The AI superusers and dismissers may be drifting further apart.<\/p>\n
\n Loading audio narration…\n <\/p>\n
Andrej Karpathy might as well be called AI’s philosopher king. The former Tesla director and OpenAI founding member often posts long, multi-paragraph explanations of AI trends on X. And you’ve probably heard one of the terms he coined that wound up as a new word in the dictionary: “vibe coding<\/a>.”<\/p>\n In his latest<\/a>, Karpathy wrote that there’s a “growing gap in understanding of AI capability.” He divided users into two populations: the everyday user who has tried out free ChatGPT<\/a>, and perhaps only tinkered with early versions of AI chatbots, and the power user who pays to employ the latest models in their work.<\/p>\n Judging by my tl there is a growing gap in understanding of AI capability.<\/p>\n The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is\u2026 https:\/\/t.co\/Kx1EwuAYmt<\/a><\/p>\n \u2014 Andrej Karpathy (@karpathy) April 9, 2026<\/a> <\/p>\n The first demographic, Karpathy wrote, will laugh at AI’s quirks and hallucinations. He referenced a viral video<\/a> where OpenAI’s voice mode fumbled the question early on: “Should I drive or walk to the carwash?”<\/p>\n “The thing is that these free and old\/deprecated models don’t reflect the capability in the latest round of state of the art agentic models of this year,” he wrote.<\/p>\n The second demographic, then, is the user who pays for Claude Code<\/a> and Codex and assigns “a much greater gravity to the capabilities, their slope, and various cyber-related repercussions.”<\/p>\n These users often employ AI for tasks like programming, math, and research, Karpathy wrote. That’s because these models are “peaky” in technical areas, he wrote, which are easier to train and can be more lucrative.<\/p>\n “Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability,” Karpathy wrote.<\/p>\n In a reply, Karpathy wrote that a friend had suggested the OpenClaw moment<\/a> grew “so big” because it was the first time this first demographic was interacting with newer models.<\/p>\n More broadly speaking, there is clearly a divide among the general population in terms of AI adoption. OpenAI CEO Sam Altman<\/a> said that AI is “not very popular,” citing data center worries and CEOs using the tech to justify layoffs. Sometimes the AI superfans and avoiders get together, forming what Business Insider<\/a> called a “Claude-gap relationship.”<\/p>\n As more people continue to experiment with Claude Code, OpenAI’s Codex, or Cursor, that could change.<\/p>\n For now, Karpathy wrote that the two demographics are “speaking past each other.”<\/p>\n