AI Agents
How agent harnesses, tools, and evaluation actually determine what an AI system can do.
A language model and an AI agent are not the same thing. The model supplies raw capability; the agent is what happens when that model is wrapped in a harness: the tools it can call, the memory it keeps, the context it's given, and the way its work gets checked. These essays look at that architecture directly: where intelligence actually lives in an agentic system, and how to tell whether an agent is reliable rather than merely lucky once.
