rewrite this title OpenAI Wants to Kill the Chatbot It Invented and Turn It Into a Superapp – Decrypt

OpenAI has nearly 1 billion ChatGPT users. Most of them use it for free. That’s the problem the company is trying to solve before it goes public.

The plan, reported by the Financial Times based on interviews with more than a dozen current and former employees, is to turn ChatGPT into a superapp—a single interface for coding, task automation, image generation, and partner apps. The redesign is internally codenamed “Aria.”

OpenAI has not confirmed the codename or a hard launch date. Changes to the website and mobile apps are expected to begin rolling out in the coming weeks.

Thibault Sottiaux, who previously ran Codex and now leads all of OpenAI’s core product and platform, told the FT the goal is an assistant “capable of helping you across everything in your life, be it personally or at work.”



“Chat is dead,” a senior OpenAI employee told the Financial Times—not exactly what you’d expect to hear from the company that built its empire on a product called ChatGPT

WeChat did it. Everyone else tried.

The superapp idea has a model: WeChat, Tencent’s everything-app, which started as a messaging platform in 2011 and now handles 45 billion messages a day for 1.4 billion monthly users. WeChat Pay and Alipay process $40 trillion in annual transactions . Chinese users book doctors, pay taxes, and hail rides without ever leaving the app. It stopped being a social network a long time ago and turned into software infrastructure in its own right.

Elon Musk has been chasing that model since he bought Twitter in 2022. He’s said publicly he wants X to be the Western WeChat. In 2025, he launched XChat encrypted messaging and X Money payments, the latter through a Visa partnership for peer-to-peer transfers. X has somewhere between 500 and 600 million users. The payments ambition hasn’t taken off. Meta tried it with Messenger bots in 2016 and WhatsApp payments for years. It also didn’t stick.

OpenAI’s pitch is different. It’s not trying to build a social network—arguably a failed experiment that started and ended with Sora. Instead, it’s starting from a platform people already use for work—and making that work platform do more work.

The Codex coding agent can write, debug, and deploy software from a text prompt. The ChatGPT Agent, launched in July 2025, can browse the web, run apps, and place orders autonomously. Workspace agents, released in April 2026, run persistent multi-step workflows inside Slack and other enterprise tools.

That’s the stack OpenAI is consolidating into one interface.

The IPO math

OpenAI’s annualized revenue has passed $20 billion. Business customers account for 40% of it—and the majority of Codex users pay, unlike the majority of ChatGPT users who don’t. Jenny Xiao, a Leonis Capital partner and former OpenAI researcher, told the FT that a year ago OpenAI was “swing for the fences” while Anthropic was “make money first.” Now, she said, “the two are converging, because both of them are trying to aim for an IPO and investors care more about money than dreams.”

Anthropic is the direct pressure. Claude Code crossed $2.5 billion in annualized revenue by early 2026—faster than any enterprise software product before it—and enterprise now accounts for roughly 80% of Anthropic’s total sales. Also, they seem to have beaten OpenAI in the IPO game, officially initiating its initial public offering process by confidentially submitting a draft Form S-1 with the SEC last week. OpenAI is now playing catch up, filing confidentially for its own IPO today.

Alex Embiricos, OpenAI’s head of enterprise product, told the FT where he thinks this ends: “When we have AGI, I don’t think there will be a large number of distinct brands. Probably there will be a single entity that I can talk to that can do whatever I need.” OpenAI is also acquiring Python toolmaker Astral to fold developer tools directly into Codex. The IPO is targeting Q4 2026.