Layer one: random draws, fixed determination
Basic draws yin or yang fairly for each of six line positions. The full three-coin method forms each line from the values of three coins. The draw itself contains randomness, but once the lines exist, predefined rules determine how the primary hexagram is formed, which lines are changing, how the resulting hexagram is formed, and which part leads the reading.
AI does not inspect the question and choose a more appealing hexagram. It cannot alter line positions, names, changing lines, or reading priority. This boundary keeps “what was cast” separate from “how it is interpreted in modern language.” An interpretation can be wrong without changing the casting result.
Layer two: classical passages come from fixed data
The site’s classical layer records the names and Judgments of the sixty-four hexagrams, the Image, line texts, the special Use-Nine and Use-Six passages for Qian and Kun, and relevant Sequence material. After a cast, these records are retrieved by hexagram and line position; a language model does not compose, complete, or rewrite them at that moment.
The ancient text of the I Ching is in the public domain. This site uses the received text as fixed reading material and checks obvious variants against publicly available editions when needed. It is a general reading edition, not a scholarly critical edition or the sole authoritative text.
Layer three: translation and editing are not the source text
The interface preserves Classical Chinese as the source-text layer. Modern Chinese expands the older language into more accessible prose, while English presents a corresponding translation or natural adaptation. These are reading aids rather than the original text, and fluent wording should not hide the fact that interpretation is involved.
Some modern Chinese, English translation, and interpretive material was drafted in batches with generative tools, then checked for structure, terminology, Traditional Chinese usage, and Chinese–English alignment. Those checks can catch formatting and language problems, but they do not mean every sentence has been approved by a specialist in the classics or translation. Where wording differs, begin by comparing the clearly labelled source-text layer.
Layer four: Basic is pre-generated; Plus uses live AI
Basic’s four modern reading sections are generated and prepared before any casting, with Chinese and English aligned to the same core interpretation. When you cast, the system retrieves that material instead of calling a language model for you in real time. Basic also has no custom-question field to send to an AI provider.
Plus, which remains in local preview and planning, is the path intended to give a live AI the already fixed hexagram, selected area and time frame, and custom question, then create one aligned bilingual personalised reading. Before any public launch, the interface must explain that the question passes through the server to an AI provider, that AI may be inaccurate, and that its role is interpretation only. The actual provider, retention, and data-use arrangements must also be disclosed accurately.
Scene images and revision: inspiration, not evidence
Reading scenes are artistic interpretations, some created with the assistance of generative image tools. They are intended to create space for pause and association. They do not participate in the cast, prove a historical meaning, or make one hexagram more favourable or dangerous than another.
Translations, modern readings, and visual interpretations may all require revision, and content pages show an update date. Transparency is not about piling up documentation. It shows how each layer was made and how far its authority extends.