astrology, like any claim about the world, earns trust by being checkable.
a reading is a structured reflection, not a forecast of the outcome. four commitments keep that distinction honest: claims specific enough to test, set down in advance, with the misses kept in view beside the hits, and every line traceable to a computed chart.
if some of these words half-ring a bell, a phrase from a grandparent, a chart someone cast the week one was born, that is welcome. this is the same body of knowledge, approached the way one might think now: as observations that can be written down, dated, and graded against what actually happens.
1 · specific enough to test
a statement that fits every possible outcome is comforting, but there is no way to tell whether it is right, because nothing could ever show it to be wrong. a good deal of popular astrology is written broadly, and there is real warmth in that tradition. Artiji aims at something narrower: an observation precise enough that a single fact about one's own life could show it to be off. when that happens, it is recorded as a miss rather than smoothed over. below is the same idea twice, once in a broad form that is hard to test, and once in a specific form one could check.
broad"one is a natural leader."
testablethe loudest placement in one's chart is Mangal, the Sanskrit name for Mars, the graha traditionally tied to drive and initiative: forward motion ahead of diplomacy. if one recognizes oneself as consensus-first by temperament, that reading has missed.
broad"a meaningful change is on its way."
testableone's current dasha, a multi-year planetary chapter in the Vimshottari cycle Jyotish has used for centuries, has recently moved from Shani to Guru, the Sanskrit names for Saturn and Jupiter. Shani reads as the slow, consolidating one; Guru, whose name also means teacher, as the expansive one. if the past few years felt like narrowing rather than widening, the dates of that handoff are off.
broad"relationships are important to one."
testablethe seventh house, the sector of the chart traditionally read for partnership, is the strongest angle in one's map. if one's own history says the work has consistently come first, that emphasis is off.
broad"the timing will make sense in hindsight."
testableone's Moon sits in the nakshatra, one of the twenty-seven lunar mansions the Moon travels through, called Rohini, to which tradition assigns a steady and patient temperament. if that is not the person close friends would describe, the placement has been read wrong.
2 · pre-registered, not fitted afterward
a common move, in astrology and in quantitative work alike, is to fit the explanation to events once they have already happened. reading meaning backward is easy, and it feels convincing. pre-registration is the discipline of setting the claim down in advance, so hindsight has nothing left to add. the observation is written and dated before the window it describes begins.
- a note written after a funding round that accounts for the round is easy to produce. a note written before it, naming the window in question, is a real claim one can hold Artiji to.
- each falsification protocol carries an
EP- identifier and a fixed claim, set down before the period it covers. the identifier is the receipt.
- this mirrors how a clinical trial is run: the hypothesis is logged before the data arrives, not after. a backtest one is free to adjust along the way is a much weaker form of evidence.
- the timestamp is open to inspection. if the record shows
EP-014 was registered in March, the commit history shows the same date.
3 · misses are published alongside hits
it is natural to remember the readings that landed and to let the rest fade. a method does the opposite and keeps the misses in view, because the misses are what give the hits their weight. a record of only successes tells one very little.
- when a pre-registered claim does not hold, it is published as a null under the same
EP- identifier, rather than quietly removed.
- showing only the successes is survivorship bias: the readings that missed fall out of frame, and the picture looks stronger than it is.
- a claim that does not land is logged as a miss, rather than reinterpreted afterward as a subtler kind of success.
- for a reader coming from an AI background, the closest analogy is an evaluation set with the failures left in, instead of a highlight reel. the miss rate is treated as part of the result.
4 · provenance
a fair question about any AI astrology is where the claim actually comes from: the chart itself, or the model writing the prose. Artiji fixes that order deliberately, so the two roles do not blur.
- the astrological content is computed from real planetary positions for a given birth moment, using the same ephemeris mathematics an astronomer would use, in the sidereal frame that Jyotish uses. that computation is the source.
- the language model does not originate the claim. its role is to render the computed output into readable English. it is the translator rather than the author.
- the computation is deterministic: the same birth details produce the same underlying chart each time, independent of how the text is later phrased.
- if a sentence in a reading does not trace back to a placement in the computed chart, it is held back rather than published. the model is kept close to rephrasing, so it is not the part supplying the astrology.
a claim specific enough to be wrong is a claim worth trusting when it holds.