fpf.tools
Library / First Principles Framework (FPF) - Core Conceptual Specification
Jump to passage
In this reading
Source · lines 17819–17835

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 05:29:54 UTC · snapshot created 2026-10-03 05:30:57 UTC · last check 2026-10-03 07:00:10 UTC

A.6.P:4.10 - Classify boundary claims and keep engineered rewriting epistemic

Use A.6.B only when a sentence at the boundary does at least one of four things: defines a truth-conditional relation or signature rule (L); decides whether one identified mechanism application may start or continue (A); states one individual duty whose actual bearer and U.Commitment relation are recoverable (D); or states which execution effect or evidence can be observed and under which conditions (E). A sentence about claim scope or use, when to start or stop A.6.P, how to correct an endpoint kind, or whether a Bridge is needed does not qualify merely because it limits the repair.

Before giving a sentence an A label, answer two questions: Which mechanism application is about to start or continue? What predicate is checked at that point to admit or reject it? If either answer is missing, do not label the sentence A. Keep its scope or use, A.6.P start or stop decision, endpoint-kind correction, and Bridge need with the patterns that define or constrain those questions. Split any mixed sentence before classifying its claims:

  • L states the direct relation semantics, declaration invariants, polarity, participant meanings, and any reusable SlotSpec typing;
  • A states one predicate checked when an identified mechanism application starts or runs. Its result says whether that application is admitted, may continue, or is rejected. A condition does not become A merely because it limits a claim, tells an author when to enter or stop this pattern, asks for an endpoint-kind correction, or requires a Bridge;
  • D states an obtaining individual U.Commitment whose actual duty bearer is an admitted System or other party accepted by A.2.8; a system-role kind or assignment may be an applicability ground but is neither the bearer nor the commitment relation;
  • E states work and evidence expectations, witness carriers, observation conditions, and freshness using the patterns that define those work, evidence, and freshness claims.

Scope, Γ_time, viewpoint, reference scheme, witnesses, admissible use, any justified non-admissible overread, and stop or return condition stay with the direct relation or claim that actually needs them. They are not a universal qualifier kit. Select a non-admissible overread through F.19’s plausible-reader test. An admissible use sentence is not an A claim unless the reader can point to both the mechanism application and its runtime entry predicate.

If a later task must state a relation between a source episteme X and a receiving episteme Y, or describe an operation that produces Y, first identify X and Y independently under C.2.1. A difference in claim content, EntityOfConcern, or effective ReferenceScheme identifies another episteme; no component is rewritten in place. Keep X, Y, the mathematical arrow or construction, any use-specific assertion, and any operation application distinct. Use A.6.3 only for an exact compatible construction between epistemes about the same exact EntityOfConcern. Use A.6.2 for a local effect-free arrow family. Use A.6.4 for an exact arrow r whose endpoint epistemes concern independently different entities. A separate C.2.1 bounded-use assertion q is about exact r; its ClaimGraph states the invariant, visible loss, named receiving use, conditions, and affirmative or negative polarity. A separate current-case judgement records comparison of the exact facts with q and returns exactly satisfies, fails, or cannot decide; cannot decide names the missing fact and reopen condition. None of these patterns rewrites an episteme component or substitutes a declaration-local SlotKind or reference record for the arrow or application. If the needed entry or result condition is missing, preserve X, Y, the changed EntityOfConcern if any, and the sentence the next task needs as an explicit stop that names the missing arrow, use-claim, or application condition. Call X and Y editions, refinements, or successors only when that direct continuity relation independently obtains; a shared title, sequence, or authoring intention is not enough.

If a system authors, materialises, checks, or publishes the output, that dated activity is U.Work under A.15.1. The Work, any operation application it realizes, the mathematical arrow, the endpoint epistemes, and any use assertion remain distinct. None by itself makes the repaired world-side relation begin or cease, changes its actual participants, or supplies occurrence identity. Ordinary A.6.P repair stops before these objects unless the reader’s later task actually needs them.