Foresee / 境见 · 2025

Ancient wisdom, now reasoned.

I Ching divination and BaZi reading, powered by an AI reasoning engine over a self-iterating match weight — 100,000 words of classical theory, distilled into 1,500 curated cases.

CHAPTER 01 · 六爻

Six lines, ten thousand interpretations.

Cast a hexagram, then read it. The AI reasons over 100,000 words of classical Liu Yao theory, distilled into 1,500 curated cases — every interpretation is a chain of explicit inferences, not a one-line oracle.

The cast — coin throw, six lines, moving lines — produces the same 本卦 → 变卦 structure a practitioner would draw on paper. From there, the AI traces palace assignment, six relatives, hidden gods and the question's relation to each line — and reads them in order.

The 1,500 cases aren't a lookup table. They're worked examples the model reasons over: when the same configuration recurs, the answer cites the precedent; when it doesn't, the model derives one and shows the steps.

CHAPTER 02 · 八字

Eight characters, one life chart.

Birth time becomes a chart of heavenly stems and earthly branches. The AI reads the chart against classical BaZi texts — four pillars, five elements, ten gods — and turns the configuration into prose you can actually act on.

Charting is mechanical — pillar of year, month, day, hour. Interpretation is where the classical texts disagree most, and where the model earns its weight: it carries the disagreement instead of flattening it, surfacing 三命通会 · 滴天髓 · 子平真诠 in the same reading when they pull in different directions.

BY THE NUMBERS

An engine, behind the reading.

100K
SOURCE WORDS DISTILLED
1,500
CURATED CASE STUDIES
SELF-ITERATING WEIGHTS
CHAPTER 03 · 自优化

The engine grades itself.

Low-match readings flag themselves. A nightly job re-weights the case-matching algorithm — the engine that interprets the hexagram is also the engine that improves it. w' = w + η · (target − match)

Every reading carries a match score against the closest precedent case. Scores below threshold are queued; the queue feeds an offline optimizer that nudges the weights — palace, season, ten-gods, hidden stem — until the same hexagram, asked again, matches more confidently.

The model never re-trains on user data. What iterates is the matcher, not the reasoner — so accuracy compounds without the classical theory drifting under it.

ROLE
FOUNDER · DESIGNER
YEAR
2025
TEAM
SOLO · AI-ASSISTED
DOMAIN
DIVINATION · AGENTIC
STATUS
PROTOTYPE · ACTIVE

The classics were never short on theory — only short on patience. An AI that reads the way a practitioner does, and grades its own confidence, is the rare case where machine reasoning lengthens a tradition instead of flattening it.