TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
A founding leader of Microsoft Research’s AI Frontiers group says the team is over after it was moved into Microsoft AI in an early-September reorganization. The account describes a conflict over research priorities, but Microsoft has not provided a response or independently confirmed the group’s status in the supplied material.
Microsoft Research’s AI Frontiers group has ended, according to a founding leader who said the team was moved into Microsoft AI in early September and that its leaders subsequently left or found other positions. The account frames the dissolution as a consequence of conflicting approaches to AI development, though Microsoft’s position and the precise employment status of all group members are not established in the supplied source.
In an account published on hunch.net, the founding leader said AI Frontiers was formed in 2023, after ChatGPT’s release, to coordinate research efforts that could struggle when pursued by individuals without enough computing resources, strong comparison models or collaborators. The group had about 50 people and sought to organize work between fully top-down management and independent, bottom-up projects.
The leader cited work on AutoGen, an agent orchestration framework, and Fara, a computer-use agent, alongside research into model architecture and training. Examples included the Next Latent objective, Dion3 and the Full Bandwidth Transformer. The account also described an internal shared code repository intended to make research tools and model-building work accessible across Microsoft Research.
The group was reorganized into Microsoft AI in early September, according to the post. Its author said the change could have brought AI teams together, but argued that the new organization’s emphasis on scaling and data work did not fit AI Frontiers’ research focus. The author said they left after trying to find a place for the team, and reported that other members of the founding leadership reached a similar conclusion. The post also said AI Frontiers would not offer internships in the coming year.
Research Priorities After the Reorganization
The account highlights a strategic question for Microsoft: how the company connects research discoveries with the teams and resources that build deployed AI systems. AI Frontiers’ leader argued that research can improve model capability or efficiency, while large-scale training and data work provide a more predictable route to product performance. Those approaches can complement one another, but the post says the group’s new organizational setting did not give its work a practical path forward.
The possible stakes extend beyond staffing. The author said a variant of the Full Bandwidth Transformer could perform comparably to a model twice its size, citing experiments attributed to ByteDance. That is a research claim reported by the author, not an independently established result in the supplied material. If such efficiency gains were validated and usable, they could affect training and inference costs. The post does not provide the experiments’ methods or results in enough detail to assess that comparison.
For researchers and students, the reported end of the group also means the loss of a named research team and, according to the author, its planned internships for the coming year. It does not by itself establish that Microsoft has stopped all work on the projects or that the cited researchers have left the company.
Products Worth Considering
Bose QuietComfort Wireless Noise Cancelling Headphones, Twilight Blue
LG UltraGear 32GS60QC-B 32-inch Curved QHD Gaming Monitor (180Hz, 1ms, 1000R, AMD FreeSync, HDR10)
Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite
As an affiliate, we earn on qualifying purchases.
From MSR Team to Microsoft AI
The author says AI Frontiers took shape in 2023, when researchers saw a need to coordinate more closely in response to the pace of AI development after ChatGPT. The group aimed to assemble stronger baselines and shared resources so ideas could be tested and developed more quickly than a single researcher might manage alone. Its work included both agent systems and more fundamental model research.
The account describes computing capacity as a recurring constraint and says the group tried to share infrastructure and code across Microsoft Research. In early September, it was reorganized into Microsoft AI, which the author viewed as potentially useful for coordinating AI efforts because Microsoft AI was larger. The leader’s criticism concerns how the change worked in practice: the author says the organization’s priorities leaned toward scaling rather than research. These characterizations come from one participant’s account, not a company statement.
“So AI Frontiers is over.”
— A founding AI Frontiers leader, writing on hunch.net
What the Post Does Not Establish
The supplied source is a first-person account by a founding leader, and it does not include a response from Microsoft. The precise date and scope of the reorganization, the number of people who left, and the current roles of remaining team members are not specified. It is also unclear whether AI Frontiers’ projects, code or research have continued within other Microsoft organizations.
The account presents incompatibility over research and scaling as a central reason the group ended, but it does not document internal discussions or provide an independent account from Microsoft AI leadership. The claims about the Full Bandwidth Transformer’s potential efficiency are also not independently substantiated here. The post’s reference to a coming year of internships does not name the year, so the timing cannot be fixed from the supplied material alone.
Questions for Microsoft and Researchers
The next useful update would be a statement from Microsoft clarifying the group’s status, where its research and staff have gone, and whether work such as AutoGen, Fara and the cited model research remains active elsewhere. Confirmation of whether the planned internship program is canceled or being offered through another team would also help prospective students.
For the technical claims, further information about the cited experiments—including publication or evaluation details, comparison conditions and whether the results have been independently reviewed—would help readers judge their implications. Until those details or a company response are available, the group’s end and the reasons described for it should be understood through the founding leader’s account.
Key Questions
What happened to Microsoft Research’s AI Frontiers group?
A founding leader says AI Frontiers has ended after being reorganized into Microsoft AI in early September. The supplied material does not include a Microsoft statement confirming the group’s status.
Why did the group’s leader say it ended?
The author said the group’s research focus conflicted with Microsoft AI’s emphasis on scaling and data work, and said they could not find a suitable path for the team’s work. This is the author’s account and assessment.
What research did AI Frontiers work on?
The leader cited the AutoGen orchestration framework, the Fara computer-use agent, and research into model objectives and architectures, including the Full Bandwidth Transformer.
Will AI Frontiers offer internships?
The author said the group would not offer internships in the coming year. The supplied source does not specify the year or say whether another Microsoft team will offer related internships.
Has Microsoft confirmed the explanation?
No Microsoft response is included in the supplied material. Details about staffing, the fate of individual projects and the company’s reasons for the reorganization remain unclear.
Source: hunch.net
