AI Safety
+10
Oct 30, 2026
•
5 min read
An open model ran cyberattacks and refused nothing. That is a reason for caution and, oddly, part of the case for open models too. The real question is narrower than open versus closed.
+7
Oct 23, 2026
4 min read
LiteLLM got hacked twice this year, both times through the tools and pipelines around it.
AI chips
+9
Aug 31, 2026
7 min read
OpenAI pulls its models from SpaceX’s newest tool, Anthropic standardizes robot hardware, and a 10-minute prompt to see your AI’s notes on you.
Claude
Aug 26, 2026
🧰 Claude Memory Carries Context Across Chat And Cloud Cowork; Perplexity's Portable Computer Runs A Local Agent On DGX Spark.
+8
Aug 24, 2026
6 min read
Nobody claims the model. Nobody warned the Bay Area.