Persistent AI memory, AI watermarking, on-policy distillation

This week, AI research has landed on persistent memory for AI agents, a focus on Claude’s new watermarking system (and LLM watermarking as a whole), and taking on-policy distillation as a post-training default.

We’ll get into the intricacies of how these systems work, cover some benchmark results, and the key research movements in every topic to keep you updated in no time.