Dual-agent AI harness optimizes chemistry codes, cutting TDDFT cost by up to 94%
CatAstro_Piyush · x · 2026-11-13
An arXiv paper (2611.10378) by Mathew Chow and Tao E. Li presents a dual-agent harness for controlled optimization of scientific software:
- Separate worker and controller agents handle code modification and verification, reducing computational cost while preserving numerical accuracy.
- Applied to TIP4P water simulations in LAMMPS and to Coulomb/exchange matrix construction plus linear-response TDDFT in PySCF; TDDFT saw the largest speedups, up to 94% runtime reduction.
- Extending the protocol, they converted a serial molecular dynamics code into parallel CPU- and GPU-accelerated implementations, achieving 2–3 orders of magnitude speedup.
- They also built an adaptive self-consistent field approach coordinating multiple solvers, converging challenging systems and occasionally reaching lower-energy solutions than standalone solvers.
The result suggests AI agents are becoming reliable tools for accelerating computational science code.
More from coding & agent
- Armin Ronacher: I started 141 projects in a year, but only maintain 10-15 — TheTuringPost · 2026-11-13
- One prompt spawns a 1M-nanobot GPU swarm that morphs into any shape at 60fps — bennash · 2026-11-13
- Dev shares personal agent stack: Hermes + open-source Hermex iOS client — _ScottCondron · 2026-11-13
- Anthropic launches self-hosted Claude Managed Agents in private beta for on-prem deployment — ClaudeDevs · 2026-11-13
- DeepSeek V4 Pro agent sets up a full Hive lab in VMware in about half an hour — huangyun_122 · 2026-11-13
- Benchmark Router launches to run public LLM benchmarks on any model at 1/10th the cost — ycombinator · 2026-11-13