Always-On Agents Need Three Layers the Model Can’t Give Them
Last week Anthropic published Building always-on agents. It’s the clearest short description I’ve read of what it actually takes to run an AI agent that keeps working when nobody is watching it. If yo…