GoKawiilAI researcher Andrej Karpathy says awareness of LLM capability is splitting into four tiers
Andrej Karpathy argues that public understanding of large language models' capabilities is diverging sharply rather than converging. He outlines a rough tier system: roughly 75% of people have barely used LLMs, 20% use free chatbots casually, about 0.2% of professionals in coding and math see agents completing weeks of work in a prompt, and an estimated 5,000 people with internal access to frontier systems are reportedly seeing swarms of agents handle multi-week software projects, cyberattacks and research-level math.
GoKawiil's interpretation of the reporting above, not reported fact.
Karpathy's framing suggests that policymakers, the public and even many professionals could be forming opinions about AI based on vastly outdated or incomplete information, since the most advanced capabilities are visible only to a tiny number of insiders. This could complicate public debate, regulation and safety planning if decision-makers are reasoning from the lower tiers of experience rather than the frontier. It also implies, in his view, that human oversight and review processes may struggle to keep pace with what the most capable systems can already do.
- Karpathy estimates only about 5,000 people worldwide have hands-on access to the most advanced, frontier-grade AI systems.
- He says roughly 75% of the global population has had little to no contact with LLMs at all.
- He warns that human review and comprehension of AI capabilities are falling behind actual progress, widening a knowledge gap.
GitHub Copilot subscription — If you're curious what frontier-grade LLM coding assistance actually feels like rather than hearing about it secondhand, GitHub Copilot puts agentic code generation directly in your editor. It's the most accessible way for developers to experience the gap the article describes, turning multi-step coding tasks into prompt-driven workflows.
See GitHub Copilot subscription on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: twitter.com, 2026-10-11
Published there as: “The gap in shared understanding of LLM capability is widening”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.