Agent-native programming
Software should not merely tolerate coding agents. It should make the correct action obvious.
We write about APIs, CLIs, errors, documentation, and infrastructure designed for software agents that need to understand, change, verify, and ship real systems. Predictable interfaces, structured output, explicit side effects — the properties that let an agent act without guessing.
Latest
Agent-native programming
Errors are recovery protocols
How to tell an automated client what failed, whether retrying is safe, what changed, and what to do next.
Read article