AI News

August 27, 2025

LangChain has a neat blog post about the generalized structure of "deep agents" (with Claude Code as its inspiration) that are better at complex task that require planning: Deep Agents post. The secret? A "todo" list.
AWS's announcement from July about it's new Agent infrastructure offerings: AWS AgentCore.
Anthropic's blog has 3 vingettes about how they used Claude (and policy) to stop criminals from committing crimes with Claude: Detecting and countering misuse of AI.

On a housekeeping note, I've started to cleanup some of the dead links in the "Reading List" on the home page.