The topic file

Agents.

Systems that take actions, use tools and carry work across steps. Follow the boundary between a demonstration and dependable use.

The reading desk

Agents briefings

Primary documents. Editorial analysis. Explicit unknowns.

AgentsReference guide

How to evaluate an agent beyond its best demo

Define the final outcome, preserve failed attempts and measure correction effort. A synthetic invoice example shows how to turn that advice into a usable evaluation.

Sources checked
Sources checked 3 min readRead