Methodology
What each check does, which sources it stands on, and — just as important — what it cannot do. A method that hides its limits isn't a method; it's a sales pitch.
Four Pillars (saju). The year, month, day and hour pillars come from a Korean almanac that respects solar-term boundaries and converts lunar dates. Each name's spoken sound is mapped to one of the five elements by where it's made in the mouth — the same rule Korean naming practice applies to Hangul initials — and then compared with your day master and with the element your chart has least of. Two schools disagree about lip and throat sounds, so we show both and score the more conservative one.
Birth chart. Tropical positions of the Sun through Pluto from astronomy-engine, weighted toward the Sun, Moon and rising sign. The rising sign needs a birth time and a city — without them, it's left out rather than guessed. The name's Pythagorean numbers (expression, soul urge, personality) are matched to the elements of your Sun and Moon.
Family. Each name against every person you add: the day-master relation, whether the name brings the element their chart is missing, and how the two names sound together.
The limits, plainly. The Four Pillars use the local clock time you enter; for births far from Korea, the hour pillar can land one slot off near a two-hour boundary. The Latin-letter sound mapping is a heuristic, not a lineage school. All of it is a tradition read for fun — it says nothing about anyone's future, and the ranking blends it fifty-fifty with the practical checks below.
A deterministic engine, and nothing else. Every axis below is a published rule applied to your candidates — feed the same names in twice and you get the same grades twice. There is no reviewer whose mood, taste, or bad afternoon can move your result.
One AI is involved, in one place: it reads the finished verdict and writes the summary paragraph in plain language. It never sees our internal scoring, and it cannot change a grade or a level — if it tries to invent a number or predict your child's future, the paragraph is discarded automatically and replaced with rule-generated text. Every report states which of the two wrote it.
We think this is more accountable than a signature, not less. A named expert asks you to trust a person you cannot audit. We would rather publish the rules, let the result be reproducible, and refund in full whenever we get a fact wrong.
No decimal scores. "87.4 points" looks precise, but the number moves with whatever weights we chose. Unearned precision is theater, and you'd be paying for theater.
No percentiles. "Top 8%" requires a sample we'd have to invent. Every verdict is three levels — Good, Caution, Risk — with the evidence written in full sentences underneath.
"Can't assess" is an answer. Where our data doesn't reach, the report says so instead of guessing.
Nothing found that follows a child around
No Risk verdicts, at most one Caution.
Worth reading — no dealbreakers
Cautions exist; none would make us re-open the shortlist.
Read the flagged section first
A Risk on teasing or spelling — the axes a child lives with daily — caps the grade regardless of everything else.
Axis weights exist but are not published — publishing the formula invites cloning, and the moat is the data, not the arithmetic. That is the one deliberate exception to full transparency, and we'd rather name it than pretend otherwise.
