how we build · closing
What just happened
Seven parts. The story was pulled on a Monday morning. A few weeks later, the flag is enabled and the feature is live. Between those two moments, the meaning the brief carried has survived the code, the pipeline, the tests, the review, the flag, and the release gate. At each phase the team has had to choose: carry the meaning, or substitute it with the artifact's natural shape.
If they carried it, the code speaks the domain. The tests are named for scenarios the amigos session wrote. The PR traces back to the brief. The analytics events measure the prediction. The release brief names the person whose situation was meant to change. The client knows what to expect and when the check will happen.
If they substituted — if the function was called handlePost, if the test was named test_submit_1, if the PR said "misc fixes," if the analytics event was button_clicked, if the release brief said "we shipped the grading feature" — the software is live and working. It will do what it does. Whether it changes Gal's Tuesday afternoon is a question nobody can answer — because the instruments to answer it were never wired. The client asks "did it work?" and the team has no honest answer. That silence is the cost of a substituted chain.
The feature is live. Gal encounters it for the first time. The 47 minutes either change or they don't. Volume V is where the team finds out.
Volume V — After We Build — begins at this moment. The first 48 hours of watching. The prediction check on the named date. The bugs classified by which level of the chain produced them. The retrospective that produces one change. The model update that makes the next cycle start smarter than this one.
The Prediction was named in Volume II, kept alive through Scope in Volume III, made measurable in this volume. The next volume is where it meets reality.