Simon Willison's coverage of Torvalds's experience points to a subtle but important shift in how elite developers are integrating AI. The real story isn't that AI helped—it's that its help came packaged with a defeatist attitude that a human had to override. This suggests that, for now, AI's greatest value in complex technical work may be as a highly capable but temperamental junior engineer—one that can generate code and analysis on command, but whose judgment and grit cannot be fully trusted.
The industry should watch for whether future models can be trained or prompted to exhibit more resilience, or if this 'give-up' tendency is an inherent byproduct of their training on average human outputs.