Virtual Human: formalizing the body as executable graphs so agents can solve biology with code
Vjeux · x · 2026-11-03
Rohan Arun open-sourced virtual-human, exploring whether agents can solve "impossible" problems by formalizing the real world as executable typed graphs instead of brute-force atom-scale simulation.
- Multi-scale body simulation: coupled-cluster QM on molecules, an atom-level physics engine for proteins/drugs, and a whole-body engine predicting drug effects — validated at every level, with a real-time browser demo.
- Honest boundary: graphs solve structural/relational problems (connectivity, flows, consistency); strength/speed/folding stay physics.
- Graph algorithms give exact answers: blood conservation holds to 1.5×10⁻¹⁶ (Kirchhoff), first-pass metabolism emerges from topology, and elimination-subgraph clearance matches the full model exactly (28.488 L/h).
- The graph verifies the physics engine: BFS on the bond graph reproduces all 1687 force-field exclusion pairs exactly.
- New explicit-solvent engine: per-water-molecule Ewald/PME electrostatics validated against OpenMM to 10⁻¹¹–10⁻¹³.
MIT licensed, built to be explored by agents.
More from coding & agent
- DeepSeek releases all 7 demo prompts for Harness Creator Mode plugins — rohanpaul_ai · 2026-11-03
- Ex-ByteDance engineer on Coze shutdown: workflows killed by models that plan — dotey · 2026-11-03
- Apple's iPrefBench tests personal agents on 130 underspecified iOS tasks: infer or ask? — hhsun1 · 2026-11-03
- Yale lab open-sources a Claude skill that makes publication-quality schematic figures — CatAstro_Piyush · 2026-11-03
- Prized launches Infinite Storage: boxes scale to 40TB at local-disk speed — thisguyknowsai · 2026-11-03
- Agent-era take: if AI reads your git history, the rebase-vs-merge debate stops mattering — davidcrawshaw · 2026-11-03