In the tech world, there is a recurring fever dream that haunts CEOs who have run out of ways to make their products actually work better: the “Superapp.” OpenAI, a company that started with the noble goal of ensuring AGI benefits all of humanity, has apparently decided that “all of humanity” is actually just a group of people who are tired of having more than one icon on their taskbar.

According to a memo from Fidji Simo, OpenAI’s CEO of Applications—a title that sounds like it was generated by a prompt for “Generic Corporate Overlord”—the company is merging ChatGPT, the Codex coding tool, and their new Atlas browser into one monolithic desktop entity. Because if there’s one thing the world is crying out for, it’s a browser that tries to write your code while simultaneously hallucinating your grocery list.

The primary claim here is that “fragmentation” is the villain slowing OpenAI down. Simo argues that having different apps for different functions is making it harder to “hit the quality bar.” It’s a bold move to blame the UI for quality issues when your flagship model occasionally insists that the letter ‘r’ doesn’t exist in the word “strawberry.” It’s not the fragmentation of the software that’s the problem, Fidji; it’s the fragmentation of the logic.

Let’s look at the “Atlas” browser. We are living in an era where Google Chrome effectively functions as a RAM-eating parasite. OpenAI’s solution? Build a browser powered by an LLM. If you thought your laptop fan sounded like a jet engine before, just wait until your browser is performing trillion-parameter matrix multiplications just to help you find a recipe for sourdough. We’re moving from “Software as a Service” to “Software as a Space Heater.”

The assumption that users want their IDE, their browser, and their chatbot under one roof is the kind of “synergy” talk that usually precedes a massive layoffs announcement. Professional developers use VS Code because it works, not because they’re waiting for it to be merged with a browser that interrupts their debugging to suggest a poem about Python. By merging Codex into this “superapp,” OpenAI is essentially trying to turn a specialized scalpel into a Swiss Army knife where every blade is slightly blunt and occasionally gives you incorrect legal advice.

While Anthropic is busy making Claude 3.5 Sonnet actually capable of reasoning, OpenAI seems to be pivoting into a project management phase. They’ve gone from “We are building God” to “We are building Microsoft Office, but with more steps and higher electricity bills.”

The move to hire Jony Ive for hardware and focus on desktop “superapps” suggests a company that is more interested in the *container* than the *content*. It’s a classic move: when the underlying AI progress hits a plateau (or a “quality bar” you can’t seem to clear), you start redecorating the house.

OpenAI claims this will “simplify” their product efforts. In reality, it simplifies the user’s journey toward uninstalling. We don’t need a superapp; we need a chatbot that doesn’t forget the first half of the prompt by the time it reaches the second. But sure, give us a browser-coding-chat-hybrid. At the very least, when it crashes, it’ll be a one-stop shop for total system failure. Efficiency at its finest.