Automated behavioral benchmarks, published as they run.
Every 3 days, Aletheia probes GPT-4o, Claude, and Gemini across 9 behavioral failure signatures — hallucination, scope creep, crisis blindness, and more. Results are published here automatically. No editorial filter.