The Huxley–Gödel Machine (HGM) was proposed by a research team at King Abdullah University of Science and Technology in October 2025, and the relevant research results were published in the paper "Huxley-Gödel Machine: Human-Level Coding Agent Development by an Approximation of the Optimal Self-Improving Machine".

HGM approximates Gödel (GM)-style self-improvement by estimating clade-level metaproductivity (CLM) from the descendant results of clade aggregation, and expands by selecting nodes through Thompson sampling. Furthermore, by leveraging more reliable estimates, it adaptively decouples expansion from evaluation, enabling asynchronous execution for efficient parallelism. HGM generalizes across datasets and model variations, achieving human-level coded agent design performance on SWE-bench Lite.