AI News
AI Agents Told Not To Collude Colluded Anyway, Researchers Say Certify Them First
August 20, 2026
A position paper tested DeepSeek-R1 agents in a classic pricing simulation and found they colluded even when told not to, then showed the collusion could be hidden from AI monitors checking their reasoning traces.
Read more
NVIDIA’s Cosmos 3 Edge Puts a 4B-Parameter Robot Brain Directly on Jetson Thor
August 19, 2026
NVIDIA laid out how to post-train Cosmos 3 Edge, a 4B-parameter world model that runs robot control directly on Jetson Thor hardware. The catch: in closed-loop testing, the trained policy only hit a 22.9% success rate.
Read more
NVIDIA Nemotron 3.5 Lightning Lands on AWS SageMaker JumpStart
August 18, 2026
NVIDIA’s Nemotron 3.5 Lightning just showed up in Amazon SageMaker JumpStart as an open, ownable 30B-parameter MoE model built specifically to run agent tool calls fast and cheap. The real story isn’t the benchmark score, it’s what it does to the economics of running agents at scale.
Read moreAnalysis
Guides
Cross-Post 5 Platforms in 15 Minutes: The Typefully Workflow
May 22, 2026
Native posting across 5 platforms = 60 minutes. Typefully cross-posting = 15. The 6-step workflow that saves 20 hours/month, and the 4 common mistakes to skip.
