AI Systems Show Promise in Scientific Discovery and Optimization

PSI’s Fable 5.2 reportedly contributed to work on a long-open quantum-complexity proof, though outside understanding and practical validation remain incomplete. Opus 5.5 and GPT-6 Astra solved all 55 TIS-100 puzzles in a rigorously replayed but narrow benchmark. DeepMind’s AlphaGenome 2 adds a biological-research signal, with claimed gains awaiting independent validation and evidence of experimental impact.