Andrej Karpathy's "AI psychosis" framing, which this paper examined yesterday as a description of the comprehension gap between heavy AI users and everyone else, has crossed from X discourse into mainstream coverage. The New Stack, Fortune, and Medium all published analyses within days of Karpathy's original post. [1]

The phrase has legs because the messenger is credible. Karpathy co-founded OpenAI, led Tesla's Autopilot team, and now works independently — which means he has no product to sell when he says the gap between what AI can do and what most people believe it can do has become a chasm. His description is precise: developers who use frontier models daily are experiencing a reality so different from the general public's understanding that the two groups can barely communicate about the technology. [2]

The industry response has been instructive. AI boosters adopted the term as validation — if even insiders feel the shift is disorienting, the technology must be transformative. Critics seized on it as evidence of a bubble psychology, where participants mistake their own confusion for insight. Karpathy himself appears to occupy neither camp. His post was diagnostic, not prescriptive.

What makes the frame durable is its refusal to take sides. "AI psychosis" names a condition without claiming to know the prognosis. That ambiguity is why it spread.

-- KENJI NAKAMURA, Tokyo