Jul 3, 2025 · 49m · no-priors
No Priors Ep. 121 | With Chai Discovery Co-Founders Jack Dent and Joshua Meier
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of No Priors, Chai Discovery co-founders Joshua Meier and Jack Dent discuss their generative model Chai-2, explaining how zero-shot antibody design and biological CAD software are transforming drug discovery into an engineering discipline.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 14.7% of the talking time here. How this is scored →
The hosts as informed peer 5.5 Guest teaching 4.8 Guest disagreement 0.3 The hosts pushing back 0.5
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 20:47 Joshua rejects standard pharma objections
Joshua directly addresses and counters common industry pushback that computational tools merely accelerate what existing labs already do without expanding the target space.
Hardest push from the hosts ▶ 30:38 Sarah challenges the discovery-centric value propositionSarah counters that industry skeptics view discovery costs as negligible compared to massive downstream clinical failure rates.
Biggest teaching moment ▶ 7:34 Jack explains physical screening vs zero-shot generationJack walks through the labor-intensive legacy pipeline of harvesting llama/mouse plasma or panning yeast libraries to illustrate the scale of Chai-2's hundred-fold hit rate improvement.
The host holds their own ▶ 15:52 Sarah maps LLM emergent properties to molecular modelingSarah displays deep machine learning fluency by pressing on whether 3D generative diffusion models develop internal grammar and conceptual abstractions similar to large language models.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| The Timing and Vision Behind Founding Chai Discovery | 5 | 4 | 0 | 1 | Sarah sets a strong historical frame regarding a decade of unfulfilled promises in AI drug discovery and asks why the timing makes sense now. Joshua explains the transition from research toy to real capability powered by diffusion models and structure prediction. | |
| Chai-2 Breakthrough in Zero-Shot Antibody Discovery | 4 | 6 | 1 | 0 | Sarah asks for a lay breakdown of the Chai-2 model results. Jack details how traditional high-throughput screening or animal immunization operates compared to their zero-shot computational pipeline. | |
| Benchmarking Across Diverse Targets and Problem Formulation | 6 | 5 | 0 | 1 | Sarah presses on the benchmark design methodology across 52 targets and epitope constraints. Joshua describes their vendor-scraping and sequence-holdout approach to treat evaluation strictly as an engineering benchmark. | |
| Model Intuition: From Structure Prediction to Generative Design | 7 | 5 | 0 | 0 | Sarah draws technical analogies to LLM emergent capabilities and vocabulary emergence. Joshua and Jack discuss how the model learns deep physical interaction principles beyond sequence similarity. | |
| Platform Access, Scientific Reception, and Wet-Lab Integration | 5 | 5 | 1 | 1 | Sarah asks about industry skepticism regarding AI antibody design and the remaining role for wet-lab screening. Joshua and Jack explain how wet labs will transition from screening to high-volume verification. | |
| Future Biotech Paradigm and Complex Molecular Formats | 6 | 5 | 0 | 1 | Sarah highlights the conventional bear case that clinical development rather than discovery is the true cost bottleneck. Jack and Joshua illustrate how upfront co-design for monkey/human cross-reactivity de-risks downstream clinical hurdles. | |
| Building the CAD Software Suite for Molecular Biology | 5 | 4 | 0 | 0 | Sarah probes the strategic defensibility of molecular AI platforms and asks how domain biologists must adapt. Joshua and Jack explain why sophisticated workflow software and multi-property prompting form a defensible moat. | |
| Engineering Rigor, Team Culture, and Company Scaling | 6 | 4 | 0 | 0 | Sarah recalls early debates in her office about developer containers and platform investments for deep learning. Jack shares concrete war stories about bisecting git commits to find expensive silent training regressions. |
Statements from this episode (19)
Meier: Previous AI bio startups had overly tight lab integration
“Almost every AI bio company before us has had some kind of very tight lab integration with what they are doing. And it almost too tight.”
Dent: Humans will engineer molecules with atomic precision within a few years
“Over the next few years, we are going to have the ability as a human race to engineer molecules with atomic precision.”
Dent: Chai-2 hits nearly 20% wet-lab antibody binding success
“We have about a two week validation cycle in the lab, and two weeks later, we see that roughly close to 20% of these antibodies actually bind their targets in the intended way.”
Dent: Chai Discovery originally targeted only a 1% antibody success rate
“We were actually only targeting a success rate of one percent. That was the company-wide goal for the entire year.”
Dent: AI will unlock new molecule classes and cure previously incurable diseases
“The idea that in the next five, 10 years that there are going to be. Entire new class of molecules that we're going to be able to discover and entire new targets that we're going to unlock and time markets that we can open up and therapeutics that we can get t…”
Meier: Most AI drug discovery papers test only one to three targets
“Most of the existing papers in this area of doing AI for drug discovery are usually looking at like one, two or three targets.”
Chai filtered benchmark targets to under 70% identity from training data
“So we actually wrote a scraper that would go and see what was in stock. We would go and pick out the protein. We would go look at what that protein sequence was. Now we need to make sure this is held out from training as well. Right? So we would take that, tha…”
Dent: Structure prediction is biotech's ImageNet moment, molecular design is Midjourney
“One piece of intuition around that is that you can almost think about structure prediction as The ImageNet moment for the field, where with structure prediction, we are asking a model to go from sequence to a predicted structure, and it's sort of like a classi…”
Dent: Chai's models achieve structure error below width of one atom
“They reason quite literally by placing individual atoms in, in three D space. And often they're getting the resolution of these structures, the error down to less than the width of one atom. When we look at the error across the entire structure.”
Meier: Chai-2 maintains success rate on targets with 25% sequence similarity
“We actually even have a slide in our paper in the supplement where we actually look at an even harder subset. So not looking at things that are, you know, up to a sequence similarity with the model, but actually pushing all the way down to 25%. So really looki…”
Meier: Chai-2 achieved a 50% success rate on tested targets
“The model is not perfect. You know, it worked in 50% of the targets that we tried. Maybe it would have been more, right, for the caveats we talked about before, but, you know, it worked in 50% of cases.”
Dent: Chai achieved 70% success rate on mini protein binder designs
“I think if you see our mini protein results, we are, I think, close to 70% on those with picomolar affinities, like really, really tight binders for every single target that we tested. So all five targets we tested worked, and 70% of the designs that we ordere…”
Dent: A CAD software suite for biology will be created
“And I think once you have that, you really enter this era where you sort of have a computer aided design suite for molecules in a way that, you know, we have maybe solid works for mechanical engineering, or we have Photoshop for create creatives and that, that…”
Meier: Clinical interest in traditional monoclonal antibodies will decline
“So what we predict will happen is people probably won't be as interested in, in the clinic for things like monoclonal antibodies. These are, you know, antibodies that are hitting, for example, like a specific epitope on a protein.”
Dent: Chai generated cross-species binding antibodies testing only 14 sequences
“We ordered actually only 14 sequences to the lab. And I think four of those were histohumans. One of those was a hit to the Sino. One of them was actually overlapped and hit both. That one now allows us to move forward with that program and gives us a whole am…”
Dent: Models will generate entire drug candidates in just 20 attempts
“The fact that we can get antibody hits in just 20 attempts, there's no reason that, that we couldn't generate intra drug candidates in that same number of attempts.”
Dent: Scaling software teams stall without a dedicated architectural steward
“I think you just learn that unless somebody is really taking care to keep the entire system in their head and is an effective technical steward of the architecture, that things just evolve and the sort of the entropy of the software takes over and slows down y…”
Dent: Chai spent tens of thousands bisecting Git history for one bug
“We've literally had to do this in Chai's history, but we've had to go and bisect Git history. Run launch training runs, you know, with a sort of a binary search to identify a small enough range of pull requests to identify a bug, then go to that, that, that pu…”