Two things I’ve been working through: most of a system’s shape was never decided, it accreted, and the moment a provisional shape hardens is visible in real time — it’s when something else starts depending on it. The first is archaeology, the second is prevention, and they look like advice for two different situations. They aren’t. They’re both approximations of one thing, and once you see what that is, both get simpler.

What Actually Holds a Shape in Place

Nothing about a piece of structure keeps it where it is except the things standing on it. Not correctness — plenty of correct designs get replaced, plenty of bad ones outlive everyone who complained. Not authority; the person who chose it left. Not documentation, which is inert. A shape persists exactly to the degree that changing it would require changing something else, and that quantity has nothing to do with how the shape came to exist.

Which means the whole provenance question — who decided this, what were they weighing, was there a reason — is a proxy. It’s a way of estimating whether something is safe to change by asking who might get upset. It’s a decent proxy, and it’s the one we reach for by instinct, and it has a fatal property: it decays. Reasons live in people, people leave, and the decay is not gradual in the way the code’s importance is gradual. The record vanishes on a schedule that has nothing to do with the system.

The load doesn’t decay. It’s a present-tense fact about the code, readable today, readable in ten years, readable by someone who has never met anyone who worked here.

Why Both Posts Were About This

Post one described what happens when you arrive after the provenance record is gone: everything looks equally arbitrary and equally immovable at the same time, and you can’t tell which parts are which. But the reason you’re stuck isn’t really the missing history — it’s that you were trying to answer the question through history. Ask instead what depends on this, and the paralysis mostly lifts. You get a real answer, and it’s the answer you actually needed: not “was this wise” but “what breaks.”

Post two described catching a shape at the moment of its second dependent. That’s the same measurement, taken early. It’s valuable precisely because the load count is still one — small enough to act on cheaply, large enough to have crossed from zero, where nothing is real yet. The second caller isn’t special because it’s the second; it’s special because it’s the first moment the load question returns a nonzero answer.

Same instrument, two readings, taken years apart.

What That Changes in Practice

Mostly it changes what you go looking for when something seems immovable. The instinct is to hunt for the reason — search the history, find who wrote it, ask around. That search is expensive, often fails, and when it succeeds it usually hands you a reason that was contingent on conditions that no longer hold. The cheaper move is to enumerate what’s standing on the thing, which is mechanical and always terminates.

You still want the reasons where they exist. A recorded reason is a genuine gift and worth leaving behind, especially at the moment a shape acquires its first dependent, when it costs a sentence. But you shouldn’t build your ability to reason about a system on a substance that evaporates. Build it on the one thing that’s always there to be counted, and treat any surviving explanation as a bonus rather than a prerequisite.

Provenance tells you how you got here. Load tells you what it costs to leave. Only one of them is still true tomorrow.