AI & Innovation

John Langford quits Microsoft, says AI Frontiers research group is over

By Anika SharmaAI & Innovations Editor, Peepals Tech
WhatsAppX
Representative image: computer circuit board

Representative image: Bru-nO / Pixabay

Researcher John Langford has left Microsoft and, in a 6 November 2026 post, declared its AI Frontiers group over after a move into Microsoft AI.

What Happened

John Langford, the machine learning researcher who co-founded Microsoft's AI Frontiers research group, has quit Microsoft and says the group is finished. He made the announcement in a blog post titled "The End of AI Frontiers", dated Friday, 6 November 2026, on his long-running blog Machine Learning (Theory). Langford says the roughly 50-person group was moved out of Microsoft Research and into Microsoft AI in early September, and that inside the new division he could not get his research used, or even seriously considered. Microsoft has not publicly responded to his criticism.

John Langford quits Microsoft: key facts

  • Langford published "The End of AI Frontiers" on Friday, 6 November 2026, and said he has left Microsoft.
  • Microsoft's AI Frontiers lab page carries an update dated Monday, 14 September 2026 that reads: "Microsoft Research AI Frontiers is now part of Microsoft AI."
  • Reports put the size of the group at about 50 people.
  • In his post, Langford describes the core philosophy of Microsoft AI as scaling, "and aggressively not research".
  • He points to a variant of the transformer design that, by his account, matched a model twice its size, which he projects could halve the cost of training and running a model.
  • The lab page still lists Ece Kamar as "CVP and Lab Director of AI Frontiers"; Microsoft has issued no closure notice.
  • Langford said there would be no internships with the group's researchers in the coming year, according to reports of his post.

John Langford quits Microsoft: why it matters

AI Frontiers was one of Microsoft's best-known research teams in the current wave of artificial intelligence. According to its own lab page, its work included AgentInstruct, the Phi family of small language models, the Fara family of computer-use models (AI that operates a computer screen the way a person does), AutoGen, Magentic-One, MagenticUI and Magentic Marketplace. AutoGen is an open-source framework that lets developers build "agents", AI programs that plan and carry out several steps on their own, and it became one of the group's best-known public projects. The lab page describes the group's focus as "efficient language models and agentic systems".

The dispute Langford describes is between two ways of making AI better. "Scaling" means building bigger models, training them on more data and for longer, which costs more computing power. Research of the kind Langford describes looks for new designs that get the same result with less. A transformer is the basic architecture behind most of today's large language models. Langford's argument is that a design which matches a model twice as large could cut costs sharply, and that Microsoft AI did not take it up. One technology site that covered his post noted that this is his own projection, and that matching a bigger model on benchmark tests does not automatically halve the bill.

Langford built his reputation on contextual bandits, methods that help a system learn which choice is best when it only sees feedback on the choice it made. Technology sites covering his exit say he spent more than 14 years at Microsoft Research. Writing on X, he said: "There's a bit of a train wreck here that's painful to express. Nevertheless, it's a significant event for me and quite a few other folks involved so I wanted to memorialize a bit."

Peepals timeline: The September update on the lab page and Langford's post are 53 days apart (16 days left in September, 31 in October and 6 in November: 16 + 31 + 6 = 53), and the search for his exit trended in India on 14 November, eight days after the post (14 - 6 = 8). Our calculation from the reported figures.

EventDateWhat it says
Lab page updateMonday, 14 September 2026AI Frontiers "is now part of Microsoft AI"
Langford's blog postFriday, 6 November 2026"The End of AI Frontiers": he has left
Search trend14 November 2026"John Langford quits Microsoft" trends

Impact

Short-term: Developers who rely on tools that came out of the group, such as AutoGen, will be watching for any word from Microsoft on who maintains them. Students hoping for internships with the group's researchers next year should note Langford's statement that there will be none.

Long-term: The episode adds to a wider debate in the industry over whether money should go mainly into ever-larger models or into research on cheaper, more efficient designs.

Who is affected: The roughly 50 researchers of AI Frontiers, developers using the group's open-source tools, students seeking AI research internships.

Key Takeaway

John Langford has left Microsoft and declared its AI Frontiers group over, saying Microsoft AI favours scaling over research; Microsoft has not responded and its lab page still lists the group as part of Microsoft AI.

Questions and Answers

Why did John Langford quit Microsoft?

In a blog post dated 6 November 2026, Langford said the AI Frontiers group was moved into Microsoft AI in early September, that the division's core philosophy is scaling "and aggressively not research", and that he could not get his research used or seriously considered there.

Has Microsoft shut down AI Frontiers?

Microsoft has not announced a closure. Its lab page, updated on 14 September 2026, says AI Frontiers is now part of Microsoft AI and still lists Ece Kamar as lab director. The statement that the group is over is Langford's own assessment.

What did the AI Frontiers group build?

According to Microsoft's lab page, its work included AutoGen, Magentic-One, MagenticUI, Magentic Marketplace, AgentInstruct, the Phi family of models and the Fara computer-use models, with a focus on efficient language models and AI agents.

What research does Langford say Microsoft ignored?

He points to a variant of the transformer design that, by his account, matched a model twice its size, which he projects could halve AI training and running costs. Coverage of his post notes this is his projection, not a tested cost saving.

Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.

#John Langford quits Microsoft#John Langford#AI Frontiers#Microsoft AI#AutoGen

By Anika Sharma, AI & Innovations EditorCovers AI & Innovation and more for Peepals Tech. All stories by Anika Sharma

Get the day's stories in your inbox

Peepals Weekly: the week decoded, free. One email a week, unsubscribe in one click.

Anika Sharma is a pen name of the Peepals Global Editorial Team, which researches, fact-checks and publishes every story. The portrait is an illustration.

Responsible for this story: the Peepals Global Editorial Team, led by editor Prashant Sonar. How we report

More from Peepals Tech

The week's top stories, in your inbox. Free.

The most important, fact-checked stories from all seven publications, once a week.