Karpathy: 75% of people have barely touched LLMs as the capability perception gap widens
karpathy · x · 2026-10-12
Karpathy revisited his April tweet, arguing the gap in shared understanding of LLM capability is widening into a sharp funnel:
- 6B people (75% of the population) have barely come in contact with LLMs at all;
- 1-2B (20%) are casual free-tier users who see derpy chatbots — a better Google search or writing aid — and often find them not very useful;
- those using this year's state-of-the-art agentic models like OpenAI Codex and Claude Code experience something entirely different.
He notes viral mockery of model quirks and hallucinations is largely based on old, deprecated free-tier models, not the latest frontier agents.
Related event: Karpathy: LLM Awareness Gap Widening Sharply(4 posts)→
More from AGI Musings
- Hinton's 2016 radiologist prediction flopped a decade later — same hype now targets mathematicians — jasondeanlee · 2026-10-12
- Fields medalist shaken by OpenAI's 700 AI-generated math manuscripts touching his work — LukeW · 2026-10-12
- Defending the 30-day 8000-sqft wet lab: engineer pushes back on biotech gatekeeping — kenbwork · 2026-10-12
- Hamel Husain: no single role wins the AI era, breadth beats moats — HamelHusain · 2026-10-12
- As mathematicians protest AI, a 1950 Wiener classic on human obsolescence resurfaces — soumitrashukla9 · 2026-10-12
- Practitioner's verdict on autoresearch: great at speeding experiments, not at frontier runs — iaindunning · 2026-10-12