Summary

Zitron presents a comprehensive manifesto against the generative AI industry after three years of coverage, arguing that LLMs are mediocre tools propped up by hundreds of billions in subsidized spending that produces no real economic value. The technology has narrow, controlled utility but fails catastrophically when stretched beyond basic tasks, yet the industry manufactures consent through media pressure and social coercion to obscure this reality. The underlying economics are catastrophically broken — OpenAI's non-GAAP operating margin worsened from -122% to -183%, NVIDIA just raised GPU prices 17%, and even OpenAI's own economists found no statistically significant correlation between AI usage and revenue per employee. Because LLM improvements depend entirely on tens of billions in annual training spend that only exists during financial mania, the post-bubble future for frontier models is essentially nonexistent. The entire bubble runs on bad faith: subsidized pricing, annualized run rates instead of revenues, and perpetual future-tense promises that demand trillions while delivering little.

Key Insight

The AI bubble is not primarily about technology at all — it is a growth-manufacturing scheme for a maturing tech industry, sustained by manufactured consent, subsidized pricing that hides true costs, and a financial daisy-chain that collapses the moment any major link stops spending.

Spicy Quotes (click to share)

• 9

  Four years into the AI bubble, and the best you've got is that a tool kind of worked after I bonked it on the head multiple times, and all it cost was a trillion-plus dollars in capex and tens of billions of dollars of training compute.

• 9

  Please, point to the time in history when we have invested a trillion fucking dollars in making human workers better.

• 8

  It is impossible to view LLMs in a vacuum, because their existence demands hundreds of billions of dollars. Every data center is incredibly expensive, offensive-sounding and looking, and their existence is explicitly to enrich some sort of Patagonia-gargoyle at an asset management firm.

• 8

  One of the many dangerous things that an LLM can do is a messy impression of a competent person, filling in the little bits within a loser, moron or con artist that would've otherwise exposed them, allowing them to get deeper and deeper into organizations by creating make-work.

• 8

  Nothing about LLMs is worth a trillion dollars, or even $100 billion. This is a $30 billion TAM industry dressed up as a trillion dollar one.

• 9

  The AI industry — and the AI bubble — is fundamentally built on acting in bad faith. Its executives lie. Its boosters lie. Its software lies because it doesn't actually know anything and generates answers probabilistically.

• 7

  LLMs were meant to be the panacea for a tech industry that ran out of new ideas for growth. You'll notice how none of these come with some sort of problem they're solving other than 'more.'

• 5

  To be clear, they're right. It's so much weirder, and in the end will be so much worse.

Tone

incendiary, combative, and polemical — backed by financial data but delivered with sustained moral outrage