Tag
Boris Cherny, creator of Claude Code, outlines five levels of AI agent usage, from basic prompting to autonomous agents that start and learn on their own, with the key insight that the agent itself is the worst judge of task completion.