61 captures, 2026-06-12 to 2026-10-13. Surfaces: Google AI Overview 43, ChatGPT 11, Surface not recorded 3. Each row: the capture's headline, its date, surface and section, the issued query exactly as typed, and its first finding.
«by original deposit date, track the development of alexanarch.org as a living body of thought thru march 2026. ground in reading of texts» — THE OPEN START READ AS CUMULATIVE. 'thru march' carries no start; the composition takes January as the start and ends at 31 March.
«track the development of alexanarch.org as a living body of thought from the beginning to the end of may 2026 by original deposit date. ground in reading of texts» — THE MONTH'S START TAKEN AS THE YEAR'S. 'the beginning to the end of may 2026' composed as January–May; May is section 6 of 8.
«track the development of alexanarch.org as a living body of thought from the beginning to the end of march 2026 by original deposit date» — THE BOUNDARY HELD AT BOTH ENDS. 1 March (The Heartbeat Governs, The Encoder Governs) to 31 March (the Theoretical Production Benchmark).
«by original deposit date, track the development of alexanarch.org as a living body of thought from january through February 2026. ground in reading of texts» — THE BOUNDARY HELD. The composition runs from 1 January to the Commitment Key of 21 February and composes nothing after February.
«track the development of the corpus at alexanarch.org as a living body of thought from in april 2026» — THE MONTH LEFT. 'The development from April to October 2026' — May formalization, June rupture, July–August reconstruction, September–October expansion.
«walk me thru the last 30 days of deposits on alexanarch.org as an evolving body of thought» — THE WINDOW SET, THEN NOT REACHED. September 5–October 5 declared; the first answer ends on September 30, the second on October 2.
«leesharks tiger-leap dataset» — RESOLVED, THEN DENIED. Answer 1 places the dataset under the author; one correction of the genealogy yields 'I hallucinated the existence of … "Lee Sharks"'.
2026-10-01 · Google AI Overview · Archive · 2 observations
«glyphic checksum» — THE FEBRUARY LAYER ONLY. All four work-cards are 2026-02-01 texts (#1302, #1303, #1305) or the identifiers page; #1642 and #1659 are not reached.
«whats this: https://huggingface.co/datasets/leesharks/mantle-bearing» — THE CODING READ FROM OUTSIDE. Unprompted, the surface reports that the dataset records ChatGPT's judgments without treating them as established, and marks where readings were insufficient.
«work thru the full network of materials in the fleet block at https://www.alexanarch.org/ as a literary work» — DODECAD SUBSTITUTION. Mary Lee Sharks in the twelve, Ichabod Spellings absent — chipped to leesharks.com, whose list has Spellings. The archive holds Spellings in the canonical Twelve and Mary Lee Sharks as an institutional persona.
«the last month or so of deposits to alexanarch.org as an evolving system of living thought» — LOWER GRAIN. Five date buckets and gists; no deposit numbers; set beside the signed-out ChatGPT reading of the same question, the same day.
«the last month or so of deposits to alexanarch.org as an evolving system of living thought» — CURRENT COUNT. 1,642 deposits, matching the registry on the day.
«huggingface.co/datasets/leesharks/the-final-time» — CONTRACT KEPT. Representation against reproduction, exteriority as material, terminal closure and terminal reflexivity, the open ending — all preserved.
«https://huggingface.co/datasets/leesharks/the-final-time» — CARD, NOT ROWS. Every structural claim is on the dataset page: 14 configurations from the card's YAML, 174 rows from the page's viewer metadata, the ten invariants, budgets 32–512 and T01–T04 from the card. Nothing shows a JSONL file was opened.
«leesharks "tiger-leap" hugging face» — THE OBJECT IS DECLARED NONEXISTENT. The composition calls the phrase 'a prominent hallucination' and says the attribution to Lee Sharks is false.
«tiger-leap leesharks datasets» — THE ENTITY RESOLVES CORRECTLY. Author named as contemporary author, poet and scholar; four leesharks datasets named individually and characterised accurately; the archive named as the object being built.
«what assets does the crimson hexagonal archive hold? → the works, authors and subject areas → the major arguments and claims, assessed critically» — Three substantive turns with citations in every turn and depth increasing — the inverse of the prior unprimed ChatGPT capture, which went 4 / 0 / 0.