01
Calculations are traceable
The system should keep enough trace data to explain why a chart, hexagram, direction, or entitlement exists.
Trust and safety
Methodology and Safety
This product treats calculation facts, model interpretation, privacy, and payments as separate responsibilities so a reading can be useful without becoming opaque or coercive.
MVP status: public safety doctrine is live and linked from every P0 tool page; deeper audits continue as engines, auth, and commerce ship.
02
Advice has guardrails
Readings should avoid certainty, fear, dependency, medical claims, legal advice, financial directives, and safety-critical decisions.
03
Payments are recoverable
Paid work must keep status, entitlement, and provider failure paths clear so a user is never left charged without a usable recovery path.
How shipped flows are reviewed
01
Define acceptance first
Each feature gets plain-language verification scenarios before code is written.
02
Separate facts from narrative
Shared libraries produce facts; app routes and model calls consume those facts.
03
Verify before tests
Browser walkthroughs check the real UI before Playwright selectors are finalized.
Product boundaries
This service is not medical, legal, financial, emergency, or mental health advice.
No reading should encourage secrecy, dependence, panic, or irreversible action.
Users should be able to delete or stop account and email flows without dark patterns.
Public SEO, sitemap, robots, and llms files must exclude private account surfaces.