From Running to Reliable
Why exit code zero does not mean the job is done, and how to define meaningful acceptance rules for automation.
Automation often treats a zero exit code as completion. For work that produces files, database records, or external effects, that only proves the process did not report an error.
The real question is whether the expected result is complete, semantically correct, and persisted.
Success is not a boolean
A single run can involve at least four distinct judgments: whether the process exited cleanly, whether its artifacts are complete, whether the result is semantically correct, and whether the actor was authorized to produce it. Collapsing them into success: true makes false-success reports difficult to detect.
Begin with observable outcomes
A reliable workflow describes its observable end state before choosing an execution strategy. The initial run, recovery, and final acceptance can then share the same source of truth.
This is sample writing used to validate the blog structure. It will later be replaced by a real engineering note with code, logs, and a concrete retrospective.