Public Proof Path

A judgement should leave an inspectable chain.

DAEPOLIS is designed so a serious reader can move from source intake to claims, evidence, review, ledger state, method checks, external anchors, and later error memory. The proof path does not ask the reader to trust a fluent answer. It shows the operational objects underneath it.

The Chain

This is the minimum public path a DAEPOLIS record should be able to show when it is mature enough for outside inspection.

01
Source / Document Intake/demo/decision-record

Input becomes reviewable material, not authority.

Source packs and uploaded documents become excerpts, candidate claims, source metadata, source gaps, and hardening tasks. Unverified material can assist a record, but it cannot silently become source-grounded proof.

candidate
02
Candidate Claims/ledger

Prose is split into atomic assertions.

Claims are separated from the generated brief so each assertion can carry support status, verification state, evidence links, falsifiers, and later updates or disputes.

claim ledger
03
Source And Claim State/ledger

Proof posture is visible beside the claim.

The ledger distinguishes provisional, source-assisted, source-grounded, contradicted, primary-supported, expert-supported, journalistic-only, inferred-only, stale, and contradicted states.

public proof
04
Decision Record/decision-record

The judgement gets a durable container.

A Decision Record holds the question, decision context, claims, evidence graph, forecasts, dissent, review state, monitoring triggers, and outcome memory.

record object
05
Review Event/decision-record

Lifecycle facts are append-only.

Creation, review, challenge, correction, approval, dispute, export, and retraction events become immutable record events instead of disappearing into chat or document history.

event spine
06
Ledger And Export/ledger

Public records expose what can be checked.

The public ledger and API expose claim proof, forecast criteria, preregistration hashes, resolution state, canonical claim activity, and audit completeness without exposing private source material.

inspection
07
External Anchor/ledger

Selected hashes can leave the database.

OpenTimestamps anchors SHA-256 digests only. Calendar-pending proofs stay pending; only Bitcoin-attested proofs are labelled anchored.

third-party time
08
Method Bite/ledger

Formal machinery can revise the bottom line.

Deterministic game-solver receipts, base-rate gates, adversarial audits, and bottom-line revisions can attach to the ledger chain when a method check weakens or rejects the visible judgement.

method receipt
09
Error Memory/ledger

Wrong forecasts become reusable intelligence.

Resolved misses can carry Brier score, attribution payload, source adjustments, argument adjustments, evaluator status, and repair instructions for future analyses.

learning loop

Where To Inspect

The public trust surface is split across concept, demo, ledger, JSON, and integration views. Each view answers a different due-diligence question.

Compiled Seed

The first live proof-path target is the public PFA Defence End-Use seed. Its JSON surface at `/api/proof-path` compiles one public analysis into claims, evidence links, a Decision Record, immutable events, hardening tasks, and forecast/resolution hooks.

This is a public demonstration seed, not a client record and not a private PFA internal artifact. It proves that the judgement chain is inspectable; unresolved forecast outcomes still resolve later through the prediction ledger.

Publication Gate

Publication is a proof-readiness boundary, not a visibility switch. Records that cannot show the minimum chain should remain private, draft, pending, or visibly incomplete.

Claim and evidence completeness
Source-hardening and claim-verification states
Decision Record present
At least one immutable record event
Forecast resolution grade or explicit no-forecast reason
Clean critical quality flags
External anchor status represented truthfully

What This Proves

The path proves that DAEPOLIS creates durable judgement objects: claims, evidence, events, hashes, reviews, method receipts, and correction memory. It does not prove that every judgement is correct. It makes correctness, incompleteness, dispute, and later failure inspectable.

The strongest record is not the one that sounds most certain. It is the one that shows how it could be challenged, how it could be falsified, who reviewed it, what proof exists outside the prose, and what the system learned when reality pushed back.