Abundant Labs launches FrontierRSI, a benchmark where Claude must improve Claude

Madisonkanna · x · 2026-11-20

Abundant Labs released FrontierRSI, a new recursive self-improvement (RSI) benchmark. The team argues existing "RSI" benchmarks miss the point: true RSI requires recursion — an LLM improving the LLM itself. "If you have Claude improve Waymo, there's no way for Waymo to improve Claude." The benchmark evaluates agents on a broad range of AI research and engineering work, plus post-training on multi-step agentic tasks rather than the single-step setup of PostTrainBench.

Related event: FrontierRSI: New Benchmark Tests Whether AI Agents Can Do Real AI Research(2 posts)→

Original post →

More from coding & agent

coding & agent channel →