Two competing visions of AI coding emerge: a surgeon's careful partner versus a metered commodity—raising questions about what role AI should actually play in knowledge work.

What happened: The essay contrasts two metaphors for AI-assisted coding. Geoffrey Litt proposed viewing the AI as a surgeon's assistant—the human does high-value work, the AI handles routine tasks. Sam Altman of OpenAI instead framed intelligence as a utility sold on a meter, like electricity or water.

Why it matters: The metaphors reveal fundamentally different assumptions about AI's reliability and role. The surgeon model assumes AI needs careful oversight and formal process (the author notes surgeons follow strict protocols and their assistants receive direct correction). The utility model treats AI as a commodity to consume. The author suggests the utility framing misnames what LLMs actually deliver—not intelligence, but a stream of 'internet consciousness' that requires human process and learning to be useful.

What to watch: The core tension is whether AI coding tools can be trusted like a trained surgical assistant (which internalizes feedback and performs consistently) or remain unpredictable streams requiring constant human judgment. The author expresses doubt that Claude, as currently built, can be relied upon for critical decisions—implying that AI-assisted work still depends on building organizational discipline, not just consumption.