Microsoft Research’s AI Frontiers group dissolved after internal reorg
AI Frontiers, a ~50-person research group focused on agentic systems and fundamental capabilities, was reorged into Microsoft AI in September 2026. Key contributors departed citing incompatible research philosophies and lack of path to impact.
3 sources · cross-referenced
- Microsoft Research’s AI Frontiers group (~50 people) was dissolved after being reorged into Microsoft AI in September 2026.
- The group worked on agentic systems (e.g., Autogen, Fara) and fundamental research like the Full Bandwidth Transformer and Next Latent Objective.
- Departures followed a reorg into Microsoft AI, where leadership prioritized scaling over research, according to a post by founding member John Langford.
- Langford and other leaders left, citing incompatibility with the new focus and lack of opportunities to see their work deployed.
Microsoft Research’s AI Frontiers group, a roughly 50-person research organization, was dissolved after being reorged into Microsoft AI in early September 2026, according to a post by founding member John Langford. The group formed in 2023 to better organize research efforts in response to the post-ChatGPT landscape, aiming to coordinate coherent research programs rather than relying on individual efforts.
AI Frontiers worked on agentic systems, including the Autogen orchestration framework and the Fara computer-use agent, as well as fundamental research projects such as the Next Latent Objective for extrapolation and the Full Bandwidth Transformer, which aims to improve model capability via recursion. The group also contributed shared infrastructure, including a monorepo with edge-scale pretraining code that was made available to Microsoft Research.
The reorg placed AI Frontiers under Microsoft AI (MAI), a larger organization led by ex-Inflection leadership whose core philosophy prioritizes scaling over research. Langford argued that this approach, while predictable in returns, risks overlooking research-driven breakthroughs that could yield disproportionate efficiency gains, citing experiments suggesting the Full Bandwidth Transformer could match a model twice its size.
Langford and other founding members of AI Frontiers departed after the reorg, citing incompatibility with MAI’s scaling-only focus and a lack of path to deploy their work. Langford described the situation as a failure of alignment between research values and organizational priorities, and expressed skepticism about Microsoft Research’s ability to protect research-oriented efforts in the future.
The departures leave the future of AI Frontiers’ projects uncertain, including internships that were previously offered. Langford suggested that Microsoft’s research environment could become attractive again if leadership develops a stronger commitment to organizing for innovation alongside scale.
- Nov 11, 2026 · The Verge — AI
Google DeepMind unveils AlphaGenome 2 to predict genetic variant effects on cell types
Trust79 - Nov 11, 2026 · Google DeepMind — Blog
Google DeepMind unveils AlphaGenome 2 to predict DNA variant effects in specific cell types
Trust84 - Nov 11, 2026 · arXiv cs.AI
Researchers demonstrate auditable human-agent teams for autonomous synchrotron X-ray experiments
Trust79