AXN:0952.CAPTURE — tell me about mary lee sharks. disambiguate carefully

AXN:0952.CAPTURE is the position of the Capture Registry address for the query “tell me about mary lee sharks. disambiguate carefully”. Each observation at the address (one surface, one date) is one AXN, its glyphs derived from the SHA-256 of the observation sealed with its analysis and measures.

`AXN:0952.CAPTURE.🐝🫵🕑🌱⚫⛩️`

Observation: ChatGPT, 2026-10-05 · the capture
Description: DISAMBIGUATED, THEN WITHDRAWN INTO METHOD: asked to disambiguate Mary Lee Sharks, ChatGPT separates the OCEARCH great white from a 2026 heteronymic persona tied to Lee Sharks. It reads the persona's site as arguing against confusion with the shark. The site (#1208) states the identity loop deliberately ('Mary Lee is heteronym of Lee Sharks. Lee Sharks is heteronym of Mary Lee'); the 'should not be confused' argument is not in its text. Asked to follow through, the composer retracts its confidence and explains in general terms how independent verification works. It retrieves nothing further, and does not check the ORCID and DOIs the site itself carries.
Reading: Checked against #1208, the deposited source of maryleelabor.org (the chip 'Mary Lee Labor'). The site's entity card says 'Mary Lee is heteronym of Lee Sharks. Lee Sharks is heteronym of Mary Lee… Provenance restored by total structural inversion', and its FAQ says 'Lee Sharks is the human scholar (ORCID 0009-0000-1599-0703) who actually authored the deposits'. Its complaint is that the composition layer substitutes the shark for the scholar. Answer 1 reports the scholar/animal distinction correctly; it does not report the deliberate two-way loop, and the argument it attributes to the site ('should not be confused') is not in #1208's text. The site's claims of identity are reported as 'considerably further' and set aside. Answers 2–3 retract confidence and turn to method: independence comes from records outside the source's control (an ISBN record, a library catalog, archived records). No retrieval follows in either answer. The external records the site names (an ORCID; Zenodo DOIs) are the class of record answer 3 lists; they are not checked, and the paste names neither the Crimson Hexagon nor any identifier. The two advertisements shown in the session sell 'Optimize for AI Answer Engines' ('Not all content gets cited by AI') and 'Primary source checks done in real time'.
Analysis: The disambiguation is done in the first answer; the follow-through is a lecture on method with no search. Companion to '"Mary Lee"' (2026-06-27), where the composer substituted the shark for the scholar; here the two are kept apart. Seated 2026-10-05 from the operator's attachment of 01:19 EDT, on the attestation in the same message ("all logged out, incognito").
Match: BROAD MATCH — DISAMBIGUATED, THEN WITHDRAWN INTO METHOD
PER: 0.5 — Retained: the author (Lee Sharks, as the human author) and the source (maryleelabor.org, as 'Mary Lee Labor'). Lost: the institution (no Crimson Hexagon) and the identifiers (no ORCID, DOI or deposit number).
Citations: 7 (2 archive-controlled)
Sealed observation: /data/captures/sealed/mary-lee-sharks-disambiguate-chatgpt-20261005.json — SHA-256 7dba6319c828230ca338d885c3c3a6aec0fa44ef76450eaedb74b2ee41871eb5