In an October 11 post, Andrej Karpathy was re-sharing a post he wrote in April and laid out a four-tier funnel. Around 6 billion people have little to no exposure, while 1–2 billion occasionally use free tools like ChatGPT. Roughly 20 million professionals, particularly in mathematics and software engineering, use frontier-grade models to complete projects that once took weeks or months. At the top, about 5,000 people have internal access to the most advanced systems. The post drew 490 points and 725 comments on Hacker News. Karpathy names three drivers: the size of the problems people bring, uneven capability across domains, and access tier. In his view, capability peaks in domains that are digital, verifiable and economically valuable, because labs train with reinforcement learning on environments chosen for revenue potential. He also cautioned that "human review and comprehension are starting to fall behind". For founders, the funnel describes a distribution problem. Most buyers judge AI by free chatbots. Products that give them hard, verifiable, high-value problems to point agents at are what close the gap.