Dylan Patel

And when you zoom in really close, it can be really obvious that this model is just really bad at this thing, and we can fix it, and you just add these up. So some of it feels like luck, but on the ground, especially with these new reasoning models we're talking to, it's just... so many ways that we can poke around. And normally it's that some of them give big improvements.

Before expert controls started.

Before expert controls started.

Before expert controls started.

Total AGI vibes. They're like, we need to do this. We need to make a new ecosystem of open AI. We need China to lead on this sort of ecosystem because historically the Western countries have led on software ecosystems. And he straight up acknowledges, like, in order to do this, we need to do something different. DeepSeek is his way of doing this.

Total AGI vibes. They're like, we need to do this. We need to make a new ecosystem of open AI. We need China to lead on this sort of ecosystem because historically the Western countries have led on software ecosystems. And he straight up acknowledges, like, in order to do this, we need to do something different. DeepSeek is his way of doing this.

Total AGI vibes. They're like, we need to do this. We need to make a new ecosystem of open AI. We need China to lead on this sort of ecosystem because historically the Western countries have led on software ecosystems. And he straight up acknowledges, like, in order to do this, we need to do something different. DeepSeek is his way of doing this.

Some of the translated interviews with him are fantastic.

Some of the translated interviews with him are fantastic.

Some of the translated interviews with him are fantastic.

There hasn't been one yet, but I would try it.

There hasn't been one yet, but I would try it.

There hasn't been one yet, but I would try it.

Very direct quotes, like, we will not switch to closed source when asked about this stuff. Very long-term motivated and... how the ecosystem of AI should work. And I think from a Chinese perspective, he wants a Chinese company to build this vision.

Very direct quotes, like, we will not switch to closed source when asked about this stuff. Very long-term motivated and... how the ecosystem of AI should work. And I think from a Chinese perspective, he wants a Chinese company to build this vision.

Very direct quotes, like, we will not switch to closed source when asked about this stuff. Very long-term motivated and... how the ecosystem of AI should work. And I think from a Chinese perspective, he wants a Chinese company to build this vision.

Accepted practice is that for any given model that is a notable advancement, you're going to do two to four X compute of the full training run in experiments alone.

Accepted practice is that for any given model that is a notable advancement, you're going to do two to four X compute of the full training run in experiments alone.

Accepted practice is that for any given model that is a notable advancement, you're going to do two to four X compute of the full training run in experiments alone.

Research gets you O1. Research gets you breakthroughs, and you need to bet on it.