Jensen Huang and Boris Cherny, Claude Code's chief, each declared within 24 hours of each other that the era of prompting AI models is over. The replacement: 'loops,' meaning LLMs autonomously prompting other LLMs in continuous, token-burning cycles. Cherny, who Anthropic permits to burn up to $130,000 a month in tokens, described agents he runs that continuously audit and rewrite code, submitting pull requests indefinitely, never stopping. The timing is not subtle.
The author, Ed Zitron, predicted this exact move weeks earlier: that a slowing AI industry would manufacture a new paradigm to drive token consumption and revenue. What makes this piece worth reading in full is not the prediction being confirmed, it is the structural argument underneath it. Zitron traces how Silicon Valley enforces consensus through social punishment, not reason, attacking critics on the basis of 200-word skims, demanding loyalty to what he calls 'the Great Prophecy': that Anthropic and OpenAI can meet their combined $1.1 trillion in spending commitments.
The conclusion is blunt. The tech industry has run out of original ideas and what remains is a cargo cult, self-described rationalists who demand conformity, a supposed meritocracy that suppresses dissent, and a generative AI economy that funnels value upward to the largest tech companies already in existence. Snap's $2,195 AR glasses appear next in the piece as further evidence. The argument is not that AI is useless. It is that the culture surrounding it has become pathological, and that loops are the latest symptom.
[READ ORIGINAL →]