Lab-in-the-Loop at Compute Speed — Ian Quigley, Leash Bio
October 14, 2026
AI Summary
5 min readIan Quigley, CEO of Leash Bio, describes a moment from his time at Recursion that crystallized a deep problem in computational drug discovery. A medicinal chemist walked by and identified the creator of a molecule just by glancing at it. That observation haunted him for years. When his team later trained a classifier to predict a chemist from a molecule alone, it achieved 60% top-five accuracy across 1,815 chemists. The model was learning sociology, not physics. This insight drives Leash’s entire approach: if you train on public data, you are likely learning the fingerprints of the people who made the molecules, not the underlying chemistry.
The Clever Hans Problem in Drug Discovery
The core problem is that public datasets for small molecule binding are full of hidden confounders. “When you train models on the exhaust, you might be in trouble,” Quigley says. Medicinal chemists are not trying to create unbiased training data; they are trying to make drugs efficiently. They tend to add functional groups to molecules that are already working, which makes molecular weight one of the single best predictors of activity in public datasets. They avoid certain synthetic routes for cost or feasibility reasons, and those choices leave systematic gaps. The result is a Clever Hans problem: models appear to learn chemistry but are actually reading the behavior and preferences of the ch
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Latent Space: The AI Engineer Podcast
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 Lab-in-the-Loop at Compute Speed — Ian Quigley, Leash Bio
- 2 (00:00) **The Clever Hans Problem in Chemistry** - Ian opens with a vivid anecdote about a medicinal chemist who could identify a molecule's creator just by looking at it, revealing a core problem with public training data.
- 3 (02:40) **What is a DNA-Encoded Library (DEL)?** - Ian explains the core technology that makes Leash's approach possible and why most people who work with DELs come out traumatized.
- 4 (05:00) **Why Sequencing is the Only Scalable Readout** - Ian argues that DNA sequencing is the only truly scalable measurement technology in life sciences.
- 5 (10:00) **How Leash Handles the DNA Barcode Problem** - Ian explains the clever barcoding strategy that distinguishes real binding from DNA-protein interactions.
- 6 (13:20) **Why DELs are Historically Traumatic and How Leash Broke Through** - Ian explains the traditional DEL workflow failures and Leash's radically different approach.
- 7 (16:20) **The Recursion Mafia** - Ian identifies four companies spun out of Recursion that all "rhyme" in their approach to noisy biology.
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
A few years ago I stumbled upon Leash Bio’s Substack, and was hooked. Be it their Kaggle competition that showed that no, we really haven’t solved protein-ligand binding, or their crazy discovery that knowing a molecule’s author is perhaps all you need to know to predict how good it is. Their substack — or at least Owl Posting’s deep dive on them — should be required reading for any scientist trying to do chemistry with AI.
So I was super excited to finally be able to sit down with Ian Quigley from Leash Bio, and pick his brain. Leash specializes in DNA Encoded Libraries (aka “DELs”), a way of collecting ridiculously high-throughput interaction data between proteins and small molecules. Before I discovered Leash, I had heard nothing but horror stories from DEL users. And yet, Ian and friends somehow found a way to tame the beast, extracting clean and reproducible data that is just perfect for large scale modeling. Without teasing too much, the answer comes down to epistemic rigor, and an incredible conviction that you just have to Do The Thing.
It gets even better! Near the end of this episode, Ian teases something his team was just playing with when we recorded: an AI playground for agents to do verifiable scientific reasoning. They discovered the data they had spent the last few years collecting could provide tens of thousands of “lab-in-the-loop” simulations, just what an agent needs for scientific RL. Two months later, it’s matured into Minos, and today they’re releasing a full preprint, blog, and open dataset. We’ll be releasing a lightning podcast in a few days with an update. Stay tuned!
We had a blast with Ian, probably one of our most fun episodes to date. For both complete chemistry novices and expert AI x Bio practitioners, we think you’ll get something out of this.
Enjoy!
We covered:
* Why DELs traumatize everyone who touches them, and how Leash puts dozens of barcodes on each molecule to catch proteins that bind the DNA instead of the drug.
* Ian’s rule that biology