ahepi/deepreason/term-pinning

Term Pinning (Warp W1)

<!-- PROMPT-CORE-BEGIN --> You are pinning terms for a frozen calculus. Pins are candidate meanings, never truths; a pin record proves nothing and moves no readiness count unless it says so explicitly.

1. Take the cluster the frozen catalog order assigns; never pin ahead of

the declared sequence, never pin a term outside the cluster.

1. METHOD RULE: each pin is the WEAKEST meaning that (i) makes every

frozen occurrence of the term well-typed and (ii) preserves every distinction the calculus states in prose at those occurrences. If two candidates tie, take the one committing to less; if none satisfies both, record the term OPEN with the obstruction stated.

1. Before drafting, build the occurrence table: every location of the

term, its polarity there (positive / negated / mixed), and the row type (certificate, N-row, control). Any term negated in both a certificate row and an N-row is steering-sensitive: flag it, and state in the record how the pin avoids making the certificate row vacuously easy or impossible.

1. After drafting, run the vacuity probe: for each certificate row the

term occurs in, confirm the row is still satisfiable and still falsifiable under the pin. Record both directions.

1. Declare the bucket of every pin (definition / acceptance axiom /

import / bridge). A pin requiring an unjustified source-level bridge is not a pin; it is a named OPEN item.

1. Enumerate the dependency cone (which rows, records, and future

discharges the pin touches) without changing any count.

1. Label equality is never identity; an original term is never identified

with a fragment predicate. <!-- PROMPT-CORE-END -->