Mythos 5.2 accurately models 28 of 34 viral capsids for gene therapy design
MoAlQuraishi · x · 2026-11-10
Anthropic's Mythos 5.2 found a new way to predict the structure of whole capsids — the protein shells that deliver gene therapies — accurately modeling 28 of the 34 capsids attempted, potentially letting biologists design capsids faster and cheaper.
- Practitioner amirshanehsaz, formerly at Dyno Therapeutics, praised its creative and compute-efficient AAV folding approach using OpenFold3-p2 and ESMFold2, with review from the Dyno team.
- OpenFold lead Mo AlQuraishi amplified the work, crediting the folding-model teams involved.
Related event: Mythos 5.2 models viral capsids with 28 of 34 accurate(2 posts)→
More from Research
- Copy-from-document speculative decoding speeds AI extraction 4.7x (up to 19x) vs naive vLLM — sh_reya · 2026-11-10
- AI writing benchmark finds Claude Code replies use vague wording 60% of the time vs 4% for humans — IanArawjo · 2026-11-10
- UCLA's Visible Touch turns touch into image overlays, boosting VLA robot success by up to 30pp — Scobleizer · 2026-11-10
- Returns to AI research sit near the knife-edge: r ≈ 1.2 — austinc3301 · 2026-11-10
- AI takeoff models boil down to three questions, says new synthesis — austinc3301 · 2026-11-10
- Testing the "RL catching up to bigger pretrains" hypothesis against METR timelines — austinc3301 · 2026-11-10