Anthropic co-founder Jack Clark moves prediction for self-improving AI forward to 2027

Anthropic co-founder Jack Clark has moved his prediction for recursive self-improving AI forward to 2027, citing new benchmark data. The TasteVal benchmark, developed by P-Zero Research, measures AI models' research taste across tasks including training data selection and defending against adversarial prompts, though it only tests two of the three defined abilities. The top model, Anthropic's Opus 5.5, achieved a compute multiplier of 2.3, reaching expert-level scores with less than half the compute required by human experts. The study found that across several model generations, the compute multiplier has roughly doubled every three months since December 2025.