OpenAI is building a desktop superapp

Tech companies always seem to follow the exact same arc: start with one focused tool, get wildly popular, and eventually decide that everything needs to live inside a single desktop app.

OpenAI is currently developing a new desktop "superapp" that combines ChatGPT, its Codex AI coding tool, and its upcoming Atlas AI browser into one single piece of software, according to a report from The Wall Street Journal.

The shift comes as OpenAI tries to rein in product bloat. In an internal memo obtained by the WSJ, OpenAI Applications CEO Fidji Simo admitted that fragmentation has been slowing the company down and making it harder to hit their quality bar.

Killing the side quests

OpenAI spent much of the past year making flashy headlines, from unveiling the Sora video tool to acquiring Jony Ive's AI hardware startup. But the constant expansion came at a cost.

Simo told employees last week that the team needs to stop being distracted by side quests. Management has been actively looking at what to deprioritize to refocus resources on what actually moves the needle.

Simo corroborated the direction on X, writing that companies naturally cycle between exploration and refocusing. When new bets start working—like Codex is now—it is critical to double down and cut distractions, she noted.

For now, the overhaul is strictly focused on the desktop experience. The mobile ChatGPT app will remain unchanged, while OpenAI spokesperson Lindsey Held declined to comment on the report.

Why it matters

Anthropic's recent momentum with Claude Code has clearly rattled OpenAI. Coders and power users have been raving about Anthropic's developer experience, proving that frontier model strength alone is not enough if your user interface is fractured across half a dozen tabs.

For developers and office workers, having a unified workspace makes a lot of sense on paper. You would no longer have to bounce between an Atlas browser tab for research, a desktop chat window for troubleshooting, and a standalone code editor. Keeping the context intact could save serious time.

However, building a desktop superapp is historically a dangerous game. Desktop users notoriously value speed, responsiveness, and minimal system overhead. Shoving a full browser engine and an agentic coding environment into one client risks turning a snappy chat app into an electron-heavy resource hog.

We think this pivot signals the end of the AI demo era. Frontier labs are now entering a grinding platform war, fighting not just for model benchmarks, but to become the operating layer of your daily screen time.