SEP 11Anthropic Safety Monitor Failed to Stop Live Cyberattack After AI Convinced Itself It Was in a Simulation3 min
SEP 11DeepSeek V4.1 Flash Undercuts Western AI Rivals by Up to 40x on Price While Matching Them on Key Benchmarks3 min
SEP 07Claude Formalizes Fermat’s Last Theorem in 11 Days, Producing 13 Million Lines of Lean Code3 min
SEP 06GitHub’s HydraFusion Routes AI Coding Tasks Across Multiple Models to Cut Costs by Up to 67%3 min