Karpathy says AI’s public preview hides elite access to frontier systems
Andrej Karpathy, OpenAI co-founder and Anthropic researcher, argues that most people see only a limited view of AI’s capabilities. He divides users into four tiers: six billion have no exposure to large language models, one to two billion use them casually as search or writing aids, and 20 million professionals apply frontier models in coding and math. Karpathy estimates just 5,000 people…
Key points
- Karpathy divides AI users into four tiers: 75% have no exposure, 20M use frontier models professionally, 5K access elite systems
- Elite systems enable tasks like software development, cybersecurity, and scientific discovery at unprecedented speeds
- AI progress is concentrated in coding and math due to verifiable rewards and economic value
Karpathy notes the gap between public perception and elite access is widening. He cites his own shift from manual coding to agent-driven development, and compares the current moment to the OpenAI-HF incident, where human comprehension struggles to keep pace with AI advancements. He also highlights that AI’s rapid progress is concentrated in digital, verifiable domains—coding and math—due to their economic value and measurable rewards. Sam Altman previously echoed this, noting an AI tool could complete in 17 minutes what a startup took three months to build.
Anthropic's Andrej Karpathy says "the external world has seen the preview" of what AI can actually do
The Decoder · 12 October 2026
Loading the full article…
This text was published by The Decoder and written by Matthias Bastian. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Society & Work
All →- Cerebras CEO dismisses AI doomsday warnings at TechCrunch Disrupt · 2 src
- Nikon disqualifies microscopy winner over AI use · 3 src
- Signal’s CEO warns careless AI use risks war, not dystopian scenarios · 1 src
- System76 bans AI-generated code in cosmic desktop contributions · 1 src
- OpenAI engineer arrested for allegedly planting recording device in office · 11 src
Comments
via GitHub Discussions