00:23:35 · Josh Meier
[00:23:35] S0: So the first step is actually getting into the tools of the drug hunters who are making the molecules themselves. So at Chai, we do a ton of work not just on the model development, but also on the product and being able to ship these models, to the world. It's one thing to just throw a model over the wall and hope that people can figure it out. It's another to build a really intuitive product experience that allows people to, like, iteratively work on these projects on the computer. In a computer aided design suite, it's, for, for molecules, it's, I know it's zero shot, right, for generating the molecules, but you need be very thoughtful about what am I zero shotting, how do I actually collect all those inputs, how do I look at the results and decide what to do next, So we put a lot of care into thinking through what does that product look like, and I think it's a really key part of the company and what we're doing here. So that's the first step, how do we actually get this into people's hands so we can use it on the right projects? And then the next is, how do organizations, actually test these models, start to trust them, and then use it for all these projects and change the way that they're working? Because if we can get these tools into people's hands, the processes around them need to change. We need to change how we discover targets. We need to change then how we manufacture all these materials that are coming out. So it actually requires a lot of thinking by these companies, and that's why we're actually investing in those kinds of deep relationships at the very senior level, with our customers to help them think through how to even adopt this AI transformation. And I think that's also one of the most exciting parts of my job, honestly, is like getting to meet these folks who are dedicating their lives to creating these medicines and then helping them get access to this technology that can allow them to do their job at just the next level.
00:25:06 · Sachin
[00:25:06] S1: How do you think about how Chai continues to scale its impact on more of a commercial level? Like, obviously, you guys have now raised a lot of money and you've needed it to fund compute and a lot of these kind of results you've seen, but how do you kind of go from more of a research organization to a commercial organization, which I'm sure you've seen Sam and the OpenAI team do?
00:25:24 · Josh Meier
[00:25:24] S0: For a while, we were really an R and D company, right? We were investing in creating, the best models possible, and we actually didn't commercialize, we open sourced what we were doing, in part because we didn't think it was valuable enough for people to pay for it, and then we tried to cross that inflection point where the models became useful for pharma, right? And at that point, we really started to become more of a commercial organization. We started investing in the product, the business development in order to make that happen. So when we think about ourselves as an enterprise organization or actually a commercial organization, it's all about how do we get this into the most impactful projects possible in the industry. So it's actually quite natural, and one of the things I love about our business model is, as our technology gets better, we unlock more projects, we create more value, which means people want to buy more Chai for their projects, and as our product becomes better, as our technology becomes better, as our validation becomes better, the business just does better as a result too, and that allows us to continue reinvesting in this. And you see a similar thing with the big AI labs, right, like OpenAI, when they come out with better models, people buy more tokens, and that gives you the money you need to then go and build the next generation of the models. And it's it's very exciting to work in a company that has that flywheel because it means that, as we get better on everything we're doing here, it really does translate into more commercial success, and ultimately that's what the company needs to do.
00:26:42 · Sachin, Josh Meier, Adam Miller
[00:26:42] S1: Because it's such a technical area, have you thought about the business model of, like, doing the work for the pharma companies versus giving them the tools to do the work? [00:26:51] S0: It comes down to how to create the most impact possible. And our customers are quite good at what they do. They know the biology really well, they know the targets really well, and in order to create the most impact, we wanna create the best product experience so that these models can actually be leveraged, for for those projects. So it's not even an either or, to be honest. Yeah. It's not do we help them with their project or do we build the tools for it? It just starts from like, how do you solve the projects in the most effective way possible? And then you end up, you know, kind of pushing on both levers. [00:27:21] S2: Commercially, one of your big partnerships is with Eli Lilly, and from what I understood a little bit today is that you will sort of discover these drugs and they help try and develop them. Can you talk a little bit more about what that partnership looks like and what the feedback loops also look like when working with them?
00:27:21 · Adam Miller, Josh Meier
[00:27:21] S2: Commercially, one of your big partnerships is with Eli Lilly, and from what I understood a little bit today is that you will sort of discover these drugs and they help try and develop them. Can you talk a little bit more about what that partnership looks like and what the feedback loops also look like when working with them? [00:27:35] S0: It's been fantastic working with the Lilly team, right? They were really impressive in terms of how quickly they adopted this technology and then how rigorously they exported it as well. They ran the right experiments to understand whether the models were working and then started to now adopt it at scale within the organization. So it's a great pleasure to work with one of the biggest pharma companies in the world, right, and that really being a great avenue to just create, massive impact with the models that we are building here, you just look at the number of projects they can now go apply it to. [00:28:02] S2: And in terms of those feedback loops, do you hear from them about the results that they're testing with, and you're able to use that data to reinforce your models?
00:28:02 · Adam Miller, Josh Meier, Sachin
[00:28:02] S2: And in terms of those feedback loops, do you hear from them about the results that they're testing with, and you're able to use that data to reinforce your models? [00:28:10] S0: Yeah. We talk to the team all the time. They're really engaged. They've been great partners to us in that too. We're, at a point in this technology curve where there's still a lot of progress being made, so to be able to have a have a company like Lilly who is stress testing our models at scale, applying it to all these projects, telling us this is what works well, this is what doesn't work well. That is so helpful for setting our own research roadmap, and then ultimately that means that we can be a very customer oriented company with what we do. When you're working on big research moonshots, it's very easy to end up working on the wrong thing. You're not getting feedback, maybe you have like a benchmark that you're pushing, but is it really the right benchmark? So when you have a company like Lilly with such broad scope and how they can use our technology, it really helps us learn about what are the right things to be working on, and I think ultimately that's key to our success. [00:28:54] S1: Josh, we kind of flew here to kind of ask this question about like, when can we expect AI produced or designed drugs to be on the market? [00:29:02] S0: And it feels like that's [00:29:03] S1: kind of still a few years off. But what is the case in which this doesn't work? Is it that scaling laws don't continue and model progress kind of drops [00:29:12] S2: off?