Mythsensus

Mythsensus takes one birth date and runs it through 26 ancient divination systems — BaZi, Vedic, Western, Nine Star Ki, Thai, Mayan, Norse runes and 20 more — then returns a single Cosmic Score (1–1,000) measuring how much they agree. Nine of those systems also vote on a rolling 4-week forecast.

The thesis: every astrology app picks one tradition and asks you to take it on faith. The interesting signal is where independent traditions converge — and where they don't. Agreement is signal; conflict is data too.

Everything numeric is deterministic. Same birth date, same result, forever — the engine is plain TypeScript and no model sits in the calculation path. AI only phrases the reading. That line matters: when a number looks wrong it is a reproducible bug, not an argument with a hallucination.

It also ships as an MCP server, so an AI client can call the engine directly instead of guessing about astrology.

Free, no signup for the full 26-system reading. EN/TH.

The engine forked without anyone noticing. The web app and the npm package each carried their own copy, drifted for weeks, and returned different day pillars for the same birth date. One engine now, ported by a build step — never by copy-paste.

Deterministic output turned out to be a testing strategy, not a purity stance. Because the numbers never move, a screenshot from three months ago is still a valid regression test.

A drop that looks like a funnel problem is usually a traffic problem. Weeks went into re-tuning the funnel before checking whether the referrers had simply stopped sending anyone.

Let the model write the prose, never the numbers. Anything you would end up arguing with is something you should have computed.

And keep one short handover file the assistant must read before touching anything — what is done, what is waiting, what not to step on. A fresh session re-deriving your project costs far more than maintaining that page.