AIOpenRSI-Index v0.1 launches to test recursive AI self-improvement
OpenRSI says the benchmark turns fully open-source projects into research environments, with AI agent runs lasting 60-plus hours on 1,000-GPU clusters.
TLDR
OpenRSI has released OpenRSI-Index v0.1, an open benchmark intended to evaluate whether AI can repeatedly improve itself and advance science beyond human-designed methods. The organization says building this version took 100,000-plus hours of H100 GPU time. It is inviting task contributors and compute partners to help develop the benchmark and says all contributors will be included as paper authors.
As AI systems enter a recursive self-improvement loop, the central question is whether it can systematically move beyond human-designed methods to genuinely extend the scientific and intelligence frontier. What’s missing is a neutral, open standard for evaluating these…
