He argues that accurate advance estimates are generally unavailable for software work in large, poorly understood systems.
The essay separates small, repeatable tasks from broader projects. Goedecke gives a text-link change as an example that can be estimated from known steps: making the change, waiting for CI, and deploying it. By contrast, he describes much programming work as research into existing code, prior art, and the effects of a change.
His proposed practice is to collect organizational context before inspecting the code: the pressure behind the request, whether delivery is mandatory, and the kind of estimate management expects. He says the available deadline can shape the scope of the implementation, using a PDF-chat feature as an example of either a more complete ingestion and search system or a simpler text-based approach.
This is an individual engineer’s account and recommendation. The source does not establish that its approach improves delivery outcomes, applies across organizations, or makes project durations predictable.