Qiushi Engine achieves autonomous scientific discovery on optical platform

A team of researchers has created the Qiushi Discovery Engine, an agentic system based on large language models (LLMs) that achieves fully autonomous scientific discovery on an actual optical platform. In contrast to earlier AI systems that merely follow established workflows, the Qiushi Engine can independently replicate a published transmission-matrix experiment, yielding significant results backed by experimental data. This innovative system integrates nonlinear research phases, Meta-Trace memory, and a dual-layer architecture, allowing it to sustain adaptive and stable research paths over extensive investigations that involve thousands of reasoning, measurement, and revision actions mediated by LLMs. This represents the inaugural instance of complete autonomous scientific discovery within a physical framework.