News · 2026-07-12

George Hotz: I love LLMs, I hate hype -- and the labs won't capture the value they create

George Hotz -- the hacker who jailbroke the iPhone and PS3, now running self-driving startup comma.ai -- published an essay titled "I love LLMs, I hate hype" arguing that AI is genuinely valuable but that the frontier labs will fail to capture the value they create. His core line: "It's not that AI won't create that much value, it's that they won't capture it." AI, he argues, is the continuation of the general computer revolution, driven mostly by Moore's law and broad progress in computing -- not by any unique thing the frontier labs own.

Key facts

• Hotz's thesis: labs will create value but not capture it, because AI extends the general computer revolution rather than being a proprietary moat.
• He set up a Linux box running the open OpenCode agent on a local GLM-5.2 model "last week" and reports it works.
• He calls anti-open-source arguments from labs "a fear of commodification" dressed up as safety or China concerns.
• Posted July 12, 2026; drew 315 points on Hacker News. Hotz's essay, Hacker News.

The hook is that this is not a skeptic sneering from the outside. Hotz opens by insisting readers "may misunderestimate how absolutely giddy I am about AI" and notes his whole career since 2014 has been AI. He uses these tools daily. That is what makes his two specific complaints land. First, "negative valence hype" -- the window-closing, falling-hopelessly-behind, perpetual-underclass rhetoric, which he says "is mostly designed to make you feel bad about yourself." Second, the "strawman jump" from "fancy autocomplete, smart compiler" to "it's gonna own the whole light cone bro," which he will "bet everything" does not happen.

The economic argument is the substantive core, and it needs one piece of background. Frontier AI is sold two ways: cheap flat-rate subscriptions (roughly $20-200 a month) and metered API access priced per token, which can run far higher for heavy use. Open-weight models -- ones you can download and run yourself, like the GLM-5.2 Hotz runs locally -- are far cheaper at scale. Hotz's claim is that the labs are pricing for a world where everyone pays the high metered rate, and that world will not arrive because "good enough" open models keep closing the gap. A Hacker News commenter (SwellJoe) put the math starkly: at subscription prices frontier models are a no-brainer, but "the frontier labs need everyone to answer yes to spending 100x what they currently spend to justify the valuations, and it's just not going to happen."

Think of it like the early PC era. Enormous value came from personal computing, but it did not accrue to whoever built the fanciest single machine -- it diffused across the whole industry as chips got cheaper and software commoditized. Hotz is betting AI follows the same arc: the value is real and huge, and it leaks out to everyone rather than concentrating in a few labs.

Why it matters: this argument is suddenly everywhere. The same day, Microsoft CEO Satya Nadella published his own version of the same worry from the opposite chair, and the day's flood of capable open-weight model releases -- a 1-trillion-parameter Xiaomi model under an MIT license, new efficient local vision models -- is the commodification thesis playing out in real time. When the labs argue against open weights on safety grounds, Hotz hears "a fear of commodification."

Hotz is not a pure booster, which is the honest caveat baked into his own piece. He softens an earlier harsher critique -- "I'm now pretty confident I'm getting better at using them and get some boost from the models" -- while still insisting "all the vibe coded stuff is still slop (where's all this new magical software that the productivity improvements should imply?)." That tension is the point: the tools are real and useful and he uses them, the productivity gains are modest and the output is often mediocre, and the trillion-dollar valuations assume a capture that the economics may not deliver. It is a bet, stated as a bet, from someone with skin in the game.

Primary source, verified: read the paper →

Ground Truth. (2026, July 12). George Hotz: I love LLMs, I hate hype -- and the labs won't capture the value they create. Ground Truth. https://groundtruth.day/news/george-hotz-i-love-llms-i-hate-hype.html

@misc{groundtruth:george-hotz-i-love-llms-i-hate-hype,
  title  = {George Hotz: I love LLMs, I hate hype -- and the labs won't capture the value they create},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {jul},
  url    = {https://groundtruth.day/news/george-hotz-i-love-llms-i-hate-hype.html}
}

Topics: george-hotz · commodification · open-weights · economics · ai-hype