OpenAI is building desktop “Superapp” to replace all of them

ChatGPT, Codex, and the Atlas browser are merging into one and the internal memo that leaked tells a more urgent story than the press release does.

For the past year, OpenAI has been shipping products at a pace that made even its biggest fans dizzy. A browser here. A coding platform there. A video app. A hardware play. On Thursday, the company essentially admitted that strategy backfired — and announced it’s collapsing everything into one.

The plan, first reported by the Wall Street Journal and quickly confirmed by OpenAI itself: merge the ChatGPT desktop app, the Codex coding platform, and the Atlas web browser into a single unified desktop superapp. The project will be led by Fidji Simo, OpenAI’s CEO of Applications, with President Greg Brockman overseeing the product overhaul and technical restructuring.

On the surface, this reads like a sensible product consolidation one app instead of three, simpler for users, more efficient to build. But the internal memo that accompanied the announcement tells a different story. One about a company that moved too fast, spread itself too thin, and is now scrambling to close the gap on a competitor that’s been quietly eating its lunch in one of its most important markets.

What’s Actually Getting Merged and Why Each Piece Matters

Before getting into the why, it’s worth understanding exactly what’s going into this superapp because the three products being merged are quite different from each other.

Together, these three products cover the full stack of what a power user or developer would want from an AI operating environment: conversation, code, and web. The superapp would let all three work in concert your AI assistant can browse the web through Atlas, write and run code through Codex, and reason through problems via ChatGPT, all within one window, without losing context between switches.

The Internal Memo Nobody Was Supposed to See

Here’s where the story gets more interesting than the official announcement. Alongside the WSJ report, an internal memo from Fidji Simo surfaced and The Verge published the key details. It’s a frank admission of what went wrong inside OpenAI.

“We realized we were spreading our efforts across too many apps and stacks, and that we need to simplify our efforts. That fragmentation has been slowing us down and making it harder to hit the quality bar we want.” Fidji Simo, CEO of Applications, OpenAI · Internal Memo, March 19, 2026

That’s a striking thing for any executive to say publicly, let alone the head of product at the world’s most valuable AI company. The memo doesn’t sugarcoat it: OpenAI launched too many things too fast, quality suffered, and fragmentation made it harder for any single product to reach its potential. The superapp isn’t just a UX improvement it’s a course correction.

Simo also warned employees to avoid being “distracted by side quests” a line that reads like a direct reference to some of OpenAI’s more experimental 2025 ventures, including its Sora video app. The message was clear: double down on what’s working, cut everything else.

The Real Reason for the Urgency: Claude Code

The official framing is about simplification and user experience. But the competitive driver is harder to miss: Anthropic’s Claude Code. The terminal-based agentic coding tool has been gaining developer traction at a pace that reportedly alarmed OpenAI leadership. CNBC confirmed that the speed of Claude Code’s adoption internally triggered what was described as a “code red” response at OpenAI.

Codex OpenAI’s answer to the same use case has been well-received but hasn’t dominated the developer market the way ChatGPT dominated the consumer AI market. That context reframes Simo’s memo significantly. When she says the company needs to “double down” on Codex, she’s talking about winning a competitive war for the developer ecosystem before Anthropic locks it up. The superapp is, in part, a weapon in that fight.

⚡ What Makes the Superapp Different From Just “Three Apps in One Window”

• Cross-product context: The AI retains memory and context across all three tools. Start a conversation in ChatGPT, continue it in Codex, reference a webpage from Atlas all in one coherent thread.
• Agentic orchestration: The AI can autonomously call on any of the three capabilities to complete a task. “Build me a web scraper and run it on this site” triggers both Codex and Atlas without manual switching.
• Unified account and billing: One subscription, one interface a major win for enterprise customers especially.
• Developer-first power surface: Combining a coding agent and a browser into the same app as ChatGPT creates something that can genuinely compete with IDEs and browser-based dev environments.
• Faster iteration: Consolidating three codebases means bug fixes and new features ship to all users simultaneously no more version fragmentation between products.

The Microsoft Problem Nobody’s Talking About

There’s an elephant in the room that the OpenAI press release carefully stepped around: Microsoft. The company has invested over $13 billion in OpenAI and runs its own deeply integrated AI products Copilot, GitHub Copilot, and an AI-enhanced Edge browser. A unified OpenAI desktop superapp that combines an AI assistant, a coding agent, and a web browser is, on paper, a direct competitor to all three.

OpenAI has historically been careful to position its products as complementary to Microsoft’s rather than competitive. A powerful, standalone desktop superapp makes that positioning harder to maintain and how both companies navigate that tension will be one of the more interesting subplots to watch as the superapp rolls out.

What It Looks Like in Practice For Real Users

Specifics on the UI and feature set haven’t been officially released, but based on the three products being merged, here’s the most likely experience: a single desktop application that feels like a cross between a browser, a code editor, and a chat interface with one AI thread running through all of it.

You could ask ChatGPT to research a topic, watch Atlas browse and pull sources in a split panel, then ask Codex to turn the findings into a structured report or data pipeline all within one window, the AI managing the handoffs automatically. For developers especially, this eliminates the constant context-switching between tools that makes current AI-assisted workflows feel clunky. XDA Developers noted that this kind of unified surface is exactly what power users have been asking for since ChatGPT launched.

The Bigger Trend: Every AI Company Is Consolidating Right Now

OpenAI isn’t doing this in a vacuum. Across the AI industry, a pattern is emerging: companies that spent 2024 and early 2025 shipping as many products as possible are now pulling back and consolidating around their strongest bets. Google has been folding Gemini into its entire product surface Search, Workspace, Android rather than running it as a standalone app. Anthropic has gone narrow and deep, focusing on Claude’s API and Claude Code rather than building a wide consumer product portfolio.

The message from the market is consistent: users don’t want more AI apps. They want one AI surface that actually does everything well. OpenAI’s superapp bet is the company’s answer to that reality. And if you want to follow along as the product takes shape, OpenAI’s official site is where the team is expected to post all major updates. Whether they can ship it fast enough and at high enough quality to matter in a market moving this quickly is the real question. Given the frank admission that fragmentation already cost them once, the pressure is on to prove they’ve actually learned the lesson.

Mas

An AI researcher who spends time testing new tools, models, and emerging trends to see what actually works.