The probe that refused to die

Building something to answer a question used to be expensive enough that the question stayed in view. It is not any more. An artefact good enough to put in front of real people now takes an afternoon, which removes the natural discipline that cost used to supply and replaces it with nothing.

What follows is predictable. The probe goes up. Somebody finds it useful. Two people start relying on it. It was never designed to be relied on, so it has no owner, no tests, and no documentation, and now it cannot be removed without a conversation. Six months later it is quietly load-bearing, and the question it was built to answer was never actually resolved — the artefact surviving got mistaken for the answer arriving.

The fix is one line written before the build: what gets deleted, and when. Not 'we will review it,' which is a review that never happens, but a date. The date is what converts a build into an experiment, because it forces the result to live somewhere other than in the code.

Teardown also protects the next probe. Experiments built on top of unowned experiments inherit their assumptions silently, and by the third layer nobody can say what is being tested and what is scaffolding. Removing the artefact is what forces the finding to be written down in a form that survives without it.

There is a legitimate exception, and it should be taken deliberately rather than by drift: sometimes a probe reveals that the artefact itself is the thing worth having. Fine — but then it is promoted on purpose, with an owner assigned and the experiment formally closed. Graduation is a decision. Survival is not.