The a16z Show Faster Science, Better Drugs
252 snips
Sep 15, 2025 
In this conversation, Patrick Hsu, co-founder of the Arc Institute, and Jorge Conde, general partner at a16z, explore how to accelerate scientific progress through virtual cell technology. They delve into the challenges hindering research and the potential impact of AI on drug discovery. The duo discusses what a transformative moment in cell biology might look like and highlights the balance between hype and substance in biotechnology. They also examine the financial implications of breakthrough medications and the evolving landscape of biopharma.
AI Snips
Chapters
Books
Transcript
Episode notes
Design Models To Suggest Testable Perturbations
- Scope virtual cells around perturbation prediction: map cell states and predict interventions to move cells between states.
- Use models as wet-lab copilots to propose testable experimental conditions rather than only ML benchmarks.
Treat Biology Models Like GPT Generations
- Hsu compares biology model stages to GPT progression and places current capabilities between early GPT generations.
- He stresses full-stack work: curate public data, generate private data, build benchmarks, and iterate models.
Bench With Textbook Biological Tasks
- Validate virtual cell models with canonical biological discoveries and drug mechanism recapitulation.
- Use textbook-known tasks (e.g., Yamanaka factors, differentiation factors) as concrete model evaluations.
