly worth it to be five percent smarter on these models, and I think that the economics are much easier, which is why the market's ripping so much.”

“Eventually you will be able to use a model to use another model. So your, that your preferred model, if you want to do something with another model, but you say, Hey go find the best model to do this task or like create this type of content and it'll just do i…”

“It's getting to a point pretty much now where the models are becoming smarter than the average person. So the average person cannot tell the difference Like if the model gets 10% smarter.”

“I think there's this period of time where it's true. People want to see how it's working through something, but then eventually when you have that level of trust with the model or the app that you're using, you just want it done.”

“I think we will get there. It's just the question of like, you know, what if it's a thousand order of magnitudes?”

“The advances in AI are so fast. That the way we build models today with fine tuning and different kinds of training may be also obsolete a year, two years from now. So there may be models that are built that don't require any human training. So revenues that a…”

“I know like I pay for, you know, whatever, six different models, like that's probably not gonna be true in three years, right? Like there's probably gonna be one that I'll end up paying for too. There's certainly no reason to pay for six right now.”

“Stagnation is actually the default, you know, it's by default, it's not the case that we're going to have this, you know, the next thing.”

“I do think that we should think a lot about like model welfare and like, you know, that like, you know, well-being and like health of these models as they get much smarter than they are today, right?”

“I think the real bottleneck is just, we got to make them smarter. I don't like, they're already so cheap. It's like two cents per million tokens or something ridiculous. I think the real bottleneck for me using them more is not their price, but them being more…”

“We think the capability overhang we have in the models that we already have today, and we will have absolutely in six months is immense.”

“You know, you see a lot of like model drops that come out, but they don't actually provide the real hooks and they do very well in the benchmarks, right? They do very well on like kind of the public leaderboards. But they don't actually provide the hooks that …”

“I don't think we've seen like across other product services, people actually invest in like real-time audio and real-time video and image stuff. And I think that's like the next iteration of the UX of how people are going to interact with AI models.”

“It's actually hard to appreciate some of those moments if there isn't a product experience that brings it to life. And I actually think that's been a lot of the gap.”

“My, our belief is that nobody's gonna have a lead in them having the best AI model for more than a few weeks.”

“The quality bump over things that are much smaller is just like the, we're hitting diminishing returns on capital investment is a lot of it. Like they're taking out of the API because like the, they can sell other things with the same GPUs and make more money …”

“Well, I think we're well past the threshold where the models are powerful enough that you can be using them to automate massive things.”

“I think one clear example here is the Amy, like probably the hardest auto gradable, like human math eval, at least in the US. And yeah, the models are consistently getting like 90 plus percent on these.”

“Even if there's no software progress out of that point, that algorithms aren't improving, just that ability to learn on the job from everything in the economy would functionally produce what looks like a super intelligence, right? No human will be, will have m…”

“Part two, we're, I think we're going to see a groundswell in how intellectual property is dealt with. A groundswell change in how IP is dealt with. Because there was, for a long time, it's been publish or perish. Get your stuff out there. Get it into journals.…”

“You can do all the display ads you want. You could do video there all you want. But if it's in the response that comes back from the model, boy, you're gonna have lots of problems.”

“Literally all the code I've written in my life will be Replaced by what will be able to have been produced by an AI model within the next five years.”

“The one new thing is AI models have, because of scaling laws, have much stronger data network effects and data modes than we've ever seen in the past.”

“We're going through this whole electrification, so to speak, period over the next three years of putting the latest models into our apps. And for us to do that, we have to basically rebuild the apps from the ground up.”

“That's the nines problem. Right? That is, each nine is maybe linear, even sublinear problem. Right? In terms of rate of progress.”

“It seems to be pretty obvious at this point that the model's just getting smarter is not gonna By default, make adoption, like accelerate massively again. They need to become more useful.”

“I really think we're in the era of, like, of productization more than the latest model.”

“If you build a wrapper that depends on the models not getting better, you're gonna have a bad time, because the models will just get better. So if your thing is, oh, it can't do long context windows, or it can't answer in pages and pages and pages, well, the m…”

“And my argument is, like, more economic in the sense of, like, if you sell them, like, tens of billions of dollars of equipment, they can make hundreds of billions of dollars of AI value, or chips, with that equipment, whereas if you sell them AI model access,…”

“This is like a six month away thing.”

“Each model makes money, but the company loses money.”

“I mean, the models, you know, if you're in the path of models improving, you will get stomped like Godzilla, but there's still plenty of opportunities all over the ecosystem, especially if you're not doing something that's in software.”

“My fundamental view is that AI has been on an exponential for the last 10 years and as part of a sort of Moore's law for intelligence, and that we are now well advanced on that curve, and there are only a small number of years for AI models surpassing the cogn…”

“The story of the last year is that there's very, very, very, very little demand for number three. There's some demand for number two, and there's an exceptional amount of demand for the most performant model for a specific Specific task.”

“In 2024, China had just 9.3 billion invested in private AI companies. There's four hundred seventy billion cumulative since 2013, more than all other countries combined. The U.S. Produced 40 notable AI models in 2024 versus China's 15.”

“The simple math would be, it's one 60 on one test, but you've got to pass differentiated tests over a long period of time, so it's a thousand tests. So de facto, by the 50th step, it's zero IQ.”

“I mean, I, you can distill these things. Like distilling it, I think is probably much easier than training the thing in the first place.”

“The actual breaking news is that the White House is considering vetting AI models before they are released. Trump admin, which took a non-interventionist approach to AI, is now discussing imposing oversight on AI models before they are made publicly available.”

“The models feel like they're getting fatter every month and they're sort of eating away at the edges of what you can do with them.”

“I've completely dropped out of the whole, like, don't make mistakes, don't hallucinate. I feel like all that's been either baked into the pre-training or post-training, or it's like even in the system prompt already.”

“So if you have a model that can hack a system, it can usually patch that system for that hack in far less than six months, like probably in like an hour, like it can see, oh, there's a exploit here and we ship the code, merged it in, and now that exploit is cl…”

“And I think that talking about curing diseases and new medicines we're going to be able to develop, all of that starts to become much more real when you have models that are at this level of assistance and capability.”

“For most of these slowdown proposals, I personally have a hard time blackpilling about them, in the sense of, like, if the, if all of this gets implemented, how frustrated will I be? Like, the models are good. I would like better models. I want safe models, bu…”