Vocabulary unification ledger

Type: types/note.md

This ledger records candidates for conceptual unification so that vocabulary reviews and definition revisions can revisit them without reconstructing the original question. The collision ledger tracks one term carrying different senses. This ledger tracks different terms that may describe the same underlying property or mechanism, including cases where different goals led to different names.

An entry records a candidate and the current writing decision; it does not establish equivalence or supersede the linked definitions. Shared examples or similar wording are insufficient: the proposed common account must explain what each concept contributes and identify any conditions under which they come apart. Unification may yield one preferred term, or retain two terms for distinct uses of a common account.

Maintenance

Add an entry when a comparison supplies a specific common account worth testing. Record the terms and their homes, the proposed unification, remaining distinctions, the decision for new text, and what would resolve the candidate. Keep these records compact; extended investigation belongs in a workshop and the resulting theoretical claim belongs in theory.

Pending entries remain here. When a candidate is adopted, put the common account and any retained distinctions in the relevant definitions or theory notes, update affected consumers, and remove the entry. When rejected, retain the useful distinction in the relevant notes and remove the entry. The commit records the resolution; the ledger does not accumulate decision history.

Addressability and near-decomposability

Status: candidate. Recorded: 2026-11-03, operator discussion of prompt semantics following the opacity coherence audit.

Terms and homes. Addressable theory exposes assumptions, scope conditions and parts for individual inspection and revision. Near-decomposability is proposed as a mechanism for local inspection: interactions within blocks can be considered while interactions across blocks are summarized or neglected within a question's tolerance.

Proposed unification. Both may express one organization into meaningful parts with manageable interactions, used for different goals. Addressability uses it for targeted criticism and revision; near-decomposability uses it for understanding behavior. At a coarse semantic level, a prompt clause can expose a commitment such as target language without exposing the model computation that realizes it. Its usefulness for both goals depends on how well that description captures the relevant interactions.

Distinctions still to test. The current minimum for addressability is weaker than this common account: separately identifiable and editable commitments can remain strongly coupled. Conversely, a system may have a near-decomposition that an agent cannot access or does not know. Determine whether these are different grades or access conditions of one property, or differences that require separate concepts. Compare the same semantic level, consumer, question and tolerance; editable text and isolated parameter blocks are not interchangeable tests.

The JSON/exception case is a constructed test: the clauses are separately editable, but their combined obligations are unsettled until scope and exception handling are specified. It shows that separate editability does not establish a usable semantic decomposition. It does not show that no decomposition is possible, or that behavior under an adopted package has strong cross-part interactions; those are further questions.

Decision for new text. Keep the current terms and definitions. Do not infer independent effects from separate editability, or absence of semantic addressability from absence of separately located parameter blocks. Treat the common account as a candidate, not an adopted synonym relation.

Resolution test. Explain both useful local revision and useful local inspection through the proposed common organization. Check the account against a prompt's coarse semantic clauses, explicit but coupled equations, and a decomposition unavailable to the inspecting agent. State what would change in the current definition of addressability if the common account were adopted, and which distinctions writers would still need. The existing matching and bounded-impact account is a comparison target: determine whether it supplies the missing conditions or establishes a substantive separation.