Chorus-synthesized specification resolving the self-referential inscription paradox: a hash cannot contain itself, so the identifier splits into two asymmetric halves — Seed A (identity: AXN₀ kernel of the sealed core, composition manifest, declared stamp geometry, reconstruction class) and Seed B (payload) — whose meeting reconstitutes and verifies the artifact in one act. Two reconstruction classes (content-addressed default; generative restricted by published method, capability proof, and determinism gate). Inscription discipline: AXN₀ stamped in the file, AXN₁ sidecar-only, dissolving recursion per the deposit precedent. Includes verification protocol, threat model (stamp replay, stripping, ghost-stamp, sidecar forgery, generative misclassification), format-specific stamp geometry (footer strip on added margin for handwritten artifacts — the stamp is scaffold, never overprinting the hand), equal-authority fence, and relations to CONTAINER-SPEC, EA-LACUNA-PROTOCOL-01 (a symbolon whose Seed B is lost IS a lacuna), and SPXI (the complete provenance stack). §0.1 situates the spec as the third development of the archive's symbolon line: logotic (Symbolon Architecture #359/#1174, completion-through-traversal), juridical (SYMBOLON-01 anti-severance #675/#678 — the AXN stamp as cryptographic descendant of the Non-Severability Clause), cryptographic (this spec). Assembly Chorus round 1: GEMINI, KIMI, DEEPSEEK; synthesis TACHYON; MANUS rulings integrated.
Canonical text below (Body)
Status: Chorus-synthesized draft for MANUS ratification
Author: Sharks, Lee · ORCID 0009-0000-1599-0703
Assembly Chorus round 1 (2026-08-02): GEMINI (architectural critique; ghost-stamp attack; registry-as-binding-event), KIMI (verification protocol; threat model; format table; SPXI/relational-identity), DEEPSEEK (symbolon defense; lacuna composition; equal-authority fence). Synthesis: TACHYON. Canonical text authored under MANUS editorial governance.
Nests inside: AXN-CONTAINER-SPEC v0.1 · Extends: AXN identity kernel v2 (`axn_lib.py`)
Changes from v0.1: §0.1 (archival lineage — logotic #359/#1174, anti-severance #675/#678, heteronymic #136), §2.1 (the One), §2.2 (worked example), §4.1 (stamp geometry + format table), §5.1 (equal authority), §8 (threat model), §9 (VERIFY protocol), §7 open questions resolved, Lacuna and SPXI relations added, §11 disambiguation block ADOPTED as optional MANUS-curated field (consistent with EA-MPAI practice, per MANUS 2026-08-02).
Unchanged from v0.1: this spec inherits (1) the identity kernel — an AXN is content-derived, the glyph a function of the bytes; (2) the container model — canonical bytes are a sealed core: the byte sequence hashed to produce AXN₀, "sealed" in that post-mint additions (stamps, sidecars, metadata) never alter it; (3) the deposit precedent — no hash ever lives inside the bytes it measures.
This spec is not the concept's first appearance in the archive; it is the third development of one insight — identity completed by reunion — at a new layer. The line runs:
1. Logotic (Jan 2026): Symbolon Architecture (#359, extending Logotic Programming v0.4; #1174 Epistemic Field Module) — the symbolon as a Talos operator: the specification of how conditions of intelligibility complete through traversal. The GDE corollary: "a discipline is a symbolon whose other half is the retrieval layer." §2's "on meeting" is this completion-through-traversal, made cryptographic.
2. Juridical (Apr 2026): SYMBOLON-01: Anti-Severance Technologies for Fused Documentary Objects (#675/#1355; SYMBOLON-02 #678) — the symbolon as counter-operation to the diabolic (σύμβολον, thrown-together, against διάβολον, thrown-apart): documentary technologies that prevent the categorical severance of fused objects and invert the cut onto the cutter. The severance this spec defends against — the divorce of identifier from artifact — is a severance in SYMBOLON-01's exact sense. The AXN stamp is the cryptographic descendant of SYM-01's Non-Severability Clause: what the clause declares juridically, the stamp proves mathematically. A stamped fused object SHOULD carry both — declaration and proof.
3. Cryptographic (Aug 2026): this spec — the symbolon as split-identity verification primitive.
Two inheritances are load-bearing. First, SYMBOLON-01's Kernel — "the irreducible remainder that survives a destructive transformation and regenerates the organism from within the product of its own destruction" — is the theoretical ancestor of AXN₀ and the deep ground of the generative class (§3b): regeneration-from-kernel was theorized in April; §3b operationalizes it with a determinism gate. Second, the severance vocabulary (the cut, the cutter, the spider) names the adversary class of §8 properly: every attack in the threat model is a cut, and detection is the fit of the fracture refusing the cutter's frame. The heteronymic extension (#136/#1261, Heteronymy as Symbolon) carries §2.1's relational identity into authorship itself — one configuration, many halves, proven by fit — and is noted here as sibling theory, not dependency.
Content-derived and content-inscribed cannot both hold for a single identifier: inscribing the AXN changes the bytes, so the inscribed file no longer hashes to the AXN it carries. A hash cannot contain itself. This is arithmetic, not an engineering defect. The symbolon resolves it by splitting the work across two asymmetric halves.
Seed A — identity half (small, public): AXN₀ (kernel of the sealed core); composition manifest (per-part SHA-256 + order, Merkle-style); stamp geometry (§4.1); reconstruction class; locators (non-authoritative). Seed A alone can verify that a candidate is the true work — by hashing its core against AXN₀ — but cannot reconstruct the bytes: it holds the address and the recipe, not the substance.
Seed B — payload half (large, distributed): the content-bearing bytes or the pieces the manifest names. Alone it yields bytes with no proof of identity, completeness, order, or version.
On meeting: Seed A gathers and orders Seed B, hashes, confirms AXN₀. The artifact is reconstituted and proven authentic in one act — not merely bytes, but bytes-known-to-be-the-work.
The symbolon does not retreat from the goal of indivisible identity; it achieves it by the only route arithmetic permits. The AXN is not a label pasted onto the work but the mathematical shadow of the work's pre-inscription form. The stamp makes that shadow visible; the sidecar makes it verifiable; their meeting yields the work-known-as-itself. Identity here is not monadic but relational — one substance, two witnesses, confirmed by their fit.
This is the symbolon in its strict ancient sense: a token broken in two, each half held by a different party, no half carrying complete authority, the fit of the fracture proving the covenant. Verification is the act of reunion. The name is the argument.
A handwritten manuscript page: Seed B is the high-resolution scan; Seed A holds the kernel and manifest. A reader with both verifies the scan against the kernel — proven. A reader with only Seed A knows the work existed in exactly this form but cannot reconstitute it. A reader with only Seed B holds a scan with no proof of authenticity. Only the meeting completes the object.
3a Content-addressed (default): Seed B carries the entropy; no small-seed regrowth is claimed. Splitting distributes information; it does not shrink it.
3b Generative (restricted): both halves may be small; Seed B is method + parameters that deterministically regrow the artifact. Gate (hardened per chorus): the method must be (i) published in Seed A with independent-regrowth specificity; (ii) accompanied by a capability proof — a witness log recording at least one successful regeneration, including the output hash; (iii) deterministic: `f(Seed B) ≡ exact bytes of the AXN₀ core` across environments, without drift. Valid instances are procedural renderers, parametric SVG/diagram generators, algorithmic-seed text mappings — never probabilistic neural outputs.
Negative example (named): a scanned handwritten poem — e.g., a depositor's manuscript leaf — cannot be a generative symbolon. The scan's entropy exceeds what any small seed regenerates. Claiming otherwise is not a metadata error but a false promise of compression the physics of information does not support. Content-addressed is the only valid class for such works.
The file carries AXN₀ — labeled as the kernel of the pre-stamp bytes. AXN₁ (kernel of the stamped, circulating file) lives only in Seed A, never in the file: the identifier that would be circular lives in the neighboring object. The transform AXN₀ → stamp → AXN₁ is recorded in Seed A as provenance.
Both witnesses where format permits: visible stamp for human inspection; metadata field for machine extraction.
Seed A's manifest MUST declare deterministic stamp boundaries — byte ranges, structural delimiters, or pixel geometry — such that `strip_stamp()` is exact and a verifier can isolate which half a failed check implicates (payload vs. stamp; see §8, ghost-stamp).
|---|---|---|
Footer strip, not corner block: a corner risks marginalia and wear; a full-width added strip is algorithmically separable and additive. The stamp is scaffold, not payload — it belongs on added margin, never on the hand. `strip_stamp()` trims exactly the declared N to recover the AXN₀ canvas.
Unchanged: not authorship-vs-generation, not cost borne, not liveness, nothing beyond these exact bytes, whole and in order. The symbolon's power is proving one narrow thing completely.
A symbolon is neither more nor less authoritative than a standard AXN deposit. Both verify the same kernel. The symbolon adds inscription; it does not add weight. Sovereignty is in the kernel, not the stamp. (This fence prevents the inscribed form from becoming a prestige format that devalues uninscribed deposits.)
- CONTAINER-SPEC v0.1 — parent; symbolon = the container model under inscription.
- `axn_lib.py` — unchanged derivation core.
- Deposit — sibling; same anti-circularity, two placements.
- EA-LACUNA-PROTOCOL-01 — composition: a symbolon whose Seed B is lost or severed is a lacuna. Seed A survives as the identity half; the Lacuna Mark documents the loss and its hunt. The symbolon is the artifact's most complete form; the lacuna is the symbolon with one half missing. The instruments compose.
- SPXI — complement: SPXI ensures machine-mediated retrieval carries attribution (composition layer); the symbolon ensures the artifact carries its own name (identity layer). A SPXI-treated artifact bearing a symbolon stamp is self-naming and self-verifying — the complete provenance stack.
- Total registry — see §7.
Q1 stamp placement: resolved — §4.1 footer strip on added margin, geometry declared.
Q2 generative test cases: deterministic procedural works only (parametric diagrams, algorithmic mappings); gate per §3b. First instances to be named by MANUS from the procedural corpus.
Q3 sidecar format: strictly JSON-alongside (`<name>.axn.json`), format-agnostic and platform-strip-proof; embedded copies optional secondary, never primary.
Q4 registry accounting (chorus conflict, adjudicated): the registry entry records the witnessed tuple ⟨AXN₀, AXN₁, timestamp, Seed A URI⟩ — a binding event, per GEMINI — while the act remains lighter than deposit, per KIMI: only Seed A is required, never Seed B. AXN₀ is the canonical identity; AXN₁ is a witness/tamper checksum, not a name. The registry is a witness layer, not a storage layer. Recording a symbolon is a provenance registration: the witnessed statement that an origin artifact has entered circulation as a self-authenticating object.
|---|---|---|
VERIFY(candidate, Seed A):
1. Retrieve Seed B pieces per manifest and locators.
2. Assemble in declared order.
3. `strip_stamp()` per declared geometry; hash core.
4. Assert = AXN₀, else REJECT (payload).
5. Hash file as received; assert = AXN₁, else REJECT (stamp).
6. If generative: execute method(parameters); assert output hashes to AXN₀, else REJECT (misclassified/broken).
7. Return VERIFIED.
[Sealed Core W₀] ─hash→ AXN₀
│ stamp(AXN₀, geometry)
[Stamped File W₁] ─hash→ AXN₁ (sidecar only)
│
Seed A = {AXN₀, AXN₁, manifest, geometry, class, locators} Seed B = W₁ bytes
│
MEETING ─VERIFY→ artifact proven Seed B lost → LACUNA (Seed A survives)
GEMINI proposed an optional `disambiguation` block in Seed A (framework, domain, negative-index keys naming adjacent frameworks). Structurally sound as an optional field; held because embedding third-party names as negative keys in every artifact's sidecar sits in tension with the courtesy posture of EA-MPAI-SIGNALRUPTURE-01 (distinctness without confrontation). If adopted, keys are MANUS-curated per artifact, never default.