Convert

The problem

A policy names a duty. People and agents already act against it. The log says what they did. It does not say which clause decided, and a third party cannot check a determination without the institution's cooperation.

Operating policy is usually less machine-evaluable than the public regulation it points at. The common reason is not hidden judgment: operating documents move the rule out of themselves instead of settling it. They are elsewhere-heavy, not judgment-heavy. And public regulation is itself weakly machine-evaluable; a clause that points at a statute often points at a statute that does not decide.

Why a third party

The team that lives with a policy resolves its ambiguities in the direction of convenience, silently, one case at a time. We have no stake in how any clause resolves, so we record the ambiguity instead of closing it. The client's team receives the gap list and adjudicates it. The register carries what was decided and what was left open, and both are visible.

What the engagement is

You get a versioned clause register built from a policy you already publish, with every clause carrying source text and a locator. Ambiguities are registered for your team's adjudication. We replay determinations against the register and deliver a divergence report by clause, with signed sample records from the run. The engagement is complete when you open a record and verify it with our service off.

Evidence

We encoded two published regulatory instruments as machine-evaluable registers.

The encodings

Instrument	Clauses	Note
SRF regime (Singapore)	104	An encoding of the MAS Guidelines; 30 ambiguities registered, none resolved
UK payment-systems instruments	21	An encoding of public PSR

Undetermined, by denominator

Of	Undetermined
all claims	7,396 / 42,188 (17.5%)SYN
claims inside the Guidelines' scope	7,396 / 18,445 (40.1%)SYN
in-scope claims not already closed by an established breach	7,396 / 17,905 (41.3%)SYN

Undetermined: in scope, not closed by an established breach, and waiting on a fact or judgment the register names per claim.

Evaluation is deterministic: the same recorded facts against the same register version produce the same determination, regardless of which agent submits them, how many run in parallel, or when the claim is replayed.

Textual ambiguity is a property of any instrument evaluated by a machine rather than read by a person, not of any instrument in particular. In 1,200 of 10,509SYN account-holder-bears outcomes, the loss lands on the account holder after an established duty breach whose causation was denied, an interpretive choice the register surfaces as an explicit, recorded input rather than a silent default.

Every record is synthetic, signed under a key generated for the run and destroyed at its end; no figure here is a measurement of real-world claims.

A register should say where the text stops deciding rather than manufacture certainty.

Terms

Fixed price for one decision type against one source policy you already publish, quoted with the eyes-only assessment of your page. Fifty percent before we start; the balance when you verify a record with our service off.

Delivery in three weeks from receipt of the source policy: one week to the register and gap list, one week for your team's adjudication pass, one week for the replay and signed records.