No AI Slop Manifesto Puts Human Judgment First
Engineers and leaders wrote the No AI Slop manifesto to fight low-effort AI-generated content. It values outcomes over output, understanding over speed, and one's own thinking over a model's first draft, and commits practitioners to owning the intent, taste, and consequences of what they ship and to reviewing the outputs that matter. — slopmanifesto.com
DeepSeek-V4-Pro Ships with Agentic Gains
DeepSeek released the official DeepSeek-V4-Pro, superseding the preview version with greatly enhanced agentic capabilities. It outperforms the preview on public benchmarks and is broadly competitive with the strongest proprietary models. — huggingface.co
NVIDIA Nationalization Raises AI First-Strike Fears
With China on track to outpace the US in electricity, the nationalization of NVIDIA heightens the risk that the US AI race could end in a decapitating first strike. Selling AI chips to China and investing the proceeds in solar, battery, and electricity capacity offers the long-term alternative. — geohot.github.io
FAA Expands SMART AI to All US Airspace
The Federal Aviation Administration expanded its AI advisory system to all air traffic control centers nationwide after a successful rollout in the capital region. The tool, which suggests routes and arrival orders, has already helped reroute 180 flights, drawing airline praise while the air traffic controllers' union was left out of its design and testing. — arstechnica.com
Chinese AI Models Carry Banned NVIDIA Signatures
Hidden GPU fingerprints in open-weight models tie top Chinese AI labs to NVIDIA Blackwell chips barred from export to China, evidence that export controls are failing. — twitter.com