Barzilay, Collins Win Pritzker Prize for AI-Enabled...

Why this matters

Barzilay and Collins have spent nearly a decade building the case that machine learning models can do real chemistry work, not just predict it in a vacuum. Their lab's track record, halicin in 2020 and abaucin in 2023, already showed deep learning flagging antibacterial candidates that human researchers hadn't prioritized. The Pritzker committee rewarding that approach, barely two years after the prize itself was created in 2024, tells us the field is done treating "AI drug discovery" as a slide-deck pitch and started treating it as a lab discipline with results that hold up under experimental validation.

For researchers, the lesson is specific: the win went to the combination of deep learning, bioengineering, and wet-lab testing, not to a model alone. Founders pitching AI-for-biology should take note that credibility now gets built on validated candidates against drug-resistant pathogens, not benchmark scores. Antibiotic resistance is a slow-moving crisis with few new drug classes in decades.

If AI-guided discovery keeps producing candidates that survive contact with actual bacteria, that's the metric worth watching, not the headlines.