Protein binder competition test: can an LLM orchestrator out-reason human scientists?
madeofAjala · x · 2026-11-21
During the adaptyv × Anthropic binder competition, Proteus Bio and FinalDose tested whether therapeutic-grade reasoning should live in humans or AI. Getting A to bind B is no longer hard (RFD4, BindCraft2, mol-1 all shipped); the real challenge is cross-species binding (human + mouse), pH-gated switching, receptor agonism, and mutant-vs-wildtype selectivity — constraints requiring reasoning beyond SOTA models. They ran three architectures, including Route A: an LLM orchestrator driving BindCraft2 and ESMdesign, raising questions of how agents choose and evaluate tool results and how faithful their reasoning traces are.
Related event: LLM Orchestrators vs Human Experts in Protein Binder Competition(2 posts)→
More from coding & agent
- Dev compiles Rails into the browser for offline, millisecond-response apps — kieranklaassen · 2026-11-21
- Vibe-coded site lets anyone independently verify the claimed Riemann proof with one Python command — sudoraohacker · 2026-11-21
- Dev builds FPS horror game in an afternoon with Fable 5.2 — TAbrodi · 2026-11-21
- Google's Stitch Loop agent autonomously splits goals into KPIs and builds fixes with Gemini — testingcatalog · 2026-11-21
- AFK Pilot adds Gemini 4 Neon support, warns of Antigravity impersonation bans — PawelHuryn · 2026-11-21
- Eric Elliott: agent skills favoring pure functions yield less error-prone code — ericelliott_ · 2026-11-21