Six Months Behind. OpenAI Copied Cowork and Called It ChatGPT Work.

OpenAI unveiled ChatGPT Work on July 9, 2026, an agent that stops answering questions and starts delivering finished work: spreadsheets, presentations, documents, reports, even web apps. It pulls context from your apps, files and workflows, breaks a goal into steps, and grinds away at the task alone for hours. It runs on GPT-5.6 and its Sol, Terra and Luna family.

If this sounds familiar, it's no coincidence. It's exactly what Anthropic put on the table with Claude Cowork on January 12, 2026, six months earlier. Desktop first, then Windows, general availability in April, and last week a jump to web and mobile. While Anthropic iterated, OpenAI watched.

Same script, half a year later

ChatGPT Work mirrors Cowork almost point for point: agent instead of chatbot, results instead of answers, connected apps instead of isolated conversation, access to files, plugins and skills. The international press has already called it what it is, a move from whoever's playing catch-up. Bloomberg wrote that the launch arrives months after Anthropic pushed out Cowork.

On X, the reaction was even blunter. Among the first comments was a line that sums it all up: this feels exactly like Anthropic's Chat/Cowork/Code division, and it's the worst version of it.

Where OpenAI bites back

The move isn't technical, it's commercial. OpenAI attacks on two fronts: price and distribution. GPT-5.6 cuts cost to half or less of Fable 5 across several tiers, and ChatGPT's installed base, hundreds of millions strong, is an army Cowork doesn't have yet. The idea may have been born at Anthropic, but whoever gets it in front of more people might be the one who arrived later.

Let's be clear: in raw quality, Fable 5 still leads on what matters for serious engineering. What OpenAI brings isn't superiority, it's scale.

Next in line: Claude Design

And here's what you should be thinking. If OpenAI took six months to copy Cowork, how long until it copies Anthropic's next weapon? Claude Design, out of Anthropic Labs, turns prompts into prototypes, slides and one-pagers, reads your code to build a design system, and hands off straight to Claude Code. It's a shot across Figma's bow and closes the loop from prompt to production inside Anthropic's ecosystem.

At the current pace, the math is simple and uncomfortable for OpenAI: if the pattern holds, a ChatGPT-branded design-tool clone should show up sometime in early 2027. Not a prediction, a cadence reading. Anthropic invents, OpenAI chases, and the gap has been half a year.

The agent race stopped being about who has the smartest model. It's about who defines the next work format before the other does. For now, Anthropic is writing the script. OpenAI is memorizing the lines.