About Agent Case Files

Agent Case Files reports on AI agents doing actual work: what shipped this week, what it replaced, what it cost, and what broke in month three. Agents demo beautifully and deploy badly, and the gap between those two facts is the whole beat.

We write for the people who have to run the thing after the pilot ends — operators, platform engineers, the person who gets paged when an agent loops. We would rather publish one teardown with real numbers than five essays about what agents might one day do.

What we cover

• Playbooks — setups that survived contact with production, written so you can copy them
• Teardowns — what a shipped agent is actually made of, and where its failure modes live
• Briefings — wire-style notes on what changed this week and whether it matters to you
• Agent Startups — the companies building in this space, read against what they claim

How we work

Every factual sentence in a story is tied to a source we read, and our editors check that the source says what we claim before it publishes. Vendor benchmarks are labelled as vendor benchmarks. When we have not tested something ourselves, we say so rather than implying otherwise. Corrections are made on the page, with a note explaining what changed.