George Hotz, founder of comma.ai and tinygrad, published an essay on June 21, 2026, arguing that while LLMs are useful now, the productivity gains are not as large as vendors claim, and there is no path to superintelligence. He partly walks back his earlier post, *The Eternal Sloptember*, stating he was probably too harsh on models' ability to program, noting that programming itself is changing and using agents is a skill he is actively getting better at. He describes setting up a local Linux environment running opencode against GLM-5.2 hosted on Hugging Face, where a natural-language command to install and configure tmux worked without any manual steps. Hotz attacks two patterns: the claim that developers are "falling hopelessly behind" and will form a "perpetual underclass," which he reads as fear marketing, and the leap from "models are useful" to "superintelligence is near," which he treats as a valuation argument. He says he gets "some boost from the models" and is "pretty confident I'm getting better at using them," but warns of cognitive fatigue as a counter-weight to raw speed gains. He questions where the flood of magical new software is if claimed gains are real, and attributes AI progress to general computing advances tracking Moore's Law.