You open your laptop. You need AI help. You launch ChatGPT for a question. Switch to Codex to write code. Open Atlas to browse something. Three apps, three logins, three separate contexts that know nothing about each other. Congratulations — you're using OpenAI's product lineup exactly as designed, and it's terrible.

OpenAI knows this too.

On March 19, 2026, The Wall Street Journal published an internal memo from Fidji Simo, OpenAI's CEO of Applications. The diagnosis was blunt: OpenAI has too many apps, and the fragmentation is killing quality. The prescription: merge ChatGPT, Codex, and Atlas into a single desktop "superapp."

Simo didn't sugarcoat it: "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."

Translation: we built three separate things and they don't talk to each other. Oops.

What's Actually Merging

ChatGPT — the conversational AI interface that 300 million people already use. The front door.

Codex — OpenAI's AI coding agent (not just autocomplete — an autonomous program that writes, tests, and debugs code on its own). Over 2 million weekly active users, 3x growth since January 2026. On March 19 — the same day as Simo's memo — OpenAI announced its acquisition of Astral, the startup behind Python's beloved uv and Ruff tools, to bolt their Rust-based tooling directly into Codex. That means dependency management (keeping track of all the code libraries a project needs), linting (catching code style errors automatically), and type-checking (verifying that data flows correctly through code) — all running natively inside the agent instead of relying on separate external programs.

Atlas — OpenAI's AI browser, launched on macOS in late 2025. Built on Chromium (the open-source engine behind Chrome), it embeds ChatGPT into your browsing. The AI sees what you see, answers questions about the page, and can take actions on websites through "agent mode" — meaning the AI clicks buttons, fills forms, and navigates for you.

The combined promise: an agent that researches a topic in Atlas, discusses findings in ChatGPT, and implements the solution in Codex — without switching apps or losing context.

Desktop only. Mobile ChatGPT stays separate. This targets professionals who actually pay.

Why Now

This isn't spring cleaning. It's survival math.

Anthropic's revenue reportedly approached $20 billion annualized by early March 2026. Their share of enterprise AI spending — money companies pay for AI tools — climbed to 40%, while OpenAI's fell from roughly 50% to 27%. Claude Code, Anthropic's terminal-based coding agent (a tool you run in your computer's command line, no graphical interface needed), has steadily eaten into Codex's market share among developers. Google embedded Gemini into everything from Android to Workspace.

Three separate products can't compete against integrated offerings.

There's also the IPO (initial public offering — when a company starts selling shares on the stock market). On February 27, OpenAI closed a $110 billion funding round at a $730 billion valuation — the largest private raise in history. Five days after Simo's memo, on March 24, OpenAI killed Sora (their video generation tool — burning $15 million a day in compute against $2.1 million in lifetime revenue) and confirmed it completed pretraining on its next model codenamed "Spud" (the initial phase of teaching an AI model by feeding it massive amounts of data). A Q4 2026 IPO looms. "We have a platform" sells much better to investors than "we have a chatbot and some side projects."

The Price Tag

OpenAI's own research shows a 6x productivity gap between their most engaged enterprise users and the average. The superapp targets that gap by killing context-switching friction.

But I've watched this movie before. Microsoft called it Office. Google called it Workspace. The "put everything in one app" strategy works brilliantly until the app turns bloated and slow. Codex needs to ship fast. Atlas needs to ship fast. ChatGPT needs to stay responsive. Bundle them together, and they all move at the speed of the slowest team.

Then there's lock-in. A superapp that handles your chat, your code, and your web browsing is the WeChat strategy applied to AI — once your entire workflow lives inside one surface, switching costs become enormous. OpenAI isn't just fighting for users. They're fighting for default position on every knowledge worker's desktop.

Back to Your Laptop

The Astral acquisition is the most honest signal here. Buying the team behind tools the Python community genuinely loves — not just impressive benchmarks — shows OpenAI cares about developer experience. That matters.

Will the superapp work? Probably. Context-switching between apps genuinely is the biggest friction in AI-assisted work today. But OpenAI isn't competing with "no AI" anymore. They're competing with Claude Code running in a terminal, Cursor embedding everything in one editor, Gemini running across every Google surface.

The bar isn't "better than nothing." The bar is "better than the best integrated alternative." And that bar moves every single week.