About Agent Loop Notes

Agent Loop Notes publishes practical guides for engineers who build AI agents: how to test them, how to design their tools, how to keep their costs down and how to debug them when they go wrong.

Every post aims to be something you can use the same day. That means clear explanations, working code you can copy, and links to the primary sources (vendor documentation, papers and engineering write-ups) so you can check the details yourself.

What gets covered

• Evals and testing for agents
• Tool and API design for LLMs
• Memory, context and retrieval
• Cost, latency and caching
• Debugging and observability

How posts are made

Each post starts from primary sources: vendor documentation, papers and official engineering write-ups. Every number, limit and API name is checked against one of those sources, and every post ends with a Sources list that links to them.

Code samples are run, or checked line by line against the current documentation, before they are published. A claim that can't be traced to a source is cut rather than hedged.

Corrections

APIs in this space change quickly. If something here is out of date or wrong, it will be fixed and the post's updated date will change.

Follow along

Subscribe with the RSS feed to get new posts as they come out.