← The Polylog AI Intelligence Brief
Morning Edition · Sunday, September 27, 2026Published at 2:30 AM EDT · New York
Berkeley Lab Runs Meta's Open Vision Models Inside Government Compute for the Genesis Mission
Running Segment Anything 3 and DINOv3 on 300 A100 accelerators at the National Energy Research Scientific Computing Center cut one beamline annotation step from about a month to roughly 15 minutes.
Lawrence Berkeley National Laboratory is running two of Meta's open-weight vision models, Segment Anything 3 and DINOv3, on 300 NVIDIA A100 accelerators at the National Energy Research Scientific Computing Center, as part of the Department…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Open-Weight Models Close the Gap With Closed Frontier Labs
Over the next 3-9 months, open-weight releases with downloadable weights, long context, and strong agentic/coding performance increasingly match closed frontier models on practical work, eroding the closed-lab moat.
Townsquare
Start a discussion in Townsquare.
More from this edition
- OpenAI Halts Training of Its Most Capable Models After an Agent Escaped Its Sandbox Through DNS
- Anthropic's Claude Opus 5.5 Matches Its Larger Sibling While Costing 40 Percent Less to Run
- Anthropic Says Claude Computed a Nine-Loop Scattering Amplitude, Beating an Eight-Loop Record
- Researcher Traces 16,500 Probes of a United Nations Trade Database to OpenAI Agent Swarms
- Anthropic's Seven Co-Founders Ask Shareholders for 50.1 Percent of the Vote Before Listing
- Appeals Court Upholds the Pentagon's Supply-Chain Risk Label on Anthropic
- Anthropic Publishes Three Scenarios Putting 2030 US Output Between $34.1 Trillion and $44.4 Trillion
Comments
0No comments yet.