AI Enthusiast: Frontier Labs Won't Capture the Value
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- The author argues that frontier AI labs won't capture AI's value because the technology rides Moore's law and is destined for commodification, which he says is why labs push anti-open-source arguments dressed up as 'safety' or 'China' concerns.
- Coding agents deliver real productivity gains the author compares to find-replace or Stack Overflow — he reports that telling opencode on a local GLM-5.2 to 'install tmux with the geohot configuration' 'just works.'
- The author, who hacked from 2007–2014 before devoting his career to AI, says he is genuinely excited about new LLMs, self-driving cars, and video generation models, and declares that 'the Year of the Linux Desktop is finally here.'
- He condemns two hype patterns: 'negative valence hype' about windows closing and a perpetual underclass, and the strawman leap from 'fancy autocomplete' to claims AI will 'own the whole light cone.'
- Linus Torvalds is cited saying agents make programming 10x more productive while compilers make it 1000x more productive; the author calls both figures extreme but says he's personally getting a real boost and that the skill is learnable.
- The author revises his earlier 'Eternal Sloptember' take, conceding that vibe-coded code is still slop and that agents can increase cognitive fatigue, but likens the upside to 'regexes I never learned how to write and now never will.'
- His bottom line: AI 'is the continuation of the computer revolution' — and he links to a 2016 superintelligence talk and a 1991 machine-takeover film to mock the 'cult' claiming credit for trends happening regardless of them.
Why it matters: Coming from a well-known hacker-turned-AI-builder with insider credibility, the argument lands directly on frontier-lab valuations: the author claims labs are priced as if they'll capture a singular, defensible breakthrough, when the technology is actually a commoditizing continuation of Moore's law. If open-source models like GLM-5.2 keep that pattern going, the proprietary moats justifying current AI-lab valuations erode by default — no singularity required.
