There is a style of building that optimizes for the next demonstration, and a style that optimizes for the next decade. They look similar for about six months. After that, they diverge so completely that they might as well be different professions.
We build for the decade. In practice that means our infrastructure is designed to run unattended, our experiments are designed to be reproducible by someone who wasn't there, and our measurements are designed to be trusted by someone who doesn't want to believe them.
Speed is a wonderful servant and a terrible architect. We move quickly inside well-built systems, but we do not let urgency design the systems themselves. Every shortcut in measurement, every unverified assumption promoted to fact, every 'temporary' hack in a critical path — these are loans taken out against future understanding, and the interest rate is brutal.
Speed is a wonderful servant and a terrible architect.
The long game also shapes how we treat knowledge. Results are recorded so they can be found, challenged, and reused years later. An insight that lives only in one person's memory is an insight the organization does not have.
Patience of this kind is sometimes mistaken for a lack of ambition. It is the opposite. Small ambitions can afford to be hasty; if the goal is modest, the scaffolding hardly matters. It is precisely because we are building toward something difficult that the foundations have to be boring, verified, and built to last.
The work compounds. That is the entire strategy.