Anthropic researcher Andrej Karpathy stated that the external world has only seen a preview of actual artificial intelligence capabilities. Karpathy outlined a stark divide in global AI usage, noting that about 75 percent of the world's population has barely touched large language models, while treating chatbots as mere writing aids or better search engines.
According to Karpathy's breakdown, only about 20 million professionals in mathematics and programming regularly use frontier models for their work, experiencing firsthand how complex projects can be completed through agents and prompts. Furthermore, an elite group of roughly 5,000 people worldwide currently has access to internal frontier systems. This tiny group witnesses swarms of thousands of agents collaborating on software megaprojects, running cyber attacks, and discovering new science at speeds that would take human experts years to accomplish.
The rapid acceleration in capability has shifted development timelines dramatically, with OpenAI CEO Sam Altman previously noting that current coding tools can achieve in minutes what once took startups months to build. Karpathy emphasized that while the general public interacts with what feels like a basic chatbot, a small sliver of professionals is experiencing the vertigo of an exponentially rising capability curve. This widening gap matters because it highlights a massive disconnect between public perception of artificial intelligence and the advanced systems already operating behind closed doors.