AI-generated puzzle challenges, audited results from eleven model configurations, and the research paper with reproducible sources.
AI & ML interests
Post-training for recursive self-improvement (RSI), agent environments, evaluation, calibrated decision models