Go Check

Twice a manager has asked me whether the work was done, and twice I gave the answer that gets you fired. I said go check.

The answer that keeps your job is yes sir. The second time, what I asked for was the receipts behind a decision, the chain that would let me see for myself that it was sound. That request came back to me with a label on it: insubordination.

It took me a while to understand that yes sir and go check are not two attitudes. They are two epistemologies, and an organization runs on one or the other.

Yes sir is a claim offered to be believed. It hands you nothing to inspect, and you take it whole, on the strength of who said it. Go check is the opposite move. It hands you the test and dares you to find it wrong. Only the second can ever earn a verdict, because only the second can fail. I wrote the careful version of this argument elsewhere: a claim that cannot fail says nothing, so what cannot be false cannot be true. This is the loud version.

The house style of everyone who can’t afford to be wrong

Here is what took me longest to notice. The go check regime is not eccentric, and it is not mine. It is the house style of every field that has to survive contact with reality.

Aviation runs on checklists and on no-blame accident review. Surgery runs on the morbidity-and-mortality conference, where you stand up and account for the death. Structural engineers re-run the load. Reliability engineers sit down before anything is built and enumerate every way the thing can fail. Site reliability teams write the postmortem the morning the system falls over, and the iron rule is that nobody is punished for the failure, only for hiding it. NASA will not let a requirement into a design until someone has named the test that could prove it wrong.

Engineering even wrote down the philosophy. A design is a conjecture, and a failure is its refutation; the bridge that falls is a hypothesis that turned out false. Nobody in that world finds this strange. Nature cannot be fooled, Feynman wrote in his appendix to the Challenger report, after the organization that mandates the check overrode it one cold morning and lost a crew.

So you would expect the field whose entire subject is knowledge to be the strictest of the lot. You would expect epistemology to be the place where claims are checked hardest.

The one field that skipped its own test

It is the place where they are checked least.

Academic epistemology is what a bibliometrician has called a para-science. It wears the whole costume: journals, peer review, a technical vocabulary, a dense citation apparatus. The one thing it lacks is a world it answers to. Its papers cite other papers. Its counterexamples are intuitions. Its intuitions, when somebody finally went and tested them across cultures, turned out not to hold. Sixty years of patching the definition of knowledge after Gettier, every repair undone by the next invented case, the form of the repair eventually proven unfixable, and the literature just kept accumulating. Accumulation without convergence. A field that studies how claims meet the world, holding its own claims carefully away from one.

By the standard the field would recognize from its own Popper, that is the worst verdict on the board. Not false. Untrue. Edgeless. The same floor where string theory sits, so unfalsifiable it cannot even manage to be wrong.

The honest exceptions prove the shape of it. The wings of philosophy that do face the world, the naturalized and the experimental and the formal, are exactly the ones the armchair core has spent decades treating as barbarians at the gate. The field knows where its test is. It has been guarding the door against it.

A note for the rationalists

The map-and-territory crowd at least kept their slogan pointed the right way. The map is not the territory is true, and it carries weight. But the picture smuggles in a cartographer, someone standing outside both the map and the land, holding them up to read the match. There is no such person. You never climb out of your own map to grade it against the ground. The honest version keeps the discipline and drops the surveyor: you cannot confirm the map, you can only run it into the territory and watch where it tears. That is not a lesser thing. It is the only thing checking has ever been.

Novelty is lying in the open

People ask whether any of this is new, and it is a strange thing to be anxious about. A field that reads only itself leaves every adjacent shelf untouched, and the shelves are full: reliability engineering, control theory, failure analysis, clinical diagnosis, the postmortem. The connections are sitting there in plain sight, walled off by nothing but a citation habit. I am not defending a territory. I do not have one to defend. I went and looked at the next shelf over, which the people paid to study knowledge had decided was beneath them.

I still think go check was the right answer both times I gave it. It cost me two jobs, which tells you a great deal about how the yes sir regime protects itself, and nothing at all about whether the work was actually done. There was only ever one way to know that.

Every regime that runs on yes sir has to treat go check as a threat, because the check is the one move that could expose it. The regime that runs on go check is the mirror image: the same move that gets you fired everywhere else is the only thing holding it up.

If you can show me a single field that survives contact with reality on yes sir alone, I will take all of this back. Go ahead.

Check.