Latent Labs’ Simon Kohl Is Rewriting the Code of Biology With Generative A.I.

Simon Kohl, recognized on this year’s A.I. Power Index, stands at the leading edge of a scientific transformation: the fusion of A.I. and biology. A co-developer of AlphaFold2—the Nobel Prize-winning breakthrough that cracked one of biology’s grand challenges—Kohl has now turned his focus from understanding life’s molecular machinery to authoring it. As co-founder and CEO of Latent Labs, he’s advancing a vision where biology becomes programmable, and drugs can be designed with the precision and speed of semiconductor engineering. Kohl’s platform, LatentX, achieves laboratory hit rates of 91 to 100 percent for macrocycles—an astonishing leap compared to the sub-one-percent success rates of traditional methods. Rather than just predicting what nature has created, the system generates what nature could create, simultaneously designing molecular sequences and 3D structures in real time. Backed by investors including Google’s Jeff Dean and Cohere’s Aidan Gomez, Latent Labs is applying these capabilities to areas where conventional discovery has long faltered, like oncology, autoimmune diseases and rare genetic disorders.

The promise of generative A.I. in biology is matched by its complexity and responsibility. Kohl is pushing back on the assumption that A.I. will make biology “easy,” and argues that the race to create novel biological systems demands new frameworks for safety and governance. From DeepMind’s London lab to Latent Labs’ San Francisco wet lab, Kohl’s trajectory traces the next frontier in scientific discovery: where the boundary between computation and creation is rapidly dissolving.

What’s one assumption about A.I. that you think is dead wrong?

That A.I. will make biology ‘easy’ overnight. Having co-developed AlphaFold2, I’ve seen firsthand how A.I. can solve incredibly complex problems like protein folding. But the assumption that this means we can computationally get perfect drugs at this moment is wrong. Biology remains fundamentally messy. A.I. currently amplifies our capabilities—at Latent Labs, we’re making biology programmable—but it still requires deep scientific intuition to ask the right questions and interpret what the models are telling us.

If you had to pick one moment in the last year when you thought “Oh shit, this changes everything” about A.I., what was it?

It wasn’t a single model release over the last few years, but rather when I realized we could move beyond just predicting biological structures to actually designing them from scratch. That’s why I left DeepMind at the end of 2022 to start Latent Labs—I saw we were at an inflection point where generative A.I. could make biology truly programmable. We’re not just understanding nature anymore, we’re becoming capable of authoring it with precision.

What’s something about A.I. development that keeps you up at night that most people aren’t talking about?

The widening gap between our ability to design biological systems and our ability to predict their broader consequences. We can now generate novel proteins and biological circuits with unprecedented precision, but biological systems are interconnected in ways we’re only beginning to understand. As we give researchers and companies these powerful generative tools, we need to develop equally sophisticated frameworks for testing safety, understanding off-target effects and ensuring we’re not creating biological complexity we can’t control.

You co-led DeepMind’s protein design team on the Nobel Prize-winning AlphaFold2 project, and now LatentX goes beyond structure prediction to actually design entirely new proteins. What technical breakthroughs enabled this leap from predicting existing structures to creating novel ones, and how does this change the timeline for drug discovery?

The breakthrough was moving from predicting what nature has created to generating what it could create but hasn’t. AlphaFold2 understood existing structures, but Latent-X co-samples sequence and structure simultaneously—designing both molecular sequence and 3D shape in real-time while following atomic-level rules. We’re authoring biology, not just predicting it. The impact is dramatic: 91 percent to 100 percent laboratory hit rates versus traditional methods below one percent. Scientists achieve in 30 candidates what previously required testing millions, turning months of experiments into seconds of computation.

Your web-based LatentX platform allows researchers to design proteins directly in their browser, making this cutting-edge capability accessible to academic institutions and biotech startups. How are you balancing the need to democratize this technology with ensuring it’s used safely and responsibly, especially given the potential dual-use implications?

We envision a future where effective therapeutics can be designed entirely in a computer, much like how space missions or semiconductors are designed today. Our platform empowers scientists with lab-validated protein binder design at their fingertips, whether they’re experts or new to A.I.-powered drug design. In democratizing access to our breakthrough science, we take dual-use implications seriously—actively participating in biosafety discussions with regulators and restricting access per international sanctions lists. Our integrated approach, validating everything in our San Francisco wet lab, means we understand real-world implications, not just computational possibilities. We prove value first while maintaining robust safeguards.

You’ve achieved state-of-the-art results in lab testing for protein binding and recently raised €47.9 million with backing from notable A.I. leaders like Jeff Dean and Aidan Gomez. What specific therapeutic areas are you targeting first, and how do you see competition evolving as more companies enter the AI-driven protein design space?

Our models are general in nature and are able to generate macrocycles, mini-binders and antibody formats from scratch. We’re keen on applications in oncology, autoimmune diseases and rare genetic disorders where traditional discovery struggles. Macrocycles are exciting—combining biologics’ specificity with small molecules’ oral deliverability. In head-to-head lab comparisons, we’ve exceeded prior work from large tech companies and academic labs. Our advantage is integrating our world-leading expertise from experience in building AlphaFold with wet lab validation and enterprise-grade platform engineering. With the biologics market growing to over £1 trillion by 2033, success depends on delivering lab-validated results, with scalable engineering that satisfies the security requirements of the industry.