Using agents to reimplement MD forcefields: huge speedups possible, but forces must still agree
rishabh16_ · x · 2026-11-13
rishabh16 envisions agents applied to computational science: beyond recent cases of agents combing codebases for improvements (e.g., fixing the softmax/x-ent bug in torch/jax), he expects agents to optimally (re)implement forcefields/potentials for MD and FEP, where large speedups likely remain untapped — citing discoveredmat's recent DFT acceleration as encouragement.
He flags the equally important question: are we optimizing FF/IP code correctly? Faster energy computations only count if forces agree too, preserving the quantities that matter.
Related event: AI agents speed up scientific computing, boosting DFT 36x(3 posts)→
More from coding & agent
- Dual-agent AI harness optimizes chemistry codes, cutting TDDFT cost by up to 94% — CatAstro_Piyush · 2026-11-13
- Claude builds 3D sets via MCP, Nano Banana and Seedance render an action teaser — azed_ai · 2026-11-13
- User moves long Codex sessions to T3, calls it a big upgrade over stock GUI — natesiggard · 2026-11-13
- TLAPS-proof LLM serving engine: 516,973 obligations at just 0.18% latency overhead — tianyin_xu · 2026-11-13
- 1Password for Agents lets AI agents sign in without risking your credentials — dean_rie · 2026-11-13
- Compound Engineering 3.38 adds /ce-retro: mine past agent sessions for workflow fixes — kieranklaassen · 2026-11-13