John Langford Says He Left AI Frontiers After Microsoft AI Reorg
A.I. / news
John Langford Says He Left AI Frontiers After Microsoft AI Reorg
Microsoft's lab page confirms the Sept. 14 move into Microsoft AI, while Langford's blog post says the new owner's philosophy is scaling and not research.

John Langford, who helped organise Microsoft Research's AI Frontiers lab, said in a Nov. 6 blog post that he left after the lab of about 50 people was folded into Microsoft AI in early September.
Langford published the post, titled "The End of AI Frontiers," on his Machine Learning (Theory) blog and updated it on Nov. 7. It does not give his job title. Microsoft's own lab page confirms the move but says nothing about departures.
What Langford says happened
AI Frontiers, known as AIF, was set up in 2023 after ChatGPT's release as a sub-organisation of Microsoft Research, Langford wrote. He lists its output as the Autogen agent orchestration framework, the Fara computer-use agent, a training objective called Next Latent, an improved Muon optimiser called Dion3 and the Full Bandwidth Transformer.
"In early September we were reorged into Microsoft AI," he wrote. He describes the new home this way: "The core philosophy there is scaling, and aggressively not research."
On the reorganisation itself he wrote: "As far as I can tell we did good work and were non-consensually reorged away." On his own exit: "So I situated my folks as best I could and left."
He also wrote that AIF built a monorepo "shared by default with all of Microsoft Research."
What Microsoft's lab page says
An update dated Sept. 14, 2026 on the AI Frontiers lab page reads: "Microsoft Research AI Frontiers is now part of Microsoft AI!" The page says the lab will "pursue our same goal of frontier and cutting edge agentic and model research at Microsoft AI (MAI) starting now."
It lists Ece Kamar, corporate vice president and lab director of AI Frontiers, as the lab's leader. The page also says "We'll share more exciting news on this transition soon!" and gives no date for it.
The lab's people page lists 42 team members and four alumni. It says nothing about a reorganisation, and it may predate the move.
The two accounts, side by side
| Date | Event | Source |
|---|---|---|
| 2023 | AIF formed inside Microsoft Research | Langford |
| Early September 2026 | AIF reorganised into Microsoft AI | Langford |
| Sept. 14, 2026 | Lab page says AIF is "now part of Microsoft AI" | Microsoft |
| Nov. 6, 2026 | Langford posts that he left | Langford |
The accounts do not contradict each other on the facts. They differ on meaning. Microsoft's page says the research goal continues. Langford says the new organisation's philosophy is not research, and that characterisation appears only in his post.
What is still unknown
Microsoft has not said, on either page, how many of the roughly 50 people moved to MAI and how many left. The lab page still lists Kamar as director, and neither source says whether that holds after the move. Whether the monorepo and the projects Langford names continue under MAI is also unstated.
The Terminal has covered lab-published research from the other side of the industry, including Anthropic's agent memory study and a study of Anthropic researchers using Claude. One consequence is already dated. Langford wrote that "there will not be internships with the folks there in the coming year."
Sources
More in A.I.
- 01Nemotron 3.5 Transcribe Posts 4.62 Word Error Rate Across Seven Languages, NVIDIA's Own NumberThe offline speech model has 4.3 billion active parameters, 40 languages and a licence that permits commercial use.
- 02Cloudflare's Security Audit Skill Has an Open Issue Saying a Made-Up Finding Can Pass ValidationThe MIT-licensed skill runs six phases of agent audits. Its tracker shows 43 open issues, including a profile-wipe report and a Windows failure.
- 03Anthropic's Agent Memory Study Puts Opus 5.5 at 0.838 With Memory and 0.655 WithoutThe Nov. 6 post tests notes, wikis and offline dreaming, and reports where a wiki beat search and where it did not.
- 04METR Says Anthropic May Already Have Crossed Its AI Research ThresholdAnthropic's amended IPO filing says the line may be reached imminently, while METR's Nov. 5 review calls it plausible that July's models already passed it.