AI News
Microsoft’s AI Frontiers Group Is Over, Founder John Langford Says, as Research Loses Out to Scale
On November 6, John Langford published an essay titled “The End of AI Frontiers” on his personal blog and announced his departure from Microsoft on X the same day. Langford, a partner research manager at Microsoft Research New York and the co-creator of the group, writes that AI Frontiers — a sub-organisation inside Microsoft Research created in 2023 after ChatGPT, with about 50 people — is over. The news was picked up by The Brief and NewsBytes on November 7. Everything in this article about the group ending is Langford's account: Microsoft has not publicly confirmed a formal shutdown.
What AI Frontiers built
The group's best-known work is the AutoGen orchestration framework for multi-agent systems and Fara, a computer-use agent. Alongside the products, its New York team did fundamental research: the Next Latent objective for extrapolation, Dion3, an improved version of the Muon optimiser, and the Full Bandwidth Transformer. It was, in other words, a research group that shipped tools developers actually use, and that is exactly the combination Langford says no longer fits inside Microsoft.
The September move into Microsoft AI
The break came in early September, when the group was moved out of Microsoft Research and into Microsoft AI. Langford writes: “In early September we were reorged into Microsoft AI.” His account of the new leadership, drawn largely from former Inflection AI executives, is blunt: their philosophy is “scaling, and aggressively not research”, and “there was plenty of evidence that they just did not really want us.” He describes the group as having been “non-consensually reorged away”, and gives his own reason for leaving in one sentence: “I am just not comfortable being at a company where the research I do has no path to real use.”
The case for research
Langford's essay is also an argument that the trade-off is a mistake. “Returns to research are real and significant,” he writes, pointing to experiments at ByteDance that he says suggest a variant of the Full Bandwidth Transformer might match a model twice its size. That is Langford's argument, not a verified benchmark: the ByteDance result is cited by him as evidence for his case, and no independent measurement is offered. The point he is making is structural rather than numerical — that a company which stops funding research that has no immediate product path eventually stops getting the breakthroughs that make the products possible.
What happens to the people and the projects
Langford says he placed his team members as best he could before leaving: “So I situated my folks as best I could and left.” He adds that “every other founding AIF leadership team member seems to have come to a similar conclusion.” His conclusion is direct: “So AI Frontiers is over. I'm sorry to say that there will not be internships with the folks there in the coming year.” On X, where he announced the departure, he wrote: “There's a bit of a train wreck here that's painful to express.”
AutoGen’s status is a separate question
It would be wrong to read the end of the group as the end of AutoGen. The framework's support had already been withdrawn separately: a change to the AutoGen README merged on October 16 (pull request 8366) states that “Microsoft is not providing bug fixes, security fixes, new features, or support for this project.” The last autogen-agentchat release on PyPI, version 0.7.5, dates from September 30, 2025. Fara, the computer-use agent, is MIT-licensed, is not archived, and was last pushed to on September 22, 2026. Langford's post does not say what happens to AutoGen or Fara, or who maintains them now, and neither fact should be presented as caused by his departure.
What Microsoft has and has not said
As of November 8, no Microsoft statement responding to Langford's essay has been found — one outlet that looked for one reports finding none. The official Microsoft Research page still describes the September move as a transition into Microsoft AI, in an update dated September 14, 2026. So the public record has two versions: Langford's, that the group is over, and Microsoft's, which frames the same move as a reorganisation. Nothing in the reporting states headcount consequences, so this article does not claim job cuts — Langford says he placed team members before leaving, which is all that is known about what happened to them.