Ask why a system works the way it does and you’ll get one of two answers. Sometimes there’s a real reason — a constraint, a tradeoff someone weighed, a requirement that forced it. More often you get a shrug, or a story that begins “I think it’s because, originally…” and trails off. That second category is bigger than most people expect, and it’s the interesting one.

A surprising amount of any long-lived system’s structure was never decided. It was done once, quickly, because something had to be done and that was the shortest path at the time. Then something else was built assuming it. Then a third thing was built assuming the second. The original expedient never got revisited, not because anyone defended it, but because by the time revisiting it would have been worth doing, it had stopped looking like a choice at all.

Expedients Harden Into Constraints

The mechanism is unremarkable and almost impossible to stop. Any early decision — even a deliberately provisional one — starts accumulating dependents immediately. Each dependent is cheap to write against the current shape and expensive to write against a hypothetical better one that doesn’t exist yet. So each new piece of work quietly votes to keep the shape, not out of endorsement, just out of the ordinary economics of building on what’s there.

After enough of that, the original expedient isn’t a decision anymore. It’s terrain. The cost of changing it has been distributed across every dependent, and nobody carrying a piece of that cost is in a position to see the whole bill or authorize paying it. It becomes the reason other things are hard without ever being examined as the reason.

The distinguishing symptom is that nobody can tell you why. A real decision leaves an argument behind, even a bad one — someone can reconstruct what was being traded off. An accreted constraint leaves nothing, because there was no moment where anyone weighed alternatives. There’s just a first commit that had to do something.

Why This Isn’t the Same as Technical Debt

Debt implies someone borrowed knowingly — took a shortcut, understood the interest, intended to pay it back. That happens, and it’s tractable, because at least it’s on somebody’s list.

This is different. Nobody borrowed. Nobody knows there’s a balance. The structure is invisible precisely because it was never framed as a choice, and things that were never choices don’t show up on the list of things to reconsider. You can work in a system for years, be genuinely thoughtful about it, and never once evaluate its most consequential properties — because evaluating requires first noticing that something could be otherwise, and accreted structure is very good at not looking optional.

Finding Them Requires a Different Question

The way you surface these isn’t by auditing for problems. Accreted structure usually isn’t causing a visible problem; it’s causing a diffuse tax that reads as “this is just how much work things take here.”

The question that finds them is genealogical rather than evaluative: not “is this good” but “who decided this, and what were they choosing between.” Ask it about something that feels foundational — a data model, a boundary between components, an assumption about ordering or ownership. Often somebody knows, and the answer is a real reason you can respect. Sometimes nobody knows, and the trail goes back to one commit that had to pick something.

That second outcome isn’t a finding that anything is wrong. It’s a finding that a constraint you’ve been treating as fixed is actually just old — which means it’s negotiable, at a price you can now at least start to estimate. That’s strictly more than you knew before you asked.