One Person Now Runs Almost Everything OpenAI Sells
OpenAI has put one person, Thibault Sottiaux, in charge of every core product the company offers: the API (the service developers use to build apps with OpenAI's AI), agent infrastructure (tools for building AI agents — programs that can carry out multi-step tasks on their own), the enterprise business, the full ChatGPT suite including ChatGPT Work and the original ChatGPT, and Codex, the company's AI coding assistant. He reports directly to OpenAI co-founder Greg Brockman (TechCrunch).
This role was formalized through an internal reshuffling this past May. Brockman took over product leadership and chose Sottiaux to run the core product and platform teams (Wired). Before that, Sottiaux was the engineering lead for Codex (Wired; YouTube). Among Codex users, he is known less by his job title and more by what he does: the person who resets their usage limits every time the product grows enough to hit a cap (TechCrunch).
That detail may sound minor, but it reveals something important. Sottiaux's style of product management is very hands-on. Think of it like a restaurant manager who personally adjusts how many tables are available when the dinner rush keeps getting bigger — rather than just hoping the kitchen can keep up. Codex's growth depends not just on how smart the AI model is, but on practical questions like how many requests the system can handle at once, how much each user is allowed to do, and whether the infrastructure can keep developers from walking away.
ChatGPT Work, another product in Sottiaux's portfolio, is a platform designed for office workers to use AI agents. It launched as part of OpenAI's Plus plan at $20 per month (TechCrunch). That decision is notable: instead of putting the agent features behind a more expensive plan aimed at big companies, OpenAI included them in the same subscription that regular individuals already pay for. For a solo office worker, that means AI agent tools cost the same as standard ChatGPT Plus.
Sottiaux's portfolio covers a wide range — developer tools, agent platforms, and both everyday and enterprise products. Normally these areas are managed separately, but having one leader across all of them means decisions are coordinated. For example, if the API team changes its usage limits, that directly affects Codex users. And the agent features in ChatGPT Work run on the same underlying systems. One product leader means these trade-offs are made together rather than in isolation.
The Codex usage-limit resets are the clearest sign of how this works in practice. Sottiaux stepping in at growth milestones suggests demand is growing faster than OpenAI's pre-set capacity, requiring someone to manually adjust it. This is a familiar pattern when a popular product grows faster than expected. The question for developers is whether OpenAI can eventually scale ahead of demand instead of reacting to it.
Putting ChatGPT Work inside the existing Plus plan, rather than creating a separate and pricier tier, tells you something about how OpenAI wants to sell agent features. By treating them as a standard part of the subscription, the company is making it easy for individual office workers to try them. Whether that approach holds up as more people use the tools and computing costs rise is an open question, but for now it lowers the barrier for the people most likely to benefit.
For developers, having one person oversee both the API and Codex means decisions about pricing, access, and usage limits are made with full awareness of how they affect both the raw developer tools and the higher-level coding product. As more of what developers build relies on agent infrastructure, that consistency matters. Conflicting rules across those layers would cause friction, and unified leadership reduces that risk.
In my view, concentrating this much authority in one person who reports to a co-founder is worth paying attention to. It puts decision-making over OpenAI's commercial products in very few hands, which can speed things up and keep the products consistent — but it also means that if a wrong call is made, it can ripple across the entire product line. The same tight connection that enables fast decisions becomes the thing that spreads mistakes.
Sottiaux's rise from running Codex's engineering to leading all core products has been visible in real time through those usage-limit resets that Codex users experience. The credibility he earned in that role now extends across a portfolio where infrastructure, capacity, and product design meet every day. For a company whose tools are increasingly central to how people work and build things, having all of that under one leader is one of the more important organizational facts in the technology landscape today.
