AI Engineering
Small Bugs Should Not Wait for a Free Afternoon
How we connected production logs to a coding agent that reproduces a bug, fixes it and opens a pull request, while engineers keep the merge.
Category
4 posts in this category
How we connected production logs to a coding agent that reproduces a bug, fixes it and opens a pull request, while engineers keep the merge.
Why I built a delivery framework for coding agents, how the loop runs from idea to retro, and the three rules that keep people in charge.
Eight dimensions that decide whether a coding agent becomes effective in your codebase, and why the limiting factor is almost never the model.
Generated code is unpredictable, hard to maintain, and expensive to validate. There is a better way: let AI generate structured declarations, and let a deterministic runtime handle the rest.