# Latent Space: 🔬Biology Is Turning Into Software — Matt McPartlon and Neil Patil, Chai Discovery — retained record

> Retained youtube record from youtube:Latent Space.

Canonical: https://highsignal.app/data/records/c3c1c1d9cd6189ae

[← sources](https://highsignal.app/data)


source: [youtube](https://highsignal.app/data/youtube) (youtube:Latent Space)  published 2026-08-11 20:48 ingested 2026-08-12 07:12

[original source ↗](https://www.youtube.com/watch?v=Qp5xklyJySI)

## retained text



```
Chai Discovery co-founder Matt McPartlon and product lead Neil Patil explain why pharma abruptly started buying AI design tools this year instead of forcing every AI-bio company to go build its own pipeline — and what specifically changed in the models to make drug design teams trust them. We trace the lineage from Chai-1, which they open-sourced for reasons that had almost nothing to do with the model itself, through the Chai-2 campaign that convinced Lilly, Pfizer, Novartis, and argenx to sign. Along the way: why their CEO says the company's real competitor is a mouse; the cryo-EM result so accurate the team's first reaction was to suspect the lab of sending back their own file; why the product looks more like SolidWorks than ChatGPT, and why Neil expects to throw most of it away; what happened when a pharma scientist saw results on a target she'd spent a decade on; why the compute market is quietly mispriced for this class of model; and the one bottleneck both guests would remove by fiat — which is not data, not compute, and not the models.

00:00:00 — What the product actually looks like
00:01:15 — Intros: Matt and Neil
00:03:22 — Four pharma partnerships
00:05:00 — The software-layer bet
00:07:00 — The 50-target challenge
00:08:12 — What is an antibody, and why target it
00:11:22 — ADCs, bispecifics, and pressing the switch
00:12:39 — How this was done before: mice and yeast display
00:14:29 — Selectivity and cross-reactivity
00:17:26 — Chai-1 and the MSA detour
00:19:41 — Sitting in the OpenAI offices
00:23:09 — Anatomy of a structure prediction model
00:25:49 — Chai-2: all-atom diffusion and crossing into design
00:28:21 — Co-designing sequence and structure
00:29:48 — How do you know a design is any good?
00:32:36 — Validation, cryo-EM, and the 0.33 Å result
00:35:26 — Chai-2.5, Chai-3, and betting on the models
00:39:18 — Building the product under pharma IP constraints
00:42:32 — Convincing scientists who hate AI tools
00:44:58 — The ten-year target
00:45:50 — Battle-testing without your own pipeline
00:47:40 — Waterfall to loop
00:49:04 — Levels of abstraction, and throwing the product away
00:52:06 — Epitope prediction: the harder problem
00:55:29 — The economics question
00:57:29 — Modalities you can't get from immunization
00:59:48 — What's really blocking science-to-engineering
01:01:51 — Buying compute is a terrible job
01:04:18 — Triangle attention and why GPUs hate it
01:05:28 — Durable execution and Temporal
01:08:27 — Complexity vs. the bitter lesson
01:09:37 — Inductive bias and data efficiency
01:15:13 — Will protein design get commoditized?
01:19:13 — No pipeline, no data moat?
01:21:13 — Per-partner fine-tuning
01:22:36 — Are all AI companies consulting companies?
01:23:42 — The highest-value token in any domain
01:24:20 — Pharma as VC: Genentech and Eroom's law
01:26:31 — Everyone is a capital allocator
01:29:23 — One bottleneck removed by fiat
01:32:07 — Takeaways
```



document fetched 2026-08-17 06:35 · canonical https://youtube.com/watch?v=Qp5xklyJySI
