Brandon (Host)

speaker

390 appearances 3 recordings 1 series first heard Jun 2026 last heard 16 Jul

Brandon (Host)’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 3 in all, peaking in Jul 2026 with 2.

Appearances

newest first · ▶ plays the moment

Latent Space: The AI Engineer Podcast · 🔬 The Coolest Diffusion Research Isn't in LLMs — Evan Feinberg & Sergey Edunov, Genesis Molecular AI · 1 Jul 2026 podcast

But at least open source, it's...

So that's something that Pearl, for example, could actually model and expect to see some performance here.

Yeah, we've been doing...

So one thing I want to ask about

I kind of keep coming at this.

So, you know, you talked about there's a lot of synthetic data and then there's, you know, there's ML modeling, which is very kind of modern general model flavor.

I want to talk about kind of the feedback between how your computation, your ML and how wet lab data went into developing this model.

And, uh,

You know, we talked about priors.

I think maybe one, it's been typically hard to scale protein ligand models well in a way which meaningfully generalizes.

And there, you know, this recent Pearl paper and some of your other recent results we'll talk about shortly have shown true generalization.

And what I'm kind of curious about is how do these different aspects feed together

to give you this power.

And specifically, just saying computation is interesting, but you have to be very careful about computation.

MD has all sorts of biases.

If you're not right, it can give you poor results.

And so I'm curious about how all of this worked together, and especially in the history of the development of the company, how did you kind of get to this point?

I find it interesting to hear that you have now started bringing in agents into your development process.

You know, it sounds like from the very get-go, your philosophy was much more, these are tools for scientists, and you make the best possible tool, and this lets them optimize whatever pipeline they're going to use.

But ultimately, this is a tool for MedChem.

Showing 261–280 of 390 · page 14 of 20