HuggingFace’s Number One Model Has 0.56 Percent of Rank Twelve’s Downloads
HuggingFace trending put a model with 44,939 downloads at number one while rank twelve held 8,032,442. The board ranks momentum, not load.
The AI Buildout Just Found a Constraint No Grid Operator Controls
Voters settled 41 data center ballot measures and kept the industry out of all but one. The 25 megawatt ceilings sit far below an 8 GW Ohio campus.
Open Weights Now Start at Eight Blackwell Ultra GPUs
Moonshot published Kimi K3.1 Pro as 1,561 GB of MXFP4 weights. The download is free. The node that runs it is not.
The On-Prem AI Server Boom Is a Price Rise Wearing a Growth Rate
On-premises HPC server market revenue rose 19.9 percent while units grew 4.4. Most of that growth is component price, not capacity.
Broadcom Brought Back the Cheap VMware Edition With a 128 Core Ceiling
vSphere Express is live at 128 cores with no published price. The cheap VMware edition is back, and the ceiling is the product.
SambaNova Is Selling Rack Capacity, and the Proof Is a Number Nobody Printed
SambaNova and Supermicro plan racks that split disaggregated inference between GPUs and SN50 chips. The 10,000 chip plan has no performance numbers.
NVIDIA Automated the Restart. The Checkpoint Interval Still Sets Your Goodput
NVIDIA Recovery logged 8,550 training failures in 90 days on one cluster. Restarts are automated now. Checkpoint cadence still drains the GPU budget.
Three of the Top Seven Models on HuggingFace Now Generate Video or 3D
Video and 3D models on HuggingFace now hold three of the top seven slots. A gaussian-splatting model from a tiny lab is the fastest riser on the board.
The Fastest Growing AI Projects Are Harnesses, Not Servers
DeepSeek Harness climbs 3,270 stars a day. Four projects exist only to shrink context, and stars no longer track code.
Antigravity Taught the Industry What an Agent Tool Looks Like, Then Lost the Sale
Google Antigravity set the agent-first design language and sits flat at 6 percent adoption while Claude Code hit 39.