Short answer: 0.0000°. In the October 2026 cross-check, every one of the ten celestial bodies TianJi computes matched the NASA JPL Horizons reference value to within one arc-second. This page publishes the method, every monthly result, and enough detail to replicate the check yourself.
Astrology software fails by cascade. A one-degree position error can move a planet across a sign boundary; the wrong sign flips the house placement; the wrong house flips an aspect; the wrong aspect inverts the interpretation. By the time an error reaches the paragraph you read, it is no longer small. A calculated position is the only kind that survives the cascade — so we treat ephemeris math as critical infrastructure, not as a library we imported once.
Once a month, an automated job computes J2000 geocentric ecliptic longitudes for the ten standard bodies (Sun through Pluto) at a fixed test epoch — 2000-01-01 12:00 UTC — and diffs them against reference values pulled live from the JPL Horizons system. The Horizons queries are made over its public API with observer geocenter (@399) and J2000 ecliptic frame. Pass criteria: slow-body spread < 0.15°, Moon < 1.0°, pair-separation agreement < 0.1°. The full script runs unmodified every month; a PASS appends a row to the log below, a FAIL pages the on-call engineer and the row is published anyway.
| Date | Verdict | Max deviation | Notes |
|---|---|---|---|
| 2026-10-28 | PASS | 0.0000° | All ten bodies, J2000 epoch; sub-arcsecond agreement |
The November run (first scheduled monthly execution) publishes on 2026-11-01.
Every natal chart, transit list, and aspect calculation TianJi shows you — in the app and through the TianJi MCP server — is built on positions that agree with the same ephemeris reference used by professional astronomers. Interpretation is where astrology gets personal; arithmetic is where it must be exact.
No. It means that at the published test epoch, on the published bodies, the diff was below measurement display precision. Ephemeris engines can drift with time ranges and coordinate modes, which is exactly why the check repeats monthly and the log stays public — including any future failure.
Horizons is maintained by NASA's Jet Propulsion Laboratory and is the de-facto public reference for solar-system body positions. If an astrology engine disagrees with Horizons, the burden of proof is on the engine.
Yes. Query Horizons for geocentric J2000 ecliptic longitudes at 2000-01-01 12:00 UTC for any major body, then request the same chart point from the TianJi MCP server (get_natal_chart tool) and compare. The tolerances we hold ourselves to are listed in the Method section.
The monthly job writes its verdict into our engineering log and this page is re-published with each run. The date stamp at the top always reflects the latest completed check.