A month ago, OpenAI handed one engineer the keys to a product used by close to a billion people every week. His name is Thibault Sottiaux, and he now runs both ChatGPT and Codex as the company's head of core products. The mandate would make most people sweat: fuse those two products into a single thing OpenAI keeps calling, with a straight face, a "super app."
Sottiaux's own take was refreshingly blunt. He called the job exciting and "mildly terrifying" in the same breath, speaking to WIRED earlier this week. Which is about the only sane reaction to inheriting a consumer product that size.
What an OpenAI super app actually means
Let's get the buzzword out of the way first. In Asia, "super app" means something like WeChat: messaging, payments, shopping, and a dozen other services tucked behind one icon. OpenAI is reaching for something stranger and bigger.
Sottiaux frames the target as the world's best personal agent, one that genuinely understands what people care about. Over the coming year, he says, ChatGPT should grow "delightfully proactive," surfacing the right thing at the right moment instead of waiting for you to type a prompt.
Picture an assistant with a long memory. It books your dinner reservation. Later, it flags that the restaurant's special contains the ingredient that wrecked your stomach last time. It files your expense report before the deadline catches you. You ask in plain language, and the machinery underneath (code being written, an API getting called, the web being searched) stays hidden.
That's the pitch, anyway. Whether any of it ships on schedule is a separate question.
The technical core of all this is Codex, OpenAI's coding tool. According to Sottiaux, the heavy lifting comes in two stages: first stretch Codex beyond coding into something that can handle general tasks, then tuck it inside ChatGPT itself. Codex has been quietly winning over nontechnical users too, which is part of why the company thinks the bet is sound.
The clearing of the decks
Here's the part that tells you how serious OpenAI is about this. To make room for the super app, the company has been killing products.
Its video app Sora got shut down. So did a research platform aimed at scientists. Several executives who ran those teams have since walked out the door. Sottiaux, meanwhile, kept rising. The timing is hard to ignore: as other bets folded, his project soaked up resources and his standing climbed.
His reporting line runs directly to Greg Brockman, who has taken charge of product across OpenAI during the medical leave of Fidji Simo, the company's CEO of applications. For a product this consequential, that's a strikingly short chain of command.
And yet the team itself stays small. Sottiaux wouldn't say how many people are building the super app today, but his Codex group numbered around 40 a couple of months back. Forty people, give or take, reshaping a product with hundreds of millions of weekly users. The math is its own kind of statement.
From Belgium to AlphaGo to San Francisco
Sottiaux didn't arrive out of nowhere. He grew up in Belgium, studied applied mathematics, and landed at Google's London office in 2015. He started on Google Maps, then moved to DeepMind, where he built the infrastructure and tools researchers leaned on for projects like AlphaGo.
That's the system that beat a human Go champion in 2016, the moment a lot of people sat up and took the field seriously. Sottiaux helped build the scaffolding behind it.
When ChatGPT launched in late 2022, something clicked. He has said DeepMind had been sitting on similar capabilities for nearly two years without shipping them. The frustration shows in how he tells it. He moved to San Francisco and joined OpenAI in 2024, first building tools for the company's own researchers, the same kind of work he'd done before.
Within months, he started on what became Codex. The tool took off. Sottiaux turned into a familiar face in developer circles, answering bug reports personally on X and sometimes resetting engineers' weekly token limits when they begged. A minor folk hero, basically.
A different audience now
Here's the catch. Codex meant serving developers and AI researchers, people who think the way he thinks. The super app means designing for everyone else. Your aunt. The small-business owner who's never opened a terminal. Someone who wants dinner booked, not a lecture on agentic architecture.
That's a genuinely different muscle. Building tools an engineer loves is one thing. Building something a billion non-engineers reach for daily is another, and the skills don't always transfer.
Why OpenAI is making this bet now
Follow the pressure. OpenAI is racing toward an IPO and fending off Google and Anthropic, both of whom keep shipping competitive models. Growth needs a new engine, and the company is betting the super app is it.
The logic goes like this: if one personalized assistant can handle your personal life, your work, and everything in between, OpenAI reclaims its spot as the obvious leader across consumer and enterprise at once. A single product to rule them all.
Bold thesis. Also not OpenAI's first crack at it.
Last year the company shipped Operator, a feature inside ChatGPT meant to browse the web for you and finish tasks. It later folded into something called ChatGPT Agent. Neither caught on in any meaningful way. Sottiaux admits those early swings fell short, which at least suggests the team isn't pretending the past didn't happen.
So the question hanging over everything is simple. What separates the super app from Operator beyond ambition and a better name? Sottiaux's answer leans on Codex being a stronger foundation and on the agent becoming proactive rather than reactive. Maybe that's enough. Maybe "delightfully proactive" (his phrase, not mine) turns out to be the feature people didn't know they wanted.
Or maybe it's the thing they find creepy. An assistant that pings you unprompted, that remembers your stomach trouble and your expense deadlines, sits on a thin line between helpful and intrusive. OpenAI will have to get the tone exactly right, and tone is harder to engineer than code.
What to watch
The tells will come over the next year. Watch whether OpenAI keeps shutting down standalone products to feed the super app, and whether more executives leave as it consolidates. Watch whether Codex's general-purpose version actually ships, and whether ordinary users take to it the way developers took to Codex.
And watch Sottiaux. He's gone from building researcher tools to owning the consumer product that defines OpenAI in most people's minds. If the super app works, he's the engineer who pulled off the company's biggest pivot. If it doesn't, that small team and those shuttered products start to look like a very expensive gamble.
Either way, a billion people are about to find out what OpenAI thinks they want from a chatbot. They didn't get a vote.