Karpathy's napkin math: only ~5,000 people have seen what AI can really do
bittingthembits · x · 2026-10-13
Andrej Karpathy — OpenAI founding member, former Tesla Autopilot lead, creator of CS231n, coiner of "vibe coding", now at Anthropic — offers a napkin-math breakdown of the AI usage gap:
- 6 billion people have barely used an LLM
- 1-2 billion use chatbots casually
- 21 million use frontier models professionally for math and code
- Only 5,000 have access to frontier systems inside the labs
His point: the gap is widening, not narrowing. Most people still judge AI by chatbots that "fumble the dumbest questions," while inside the labs, agents are melting problems that used to take days or weeks of work.
Related event: Karpathy: LLM Perception Gap Is Widening Into a Steep Funnel(24 posts)→
More from AGI Musings
- RLVR produces "evolved modularity" AIs; 3D reasoning gains confirmed as RL-trained, not emergent — jessi_cata · 2026-10-13
- Will Depue argues progress will soon be data bottlenecked, not intelligence bottlenecked — mallow610 · 2026-10-13
- Two GLP-1 drugs outearn OpenAI and Anthropic 2-to-1, a case for AI-driven R&D — Afinetheorem · 2026-10-13
- Language is a latent space for world knowledge, at the cost of information loss — docmilanfar · 2026-10-13
- Thirty years of moving the goalposts: what AI keeps revealing about humanity — r0ck3t23 · 2026-10-13
- The day we can't tell cricket sounds are real: AI and the trust divide — spilltrend · 2026-10-13