Andrej Karpathy, former Tesla AI chief, highlighted a growing divide in how large language model (LLM) capabilities are understood globally. He estimates that about 6 billion people, roughly 75% of the world population, have had little to no interaction with LLMs. Meanwhile, around 1 to 2 billion users engage with free-tier ChatGPT-like tools casually, often treating them as enhanced search aids or writing assistants, according to his April tweet.
Karpathy breaks down LLM users into distinct groups: casual users who see chatbots as simple tools, and a smaller professional cohort of about 20 million people who use advanced LLMs for complex tasks in math and coding. This professional group experiences firsthand how LLM-powered agents can complete projects in minutes that previously took weeks or months. He contrasts this with his own experience less than a year ago, manually coding line by line before the rapid rise of agent-driven automation.
The widening gap in LLM understanding reflects broader market dynamics where most of the global population remains unfamiliar with frontier AI capabilities, while a niche group leverages these tools to transform software development and other technical fields. This divide underscores challenges in AI adoption and education, as well as the uneven impact of AI advancements across sectors and geographies.
Karpathy’s observations were shared on Twitter on October 10, 2026, providing a snapshot of LLM adoption stages worldwide and emphasizing the rapid evolution of AI tools within the last year.