Andrej, a person who coined the term vibe coding, ex OpenAI currently working on Anthropic, published a series of posts this year, one in April and another one in October 2026, about why different people treat progress in AI differently. And so what he has done, he has basically created a funnel or described the funnel of 8,300,000,000 people living on the planet Earth. So out of this 8,300,000,000, around 6,000,000,000 have almost never interacted with AI. They do not have chats with ChatGPT, with Claude, or with any other chat providers.
Potentially, used the AI products without even knowing that large language models or other general AI tools are inside. Then there is another one or 2,000,000,000 people which have used free tiers of the models. And they have tried to use it maybe one year ago, maybe two years ago. They interacted with it, and mostly they use it as a kind of advanced Google search. They're not mostly impressed around this technology, think that it's nice, but it's not that powerful, or they cannot yet use it for their own tasks.
Then there is other group of people. I think I'm inside this other group of people. It's around 20,000,000 people. These 20,000,000 people are paying for subscriptions to the frontier models. They have tested Claude Code, tested Codex from OpenAI and other powerful frontier models. And they know that the progress and especially progress in programming, in mathematics, and these type of tasks is amazing. And especially in 2026 or in the second part of 2026, it's something unbelievable. That's my own experience. And then the final group is around 5,000 people who are working in laboratories like Anthropic, like Google, OpenAI.
And these ones, these people, they are having access to the frontier models and to the models which are not yet published. So they see this exponential growth inside, maybe three to six months before other people in the outside world can see it. And Andrej is one of these people. So this funnel is very interesting, and it's actually first time I see something like that, which is just a napkin math. It's an amazing explanation why there is so different views on what AI is, what AI can do and cannot do.
