AlphaFold Predicted 200M Protein Structures in Two Years, Ending 50 Years of Manual Work
r0ck3t23 · x · 2026-11-03
A reflection on protein structure determination: a protein's shape decides its function, and solving shapes enables understanding disease and designing drugs. By 2020, after 50 years of lab work, humanity had solved only 170,000 structures — some scientists spent entire PhDs on a single protein.
Key points:
- In 2020, DeepMind's AlphaFold began predicting structures in minutes with near-lab accuracy
- Two years later it released predicted structures for 200M+ proteins — nearly every known protein — free for everyone
- The work won the 2024 Nobel Prize in Chemistry; millions of researchers in 190+ countries now use it
The author's core point: the scientists who solved one shape at a time weren't slow — every hand-solved structure became training data for AlphaFold. The people who did the hardest version of the work made the easier version possible, a pattern about to repeat everywhere.
More from AGI Musings
- AI researcher: drop the AGI longevity fantasy — humans may hold volition for ~13 more years — danfaggella · 2026-11-03
- Hoffman's theory goes viral again: evolution built consensus reality, not truth-seeing minds — GabGarrett · 2026-11-03
- AI safety founders publish personal cyber-defense and preparedness guide for the AI age — harris_edouard · 2026-11-03
- If AI disrupted any other industry as fast as software, there'd be strikes everywhere — andreisavu · 2026-11-03
- Aumann's agreement theorem, clarified: what persistent disagreement really implies — _onionesque · 2026-11-03
- Revisiting 50 LLM findings: 44% expired within 15 months, only 6% strengthened — james_y_zou · 2026-11-03