Reproduce “Closing the Analyze-to-Design Gap: An Authorization-Bounded Agent for Self-Driving Laboratories” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yimeng Liu, Eric Isbrandt, Felix Katzenburg, Kevin P. Greenman, Han Hao, Yang Cao, Varinia Bernales, Alán Aspuru-Guzik

The authors present an agent that interprets results and selects LC–MS samples, while laboratory services validate and execute its requests. In a physical oxidant screen, potassium permanganate at the highest tested loading gave the strongest product response. In archived replay, Claude Opus 5 found the best well after about seven injections, versus 14.5 expected under random ordering, with no informed baseline tested.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.