Blog
Filtered by: testing× clear
How to Build a Local Agent Bridge: Context and Failures
The HTTP in a local agent bridge is twenty lines. Compiling context and designing failures is where the real work is. Here is how both are built.
Three Bugs My Test Suite Could Not Find
A green test suite and a working system are different things. Three bugs from one afternoon of real use, and why each was invisible to the tests.
Offline Evals as a CI Gate
Use small golden datasets and CI checks to catch prompt regressions before users do.
Your Agent Passes Every Test and Still Gets the Date Wrong
Your agent testing strategy is broken. Build retrieval, tool parameter, and end-to-end evals that predict production behavior.