00:12:59 · Adam Miller, Josh Meier
[00:12:59] S2: Focusing on the product for a moment, it's been described by some as CAD for molecules, so when I actually sit down and look at the software, like, what am I actually seeing and how does it work? [00:13:09] S0: Yeah, well, like to think of the protein folding models. If you're looking at the structures on the product, it's almost like an atomic level microscope. Protein folding has advanced to such a point now where, we can now predict how antibodies and antigens interact with a very high degree of accuracy. If you look back just a couple of years ago, you know, you look at the state of the art AlphaFold, right, a few years back, it could only predict these antibody antigen complexes correctly about like 3% of the time, and then now with our latest models, it's improved by an order of magnitude, right? So we're now in a space where we can really see what these molecules look like on the computer, and that really changes the way that we design things because when you see something on the computer, there's a very good chance that it looks like that in the real world. Like, one of the craziest things we saw here is if you take those structures that you're seeing on the product, and then we went into the lab, we used a technique called cryo electron microscopy. We literally used electron microscopes to see at the atomic level what's going on here. The error between what we saw in the lab and what we saw on the computer, was... The error was was less than the diameter of an atom. So it's it's crazy how accurate these AI systems have gotten in biology.
00:13:09 · Josh Meier, Adam Miller
[00:13:09] S0: Yeah, well, like to think of the protein folding models. If you're looking at the structures on the product, it's almost like an atomic level microscope. Protein folding has advanced to such a point now where, we can now predict how antibodies and antigens interact with a very high degree of accuracy. If you look back just a couple of years ago, you know, you look at the state of the art AlphaFold, right, a few years back, it could only predict these antibody antigen complexes correctly about like 3% of the time, and then now with our latest models, it's improved by an order of magnitude, right? So we're now in a space where we can really see what these molecules look like on the computer, and that really changes the way that we design things because when you see something on the computer, there's a very good chance that it looks like that in the real world. Like, one of the craziest things we saw here is if you take those structures that you're seeing on the product, and then we went into the lab, we used a technique called cryo electron microscopy. We literally used electron microscopes to see at the atomic level what's going on here. The error between what we saw in the lab and what we saw on the computer, was... The error was was less than the diameter of an atom. So it's it's crazy how accurate these AI systems have gotten in biology. [00:14:17] S2: And also from the product perspective, in terms of what the actual outcome is, would you describe it as it collapsed the amount of time it's taking to discover and design drugs? Is that the fundamental thing?
00:14:17 · Adam Miller, Josh Meier
[00:14:17] S2: And also from the product perspective, in terms of what the actual outcome is, would you describe it as it collapsed the amount of time it's taking to discover and design drugs? Is that the fundamental thing? [00:14:28] S0: I mean, that's for sure. I think it's, I can't think of anything else in AI that takes a process that might have taken months or years, you know, discovering an initial molecule, and can now do it with twenty four hours of compute and then, you know, a couple of weeks of testing in the lab, like that's just incredible just to think about like what that compression of time looks like. But actually time is not the thing I'm most excited about, it's actually the ability to make better medicines. A lot of the field right now, is, again, drug design is really hard. There's a lot of projects that are quite incremental that are out there, and I think that by making it easier to design these complex molecules, hopefully we can raise the bar for what people are working on, and then allow patients to really get the treatments that they need.
00:15:08 · Adam Miller
[00:15:08] S2: Something that stayed with me from this interview is just how much of Chai's story comes back to trust. Drug hunters won't use tools they don't trust. Pharma companies rely so much on models, and that's why security needs to be airtight. Chai's customers' projects are, in their own words, top secret, and that's why they need to invest so much in security. And that is a big reason why they've been able to partner with companies like Eli Lilly. And that trust equation isn't just unique to biotech it's something that every company needs to think about when they start to engage with serious customers. It's not a question about is your product actually good? It's a question of do they trust you? We've heard of deals that have stalled for months because things like SOC two and ISO 27,001 weren't installed. These problems are exactly what our sponsor Vanta was made for: continuous monitoring across every framework, dashboards set up for you, and what used to take months now takes weeks. This is what lets ambitious teams focus on their product while enterprise trust gets built in the background. If you're building something ambitious and want to unlock real enterprise relationships, head to Vanta and get $1,000 off from us. Head to vanta.com/sachaninandable.
00:16:22 · Sachin
[00:16:22] S1: Speaking about the AI landscape for a second, I think people are recognising the limits of LMs, you know, text based learning about the world, and now there's a whole kind of movement around working world models and trying to understand the world without just text. How have you guys kind of cracked this? Like biology is obviously such a hard problem to solve and probably very different to what you're working on at OpenAI. Can you give us a kind of look inside your kind of thinking process and how you've kind of gone around the complexities of the real world?
00:16:51 · Josh Meier, Adam Miller
[00:16:51] S0: Yeah. I think a lot of it comes down to some kind of scaling law at the end of the day. Like, there is a process here that at its core seems like it can be learned. Something like protein folding, for example, predicting the atoms in three d space, like, given a sequence, you know, if you put either learn the problem better, like use better algorithms, put more compute so the model can process better, or more data, these are levers that tend to, you know, make progress on the problem. So if we believe that these things are predictable, then it's just a matter of getting those ingredients into a place to solve the problem, and I think that's ultimately what keeps us going, you know, through the hard challenges. Cause you asked me about doubts before, right? Like, you know, it's always like, this has gotten so good, can we make it better? Right? And I think we're just convinced that you can almost always make it better, you can keep going through that scaling law, and it's not always going be easy, there's going be a lot of work to do. People often ask us like, from Chai one to Chai two to Chai three, like, how did you get these big improvements? Like, what what was the secret from one model to the next? And I wish it was as easy as one secret that we had to find, but it's really compounding, returns. It's, you know, every week or every month, you know, making five to 10% improvements that you stack those together up for a long time, and you end up in a very different place than where [00:17:59] S2: you stand.