Latest / Models & Agents — Video Edition
Ep 157: Claude just showed it can autonomously improve alignment in other models on a single GPU…
Models & Agents Claude just showed it can autonomously improve alignment in other models on a single GPU — shifting alignment work from human teams to model-driven loops. What You Need to Know: Anthropic released research where Claude researched, proposed, trained, and tested alignment fixes for smaller models in 48 hours on one GPU. Cohere shipped Parse 5, a 2.3B vision-language model aimed at high-volume document parsing at $1.50 per 1,000 pages. Meta researchers trained an 8B model with EvoHarness-RL to match Claude Opus 4.5 on long-horizon agent tasks. Builders should watch how these…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.