Unknown never seals.

Somewhere right now an automated system is deciding whether a person gets something: a benefit, a claim, a treatment that needs approval first. It has a rule and it has some facts. One of the facts is missing. What does it do?

Most systems answer the question without noticing they were asked it. The missing field is empty, the empty field fails a check, the check returns false, and false becomes no. Nobody decided that a gap in the record should count against the person. It just did, because that is what an empty field does when nobody tells it otherwise.

Crimp is built so that never happens quietly. Its rule is short enough to fit on a badge: unknown never seals.

Three answers, not two

A decision in Crimp starts with a rule and a set of facts. The agent brings both, and it never brings a conclusion: there is no field in which it can assert one. Crimp evaluates the rule, works out the outcome, and seals the rule before anyone knows which way it went, so nobody can tune it after seeing the result.

When every fact the rule needs is present, the outcome is yes or no. When one is missing, the outcome is neither. The decision stays open. It does not seal, it does not collapse into a denial, and it does not drift into an approval because the system was in a hurry. “We don’t know” is a third answer, and it is reported as one.

That sounds like a small distinction. It is the whole difference between a system that decides and a system that guesses.

When a fact breaks

Sealing a decision is half of the job. The other half is what happens afterwards. Facts change: an income is corrected, a document turns out to be the wrong one, a test result is reissued. In most systems the decision made on the old fact simply stays where it is, until someone remembers to look at it again. Usually, nobody does.

In Crimp each fact is crimped to the decision it supports, the way a connector is crimped to a wire. Crimp re-evaluates the sealed rule against fresh facts. If the rule no longer holds, the seal lapses and the decision withdraws itself. No authority overrode it and nobody won an argument. A fact changed, and the decision followed the fact.

Why this matters for the decisions nobody appeals

Appeals are the usual safety net, and the numbers show how thin it is. In 2024, Medicare Advantage insurers made 52.8 million prior authorization determinations. 7.7% were denied in full or in part. Only 11.5% of those denials were appealed, and 80.7% of the appeals were overturned in full or in part.

A reversal does not prove the first decision was wrong: some appeals succeed because a missing document was added. That is exactly the point. A decision that was waiting on a fact should have said so. And for the denials nobody appeals, the figures tell us nothing at all. A system that re-checks a decision whenever a fact under it changes does not depend on the person finding the energy to complain.

What Crimp does not do

Crimp reports; it does not authorise. It holds no credentials, has no outbound access and executes nothing. When something asks what stands, Crimp returns the determinations that still hold, and the caller decides what to do with them. An empty answer means nothing has been decided. It does not mean allowed.

That restraint is deliberate. A decision gate that could also act would be tempted to fill its own gaps. Keeping Crimp on the reporting side of the line is what lets it say “unknown” and mean it.

The flight rule underneath

Our fourth Flight Rule is Bend, don’t break: when we don’t know, we say so, and unknown never quietly becomes yes, and never quietly becomes no. Ratchet applies it to actions, by marking an effect indeterminate instead of retrying blind. Crimp applies it to decisions about people.

Crimp is open source under Apache-2.0, with a signed v0.1.0 release published 13 Sep 2026, so the rule that enforces this is a rule you can read. A decision gate you can’t inspect is just another black box, and a black box is where unknowns go to become answers.