John Langford, a machine learning researcher who helped organise the AI Frontiers group inside Microsoft Research, wrote on the blog hunch.net on November 6 that the group is finished and that he has left, according to FourWeekMBA. His post says: "So AI Frontiers is over." AI/TLDR reported that the roughly 50-person team behind Autogen and the Fara agent was folded into Microsoft AI in September. Langford's post says he left, and that every other founding leadership team member "seems to have come to a similar conclusion."
What AI Frontiers was
Langford writes that in 2023, after the ChatGPT release, it was clear to him that research needed to be more organised to succeed than was typical within Microsoft Research. Several other people thought similarly, he says, and together they created the AI Frontiers sub-organisation within Microsoft Research. He writes that he was under the general impression the organisation was succeeding in doing pioneering things, given its small size of about 50 people.
Much of the group's work, the post says, was figuring out how to make agents work, for example the Autogen orchestration framework and the Fara computer use agent. In New York, several researchers were interested in new fundamental capabilities: the Next Latent objective, Dion3 and the Full Bandwidth Transformer. The group also built a monorepo shared by default with all of Microsoft Research, which held the edge-scale pretraining codebase Langford discussed in a COLM keynote the previous month.
The reorganisation into Microsoft AI
Langford writes that in early September the group was reorganised into Microsoft AI. He says that on the surface this could make sense, since combining two AI organisations can concentrate effort, and that Microsoft AI is much larger. In practice, he says, there are substantial issues, because the philosophy of Microsoft AI's ex-Inflection leadership is incompatible with his group's. He describes that philosophy as "scaling, and aggressively not research."
The post does not dismiss the other side's logic: it says there is plenty of evidence that bigger models are better models, and that from a business perspective returns to scale are far more predictable than returns to research. In a scaling-only environment, Langford says, the primary effort goes into getting, making and refining data, which is quite a step away from what his group was working on. He also writes that there was plenty of evidence the Microsoft AI side did not really want them, and that he considers the scaling-only focus "both misdirected and not fun." As far as he can tell, he writes, the group did good work and was "non-consensually reorged away."
Langford's cost argument
Langford makes his case for research in terms of cost. He writes that ByteDance experiments suggest a variant of the Full Bandwidth Transformer might be comparable to a model of twice the size, implying significant savings in training and inference for the same capability. He then asks how much it is worth to halve training and inference cost, given that the industry has investments approaching a trillion dollars. FourWeekMBA noted that this is Langford's reading of another lab's results, not a figure it has checked.
Why he did not go back to Microsoft Research
Langford writes that he left after settling his team as best he could, and that every other founding AI Frontiers leadership team member seems to have come to a similar conclusion. He gives two obstacles to returning to Microsoft Research: he is not comfortable at a company where his research has no path to real use, and he fears a breakdown of the basic deal there, that good work leads to good things. He writes that he could not get innovation into use, or even seriously considered, while in Microsoft AI. He also writes that there will not be internships with the group in the coming year.
What the accounts say and don't say
AI/TLDR pointed out that Langford is describing his own view and that Microsoft is not quoted in the post. FourWeekMBA reported that it did not find a Microsoft statement responding to the post in its searches. It also noted that the post does not say where Langford is going next and does not name the other founding leaders. By FourWeekMBA's arithmetic from the post's own dates, the group ran for about three years.
Langford wrote: "So I situated my folks as best I could and left."
Sources
AI/TLDR (2026) 'John Langford Leaves Microsoft: AI Frontiers Ends', AI/TLDR, 6 November 2026, available at: https://ai-tldr.dev/releases/microsoft-research-ai-frontiers-ends/ (Accessed: 8 November 2026).
Cuofano, G. (2026) 'Langford Exits Microsoft: Its 50-Person AI Frontiers Is Over', FourWeekMBA, 7 November 2026, available at: https://fourweekmba.com/ai-langford-exits-microsoft-its-50-person-ai-frontiers-is-over/ (Accessed: 8 November 2026).
hunch.net (n.d.) 'The End of AI Frontiers', hunch.net, available at: https://hunch.net/?p=13764006 (Accessed: 8 November 2026).