OpenAI's Codex triples to 15 million users in two months — AI coding is now mainstream, not early-adopter
OpenAI's Codex coding agent crossed 15 million active users by 13 August 2026, up from around 5 million weekly active users in June, with OpenAI reporting that non-developer 'knowledge workers' are now adopting it faster than engineers — a search and adoption curve that mirrors what we're seeing across Claude Code, Cursor and GitHub Copilot.
18 August 2026
OpenAI’s Codex coding agent passed 15 million active users by 13 August 2026, according to OpenAI’s Tibo Sottiaux, up from roughly 5 million weekly active users in June — a tripling in under two months, largely credited to the launch of its desktop app. The detail worth pausing on isn’t the headline number, it’s the composition: OpenAI has said knowledge workers who aren’t professional developers made up around 20% of Codex’s user base in June and were adopting the tool faster than engineers were. This isn’t just more developers picking up another coding agent. It’s a widening pool of product managers, founders and operators who are now comfortable enough with an AI coding tool to use one directly.
Codex isn’t alone in this curve. Claude Code, Cursor and GitHub Copilot have all published similar adoption and revenue growth numbers through 2026, and JetBrains’ most recent developer survey had all three tools registering meaningful usage for the first time. What’s notable is the speed: tools that were niche a year ago are now the default way a growing share of technical and non-technical people expect to interact with code at all.
So what
If you’re commissioning software and haven’t already asked your development partner how they use AI coding tools day to day, this is the moment to ask. The gap that matters isn’t “do they use Claude Code or Codex or Cursor” — it’s whether AI-assisted output goes through the same review, testing and architectural discipline a senior engineer would apply anyway, or whether speed has quietly become the only metric that counts. Adoption at this scale means AI-assisted development is now the default, not the exception; the differentiator is what a team does with the code once it’s generated. That’s the gap we build for — see our AI-assisted development approach, or get in touch if you want software built with AI tools used properly rather than just quickly.