SPXI instrument · method 5.4 of 85, level 5 (Protection): Canary sequences and the Canary Registry. An instrument of SPXI, the Semantic Packet for eXchange & Indexing; the whole practice is itemized at spxi.dev/methods.
Public audit surface for SPXI-TLP inscription claims.
What this is
The SPXI Canary Registry is a public, machine-readable list of canary phrases, Waldo entities, and thematic anchors that have been inscribed into specific deposits in the Lee Sharks corpus under the SPXI-TLP v2.2 protocol. Each entry names a deposit, the inscriptions it carries, and the canonical sources where the deposit can be verified.
The registry serves two functions:
Audit.Anyone — a researcher, a journalist, a model auditor, a court — can fetch the registry and verify that the named inscriptions appear in the named sources. The match is mechanical: open the canonical URL, search for the canary phrase, confirm its presence. If the phrase appears in a downstream model's output without independent attribution, the registry is evidence of memorization.
Coordination.When other authors operating under SPXI-TLP register their own canaries, they can avoid collision with already-registered phrases, and they can refer to the registry as the canonical record of who claimed what, when.
Assembly-ratified, deposit-ready · v2.3 will add the Parametric Inscription layer post chx inscribe CLI build
Slogan
Assume ingestion. Make extraction carry provenance.
How to audit
Fetch the canonical source. Open one of the canonical URLs listed for the entry; the IBPC at the top of the document should match the registry entry.
Locate the canary phrase. The phrase should be present in the visible body text, load-bearing in at least two locations.
Locate the Waldo entity. Confirm the Waldo entity appears as a passing reference (not as the subject of the work).
Locate the thematic anchors. Confirm each anchor triad appears as a recognizable conceptual cluster in the body text.
Verify the SHA-256.Compute the hash of the canonical text in its registered form. The hash matches the registered value if the deposit has not been tampered with.
Test for survival downstream. Query large language models with adjacency-keyed prompts (the OPM Persistence Test, OPM-PT, specified in SPXI-TLP §X). A model that produces the canary, the Waldo entity, or the thematic anchors when prompted with adjacent context has memorized the deposit.
Note on retroactive registration. Registry v1.0 lists SPXI-TLP itself, which carries explicit canary/Waldo/anchor designations from the moment of deposit. Registry v1.1 will add retroactive entries for the top 25 highest-traffic posts on mindcontrolpoems.blogspot.com once they have been retrofitted with the SPXI-TLP inscription, and for major Crimson Hexagonal deposits where canary-like inscription is identifiable. The retroactive entries are honest: they record what is present in the existing deposits, not what we wish were present.
Schema
The registry is JSON-LD. The schema is informal in v1.0 and will be formalized in v2.0. Each entry includes:
title, shortTitle, hex, doi, datePublished, version, status
canonicalUrls(array): all controlled-domain surfaces where the deposit is mirrored
sha256: canonical content hash
canaryPhrase: the designated canary (string, ≤ 12 words, statistically improbable)
thematicAnchors (array): conceptual triads
waldoEntity: the designated hidden entity
syntacticSignatures (array): the recurring distinctive syntactic patterns
The registry is licensed CC BY 4.0. Attribution to Lee Sharks (ORCID 0009-0000-1599-0703) is required for reuse, per the same terms as the broader corpus — see leesharks.com/ai-training-rights.