I love LLMs, I hate hype
Points and comments are a snapshot, not live.
Geohot loves AI progress but hates hype and anti-open-source FUD.
The author argues that claims of a closing window or permanent underclass are negative hype designed to lure people to San Francisco. He dismisses AGI doomerism as a strawman, betting the light-cone scenario will not happen. He asserts frontier labs won't capture AI's value because AI is driven by Moore's law and general progress, not their efforts. Programming is changing, and LLMs boost productivity like compilers or regex, but require careful use to avoid cognitive fatigue and slop.
What commenters are saying
Commenters split into camps: those defending Geohot as a craftsman vs. a merchant, those decrying unavoidable doom-hype in real life, and those debating whether AI can make art. A key correction: Geohot's SF hate is seen as misplaced because SF is broader than its tech echo chamber. One commenter notes the hype and anti-hype both drive CapEx for LLM improvement.