A few months ago, I started working on this program called Lumos Fellows, where we help high schoolers and university students learn how to build and grow apps using AI coding tools.

At the beginning, we weren’t sure there was even a real need for this, because it just seemed so obvious to us that if we were in high school right now and interested in coding, we’d be using AI coding tools like Cursor, Claude Code, Codex, etc.

Before we really started the program, we interviewed 32 high schoolers and asked them a bunch of questions, including whether they had heard of AI coding tools like Cursor. Also, for context, these weren’t just random high schoolers. They were already pre-selected from an application pool of 200+ students. These were students we were already impressed by, either through their test scores or extracurriculars. Almost all of them already had some form of coding experience, such as taking AP Computer Science.

Internally, we guessed how many had used a coding tool like Cursor before. We ended up converging around 30 to 40%.

When we finished our last interview, only one person had used it, and only two had heard of it at all. One of the two was already in college at Berkeley, and the other was someone we had known in real life and had invited to apply because we knew they had already tried building apps before. So realistically, the percentage of high schoolers who had used something like Cursor was 1/32, or about 3.1%.¹

Since we did this in January 2026, these numbers have likely changed. It was a clear indicator that we’re still very early in the adoption curve.² We wouldn’t be surprised if we were to run this again next January and see a number close to 50% or even more. “The future is already here — it’s just not evenly distributed.”