The Death of ChatGPT Atlas: How the Anthropic Threat Forced OpenAI to Kill Its “Side Quests”

More than nine months after its high-profile launch, OpenAI’s web browser is already being regretted. Announced alongside the rollout of GPT-5.6, the abandonment of ChatGPT Atlas on August 9 marks an admission of failure in the face of Google Chrome, but above all a major strategic pivot. With Anthropic aggressively targeting the enterprise market, Sam Altman and his teams seem to no longer have time for distractions.This is the era of streamlining and the desktop “super-app.”

A flash in the pan named Atlas

Remember: in October 2025, Silicon Valley swore by nothing but reinventing the web browser. It was amid this frenzy that OpenAI unveiled ChatGPT Atlas, a Chromium-based browser designed to merge online search with generative AI.

The ambition was lofty, but reality turned out to be duller. Just days after its release, the browser was already under fire from EFF critics, who accused it of being a privacy nightmare for sucking sensitive personal data, including medical data, into its shared memory. And in our comprehensive test, we highlighted the burdens of an agent mode that was too slow, tab-sorting bugs, and restrictions deemed excessive, such as the blanket blocking of all adult content.

With users rapidly losing interest and Google Chrome’s dominance still intact, the Atlas adventure looks like a sink of technical energy for a modest return on investment.

“Code red” versus Anthropic: no more side quests

Why pull the plug on Atlas so quickly? One explanation put forward comes down to one word: Anthropic. The parent company of Claude is racking up success among professionals with tools such as Claude Code and Cowork. Internally at OpenAI, tensions reportedly rose a notch.Fidji Simo, the company’s head of applications, did not mince his words in a memo reviewed by the Wall Street Journal: the company must act as if it were under “code red” and stop letting itself be distracted by “side quests”.

“We realized that we were spreading our efforts across too many applications and technology stacks, and that we needed to simplify our approach”, Fidji Simo reportedly said internally. In this light, maintaining a full browser in the face of Chrome no longer made strategic sense for the start-up.

The response: the rise of the “super-app” and ChatGPT Work

Rather than spreading its engineers across video (Sora), a web browser (Atlas), a desktop application and a coding tool (Codex), OpenAI would now seek to centralize everything.

Atlas’s browsing and screen-analysis technologies would not disappear entirely: they would be transferred to a new official Chrome extension and integrated directly into the ChatGPT desktop app. The stated goal is to eventually merge the ChatGPT app, the Codex coding tool and autonomous-agent capabilities into a single interface for PC and Mac, the famous “super-app”.

This transition has in fact already materialized this July with the release of ChatGPT Work, an agent capable of manipulating local files, automating repetitive tasks on Slack or Teams, and interacting directly with professional tools. The structure is that of a three-branch pyramid: a unified desktop application that sits atop ChatGPT for the general public, Codex for code, and ChatGPT Work for agentic tasks in business.

A still uncertain bet

By abandoning Atlas, OpenAI is not giving up on its ambitions, but is redirecting its efforts toward what is internally presented as the priority battle of 2026: that of agents capable of working in the user’s place. The question remains as to whether this super-app will be able to deliver the promise of fluidity that the Atlas browser itself never managed to fulfill.

👉🏻 Follow the latest tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.