AI & Safety
Private beta · Version 2026-07-26
The calculation engine selects structured hypotheses. External AI, when enabled, is used to rewrite those hypotheses into clearer language. AI does not prove that an event occurred or will occur.
Local and external modes
Local mode: report wording is produced by deterministic project rules. The current question is processed by this server and is not sent to an external AI provider.
External AI mode: when you turn it on, the structured calculation package and current input are sent to the configured provider for wording. A quality gate and compliance audit run before output is returned. Provider responses may still be incomplete or wrong.
How traditional systems are used
- Your own wording and current reality remain the highest-priority signals.
- Meihua, Qimen, Liuyao, Bazi, astrology, and blood-type folklore are hypothesis-generating tools, not scientific proof.
- Uncalibrated Qimen and Liuyao structures are retained as context and do not change production event weights.
- A report shows no more than three event hypotheses and asks you to lower any hypothesis that does not match reality.
Safety controls
- High-risk crisis text is intercepted instead of receiving a symbolic report.
- Medical, legal, financial, tax, emergency, and guaranteed-outcome language is blocked or audited.
- Internal formulas, pseudo-probabilities, and deterministic destiny claims are not shown as user facts.
- Suggestions favor small, reversible actions and observable reassessment conditions.
How to use a report
Treat each event as a working hypothesis. Compare it with concrete records, messages, dates, transactions, and other people’s actual statements. If it does not match, mark it as not happening and do not extend the interpretation. Stop and seek qualified help whenever the issue involves health, safety, law, money, or immediate risk.