Every architecture decision has a winner and a set of losers - assuming anyone named the losers in the first place. Often they were never articulated at all, and where they were, we still only record the winner. Within a year nobody can say what was rejected or why, which means nobody can tell whether the decision was reasoned or merely fashionable at the time.

You have probably watched this go wrong recently. Work arrives fast from a coding agent, or from a colleague leaning on one, and it looks fine. Nothing needs to break. Someone simply asks why it was done this way, and nobody can answer: not what the alternatives were, not what was traded away, not what would have to change to reverse it. The justification was outsourced along with the work. Agency went with it, and nobody noticed at the time.

Oliver Sturm has spent nearly forty years creating software, much of it in rooms where these decisions are made. In this keynote he replays real decisions with their rejected options restored, and argues for a discipline that costs almost nothing: if you can't say what you didn't choose, you haven't chosen