I struggle to believe that AI R&D is ingenious

The author expresses skepticism about claims that AI excels at autonomous research and development, citing multiple projects where AI systems demonstrated poor judgment in identifying important literature, designing informative experiments, and recognizing valuable ideas. While acknowledging that AI performs better on tasks with fast feedback loops and clear optimization targets, the author argues that AI R&D capabilities are significantly more limited than commonly believed, as AI frequently overlooks straightforward analyses and pursues trivial research directions.