Family B: Repeatability

A definition that gets used once is a dictionary entry.

The first family settles what the terms mean. That is necessary and it does nothing on its own, because a term that means the same thing to everybody is still only valuable if something happens the same way as a result. Repeatability is where a defined term becomes a repeated act, and the two isms in this family are about what that repetition buys and what it quietly costs.

The purchase is enormous and it is not the one people usually name. A repeated act stops being decided. Somewhere between the fiftieth and the five-hundredth performance, the deciding drops out of it, and what used to require a person's attention now runs without one, at speed, while the person thinks about something else. That transfer is the single largest gain available to a human being in a working day, it happens to everybody, and it happens whether or not anybody plans for it.

The cost is in the same mechanism. An act that has stopped being decided has also stopped being examined. It will keep running after its reason has expired, at the same speed and with the same confidence, and the person running it will tell you sincerely that they are following the procedure, and there is no test available to them from the inside that would reveal otherwise.

So two isms, and they divide the work between them. One is about the promise of consistency, which turns out to be the wrong promise and has to be corrected before it can be defended: holding the method does not produce the same result, it produces a stable spread, and the spread is a measurement rather than a failure. That correction changes what a variable result means, from somebody's fault to the system speaking.

The other is about the conversion itself, from a decision made each time to a rule that runs. That one is the closest thing this creed has to a central claim, and it draws the first serious opponent: the moment a rule governs anything anybody cares about, the conditions it reads stop being observations of the world and start being things people produce.

Both chapters in this family end up in the same place from different directions, and it is a place neither of them can get out of. A rule that is still doing its job and a rule that outlived its job are indistinguishable from outside. Both run smoothly. Both produce consistent output. Both pass every test anyone in this family can propose, because the tests were built to detect inconsistency and there is none.

That is the family's real contribution and it is not a comfortable one. Repeatability is how you get the enormous gain, and it is also how you lose the ability to see that the gain has stopped arriving.

Which is why the two chapters keep arriving at the same missing instrument. Neither can tell you whether a smoothly running thing is still worth running. Both can tell you how to make it run smoothly, and both end by admitting that smoothness was never the question.