The Reading Method: Answer & Specificity Contracts

The pathology of the generic reading

AI-era divination fails in one typical way: the symbol-level analysis is eloquent, and then the advice lands on "stay open-minded," "go with the flow," "trust your intuition" — words that can never be wrong, precisely because they say nothing. Three causes: the reading was never required to answer the question first; the advice was never required to cost anything; and four different systems were sharing one "gentle companion" voice. Ask Akasha's remedy is to write the discipline down as published contracts and run every reading on top of them.

The Answer Contract: answer what was asked

Questions are typed first, and each type carries hard requirements:

The answer comes first, analysis after — your actual question is never buried under a heap of symbolism.

The Specificity Contract: advice takes shape

Symbol to meaning is retrieval (the card-books have it); meaning to application is deduction — the reading must multiply the concrete people, places and stakes in your question by what this spread specifically points at. Core clauses:

• The four-piece advice section: a diagnosis naming your position in this configuration (it must reuse concrete nouns from your question); one principal recommendation; one taboo specific to this situation (named as an action, never "avoid impatience"); one watermark;
• The principal recommendation must place a bet: verb-first, concrete object, nameable deadline, and a cost or exclusive choice ("either… or…", "if…, first give up…");
• The platitude blacklist: dozens of phrases — "go with the flow," "stay open-minded," "work on communication," "focus on self-improvement" — are treated as contract violations; their appearance forces a rewrite;
• Each system generates advice by its own method: tarot by action-mapping, the I Ching by line-oracle transposition, astrology by windows and outlets, Bazi by the useful-god derivation chain — four arts, no longer one shared mouth.

The falsifiability check: the only quality question

The final question asked of every generated reading: cover the card names, hexagram names and placements — could this advice be handed unchanged to any stranger? If yes, rewrite. Only words that could be proven wrong deserve to be called advice. It is the hardest discipline distilled from the great divinatory traditions: masters place bets; mediocrities hand out blessings.

Signs the situation has changed

Every reading ends with a short section: what observable signs would mean the situation has shifted and the advice above has expired. This is the reading's humility and its precision at once — divination should not pretend to be permanently right, and advice that can expire is far more honest than platitudes that cannot.

The factual foundation

Every calendrical and factual element is computed server-side and injected; the AI is forbidden to invent:

• Tarot cards and hexagrams — cryptographic randomness and algorithmic casting; the AI cannot alter them;
• Stem-branch days, moon phases, solar terms, planetary hours — astronomical computation;
• The Bazi passing year — server-given at the Lichun boundary; citation only;
• Astrological transits — a real 14-day sky table; nothing beyond the table may be said.

Versions and evaluation

Every module's system prompt carries a version number, and every revision runs automated evaluation: a fixed question set against real model calls, accepted on four axes — the question answered; the advice shaped (deadline, bet and signal section all present); zero blacklist hits; and pairwise surface-overlap between the four modules' advice on the same question below threshold (the direct measurement of sameness). Only a four-axis pass ships. The ritual lives in the front of the house; the engineering rigor lives in the back — we believe the two are not in conflict, and that neither is optional.