Source
Corpus collected from 446 Chiron Intuition transcripts through July 7, 2026. Every quoted line is verbatim and links to its source video. Links are
video-level, not timestamped: the corpus stores the video id but not the moment, question, surrounding cards, or original card order, so those are not shown. A grounding pass confirmed each read traces to its cited video.
What this page mixes (three sources, labeled)
The page is not a single voice. It carries three kinds of text, kept distinct:
- Verbatim Chiron quotations — the pair passages and the raw card mentions, each source-linked.
- Chiron's curated meaning — the keyword line and one or two sentences per card, from her own cheatsheet, plus a clarifier note when one names the other card.
- Automated summaries — the blocks marked "Automated corpus summary" are generated from keyword tags, not quoted from Chiron. Each states the rule that produced it.
A fourth, outside layer is the Bunning structural label. It is not Chiron's, so it sits behind an "outside comparison" toggle under each pair's evidence, never ahead of it.
The constructed corpus reading
The "Everything in one read" box builds a plain reading from fixed templates over verified fields, not generative prose. Each card gets a
position role (where it begins, what develops, where it heads) and its cheatsheet meaning. Each connection gets one
bridge sentence, and every bridge is sourced from Chiron's own material: the cheatsheet clarifier line when one card names the next, otherwise a statement of the recorded corpus evidence for that pair (direct, clarifier, or sequence passages). Where Chiron records no connection, the bridge names the entered sequence and the evidence strength, and claims no interpretive transition.
Bunning never supplies a bridge: an outside framework does not narrate a reading built from her corpus, so the structural label stays under each pair's "outside comparison" toggle and in the evidence view. Three layers are switchable:
Plain reading (templates only),
Evidence view (the meanings, bridge source, relationship, and evidence counts behind each line), and
Raw corpus (every quotation, source link, and alternative). The box is labeled "constructed corpus reading" so it is never mistaken for Chiron's continuous words. A fully hand-authored bridge sentence for every one of the 3003 pairs is a curation task, queued; today's bridges are template-derived from verified fields.
Topic focus (removed)
A "focus on a topic" box was tried and pulled on 2026-07-15. It was substring keyword search: the word "work" matched the verb, not a job, and it surfaced passages from card pairings that were not even in your spread. The corpus carries no per-passage topic tags, so a topic filter cannot honestly show "her words on your topic for your cards." It would need topic tags added during curation.
How order is considered
The stored corpus is
unordered co-occurrence: it records that two cards appear together, not which was drawn first. To honor order anyway, the page reads each passage and checks
which card Chiron names first in her own words. For a pair in your order, passages where she says the first card before the second lead the result, and swapping the two flips it. This is an
inferred approximation, not a stored spread position, and it only works when both card names appear in the passage. True ordered syntax (with each card's stored position, timestamp, and question context) still needs the transcript rebuild, which stays queued.
Coverage
This build
finds at least one corpus passage for 2822 of the 3003 possible unordered pairs (94%) across the 78 cards, from
26,169 combination passages. Most of that is loose co-occurrence, not explicit combination reads: only
1,892 passages (7%) are direct, while nearby mentions are about 42% and flagged material about 35%. So "recorded evidence" here means recorded co-occurrence at some strength, not a confirmed combined interpretation for every pair. A pair she was never recorded reading shows
“No recorded Chiron pairing found”, and now offers an optional
constructed reading built by AI in her voice from her own card meanings, clearly labelled as constructed and kept out of her quoted words. See the next panel.
Constructed readings (AI, for uncovered pairs only)
Where Chiron has
no clear recorded combination for a pair, meaning no unflagged direct, clarifier, or sequence passage, the page can show an
optional constructed reading in a dashed, labelled box below the evidence. It is generated by AI (claude-opus-4-8) in her voice, built
only from her own verbatim cheatsheet meaning for each of the two cards, and the two source meanings are shown beside it. The individual card meanings are hers and attested; the way they combine is constructed, never presented as her recorded words and never mixed into her quoted evidence. This targets the
569 pairs with no clear combo (181 with nothing recorded, 388 with only nearby mentions); the
2,434 pairs she did record are never touched or overwritten. Constructed entries live in a separate file (
ai_readings.json) so her corpus stays exactly as transcribed, and a pair with no authored reading keeps its honest empty state rather than showing a placeholder.
Preservation audit
An acceptance test (
audit.py, shipped beside this page) checks that every corpus record reaches a rendered result. Result:
26,169 combination passages (direct 1,892; sequence 7,334; clarifier 5,966; nearby 10,977), of which
16,991 are clean and
9,178 are flagged (fragment, negative, third-card, or near-duplicate). Plus
3,523 standalone card mentions. Every passage is reachable through the pair's grouped "view all" list, every mention through each card's raw toggle, and
181 pairs are confirmed gaps with no passage.
Unreachable records: 0.
The structural lens (Bunning)
Each pair is classified against Joan Bunning's
Learning the Tarot. Two labels come from a
verified relationship table (
bunning_relations.json) extracted from her per-card Reinforcing and Opposing Cards pages plus her Lesson 15 named pairs; two are computed from card identity, which Bunning treats as reliable; the rest default to Occasional:
- Opposing pair — verified: one or both cards list the other on their Opposing Cards page in Bunning's book (for example the Magician and the High Priestess).
- Reinforcing pair — verified: one or both cards list the other on their Reinforcing Cards page (for example the Empress and Nine of Cups). Same suit alone never qualifies.
- Unclassified — a Major Arcana pairing whose Bunning relationships are not yet extracted. Kept distinct from Occasional: it means verification is pending, not that no relationship exists.
- Ace pair — computed: both cards are Aces.
- Court pair — computed: both cards are court cards.
- Occasional pair — two Minor Arcana cards where neither lists the other in the extracted references, so there is no named relationship. Any two cards can still be read together for one reading.
Coverage note: the verified table currently covers the Minor Arcana card pages plus Lesson 15's named Major pairs.
One-sided listings are included, not excluded: a pair is added if either card lists the other, since Bunning's card pages are sometimes directional. Minor pairs with no entry fall to Occasional; Major pairs not yet extracted fall to Unclassified, never to a false "no relationship." Completing the Major Arcana extraction is queued.