Transcribed by Andrew, 6 November 2026. The wording below is verbatim; the only change is that the single block of text has been divided into paragraphs for readability.
Did you know, for basically all of history, we made drugs by guessing, but now AI has changed all of that because it's learned how to design new molecules. So let me tell you about the problem that it's solving.
Almost every drug in your medicine cabinet works by binding a specific protein in your body. Painkillers bind to pain receptors, cancer drugs bind to tumor markers, and antibiotics bind to bacterial proteins. And to make a new drug, you basically had to find a molecule that binds to your target, and you did this by testing millions and millions of candidates, mostly by trial and error. It cost billions of dollars and it would take decades just to find one drug.
So for years, AI has become better at predicting which molecules might bind. So companies like DeepMind's AlphaFold cracked the shape of proteins, and that was huge. But you still had to test enormous libraries to find the one that actually worked.
Now labs like Chroma, Proficient, and Evolutionary Scale have done something bigger. Instead of searching for a binder, they actually design one from scratch using generative AI models trained on the language of proteins, the same way that GPT was trained on the language of English. You give the AI a target, a specific spot on a specific protein that you want to attack, and it generates a brand new protein specifically designed to bind there. This is the one-shot milestone that the drug discovery world has been chasing for decades because it changes the entire economics of medicine. Instead of scanning millions of different options, you generate the exact molecule that you need.
So comment below what you think about this and follow for more AI, math, computer science, and tech breakdowns.