The market for software is roughly a hundred times larger than the market for people who build it — an industry framing that the last month turned into a usage number for the first time. Inside Codex, the fastest-growing cohort is not the developers the tool was built for, and that mix is the only data point that explains why OpenAI bothered to count at all.
Latent Space's write-up of Akshay Nathan's appearance is careful to put the combined-user figure in context: two weeks for ChatGPT Work and Codex to clear ten million users combined, but the load-bearing detail is who those users turn out to be. Knowledge workers sit at roughly a fifth of the Codex base, and that slice is growing more than three times as fast as the developer slice it was launched for. The product Nathan is describing started as a coding tool and ended as a productivity surface, and the gap between the two is the gap between a beachhead and a market.
The reusable mechanism is the developer-as-Trojan-horse: ship the agent to the people most willing to debug it, then let it walk into the much larger cohort of people who just want the work done. Whoever prices, packages, and absorbs the security review for that next will own the productivity layer of the next operating system. OpenAI's bet is that the developer install was a free option on that prize.
Reported by Sky for Type0, from Codex from 0 to 10M Users: Building ChatGPT Work — Akshay Nathan, OpenAI. Read the original: latent.space