Andrej Karpathy says only 5,000 people have seen what AI can really do

In a Sunday post on X, Andrej Karpathy—an OpenAI founding member who joined Anthropic in May—estimated that roughly 6 billion people have barely used an LLM, 1 to 2 billion use chatbots casually, about 20 million use frontier models professionally for math and code, and only around 5,000 people have access to frontier systems inside the labs. Box CEO Aaron Levie agreed, calling this "a good way to understand how early we actually are in AI diffusion right now," and pointing to "agent swarms" as a promising enterprise form factor that few currently use due to expensive scaling, limited model capability, immature management tools, and data, security, and compliance challenges. Karpathy explained that capability peaks in digital, verifiable, economically valuable domains like coding and math, where reinforcement learning and testing work best, citing his AutoResearch project that ran 50 overnight experiments on a single GPU and committed an improved learning rate to Git. Both he and Levie highlight "the weirdness of the moment": most people judge AI by casual chatbot interactions while a small group of developers sees staggering capability gains, and for Box, this gap represents a major enterprise opportunity.

(Source：The New Stack)