From 74f2cc9cd7c6d34085fd72a1654cca0230bfd1ca Mon Sep 17 00:00:00 2001 From: "will.anderson" Date: Sun, 23 Aug 2026 13:13:04 -0500 Subject: [PATCH] =?UTF-8?q?=C2=A734c:=20five=20encoding=20contracts=20-=20?= =?UTF-8?q?no=20headcount=20on=20truth,=20forks=20as=20citizens?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docs/architecture/CALCULATIONS.md | 64 +++++++++++++++++++++++++++++++ 1 file changed, 64 insertions(+) diff --git a/docs/architecture/CALCULATIONS.md b/docs/architecture/CALCULATIONS.md index b4a3ee3328..6d87cf0afd 100644 --- a/docs/architecture/CALCULATIONS.md +++ b/docs/architecture/CALCULATIONS.md @@ -1155,3 +1155,67 @@ So the boundary protects the dissenter structurally: "the world would think we're crazy" is not an argument against a chain of receipts. Crazy is a headcount word. The chain does not count heads. + +## 34c. ENCODING THE BOUNDARY — design contracts, not sentiments + +Translating §34b and its coda into mechanisms the storage and +retrieval layers must enforce: + +### C1 — evidential weight has NO headcount term + + weight(claim) = f(tier, chain_match, grip, decay) + # strictly NO count(holders) + +Invariant test (must hold in code): duplicating a claim across N +parties without new evidence changes NOTHING about its weight. +Ten copies of a rumor weigh exactly one rumor. Confidence is a +function of VERIFICATION DEPTH, never POPULATION. + +### C2 — forks are first-class citizens + + FORK { + versions: [all sides, retained], + opened_at, state: open | resolved(evidence_ref), + resolution_by: arithmetic | still-open + } + +Unresolved forks are entities: retrievable, dated, NEVER garbage- +collected for being old or alone. A fork may only close by +evidence reference (chain math, re-derivation). There is no +code path that closes a fork by counting. + +### C3 — agreement outputs are stamped as INTENT + +Any record whose status derives from signatures carries: + + by_agreement: true # this is a decision, not a discovery + +Downstream reasoning must treat by_agreement records as plans +and commitments — queryable, honorable, binding on ACTION — +but they can never be cited as evidence inside C1's weight. +The schema makes "we all agreed therefore it's true" +UNREPRESENTABLE: intent records lack the evidence field, +structurally. + +### C4 — the minority surface + +Retrieval must be able to answer: + "what does ONE party believe against everyone?" +as a first-class query (minority views), because stagnation +hides exactly there. A lone position linked to an open fork +is surfaced with its fork, its dates, and its verification +tier — visible forever, weighted honestly by R1-R2 alone. + +### C5 — the alarm is not a verdict + +Fork alarms trigger investigation workflows, never automatic +resolution. The system's loudest event produces its most +careful state: everything held, nothing elected, evidence +sought. If evidence cannot decide, the fork stays open — +"we do not know yet" is a representable, respectable, +permanent answer. + +These five contracts are binding on semantic.ts and the storage +layer whenever they ripen (§29). They are the difference between +a system that TOLERATES dissent and one that cannot even +represent majority-coerced truth.