What's this about?
Sequoia hosts the Chai Discovery team arguing that drug design is "another scaling problem" — that molecular design, like language and vision before it, yields to more data and compute rather than clever hand-crafted priors (Aug 4).
What exactly is "the bitter lesson"?
Rich Sutton's famous observation that, over AI's history, general methods that leverage more computation reliably beat human-engineered domain knowledge. Applying it to drug design means betting on scaled learning over biochemists' hand-built rules.
Why does it matter?
It's the scaling-hypothesis argument extended into biotech — the same logic behind this week's math and agent results, now aimed at molecules. Worth a listen for whether the bitter lesson actually holds in a domain with far less data than text.