A reading arrives looking finished. Named parts, a role for each, a fear, a protective strategy, a gift underneath. It reads like a conclusion, and conclusions are easy to accept without asking how they were reached.
So here is the walk from one end to the other — geometry in, named parts out — including the places where the process is making a judgment call rather than a measurement. Those places are the interesting ones, and leaving them out would make this a brochure rather than an explanation.
Step one is not interpretive at all
The pipeline starts with arithmetic. Date, time, and place produce planetary positions along the ecliptic, house divisions, and the angular relationships between every pair of points. Run it twice and you get the same numbers. Run it on someone else’s data and you get different ones.
This step deserves no credit for being objective — it is objective in the way a ruler is. But it does establish something worth holding onto: the input to everything downstream is fixed, personal, and not derived from anything you told us about yourself.
You cannot flatter a chart. It has no idea what answer you were hoping for.
Translation: where meaning enters
The second step maps geometric configurations onto psychological themes — control, vigilance, longing, withdrawal, intensity, exposure. That mapping lives in an ontology layer: structured data encoding what a given placement or relationship has long been taken to represent, expressed in an IFS-informed vocabulary rather than a traditional astrological one.
This is the step where interpretation happens, and we would rather say so plainly than let it hide inside the word “engine.” No amount of downstream arithmetic converts an interpretive mapping into a measurement. What it can do is apply that mapping consistently, to everyone, without the drift a human reader would introduce over a long afternoon.
Consistency is not the same as correctness. It is, however, the only thing that makes correctness discussable.
Scoring: not every signal counts the same
If every mapped theme were weighted equally, the output would be mush — everybody would score moderately on everything. So themes are weighted, along a few axes:
- Exactness. A relationship that lands close to precise carries more weight than one that is barely within range.
- The points involved. Some points in a configuration are structurally more central than others, and a theme carried by a central point counts for more.
- Convergence. The heaviest factor. When several independent parts of a configuration point at the same theme, that theme is treated as strongly present. One signal is a hint. Five signals saying the same thing is a structure.
Convergence deserves the emphasis it gets, because it is the only one of the three that guards against a specific failure. Any sufficiently rich symbolic system can find support for almost any claim if you go looking one signal at a time. Requiring several independent points to agree before a theme is treated as strong is a way of making the system harder to talk into things — including harder for us to talk into producing the flattering, agreeable output that a reader would enjoy most.
The output of this step is not a portrait. It is a ranked list — some themes barely registering, a handful showing up repeatedly, most sitting somewhere unremarkable in between. Most of what a chart contains is, correctly, discarded as noise.
Clustering: from themes to a named part
A ranked list of themes is not yet a part. A part is a coherent role — something with a fear, a strategy, and a reason. Clustering is the step that assembles one.
Related themes that reinforce each other are grouped. A cluster built around control, vigilance, and fear of exposure has the profile of a manager. A cluster built around intensity, withdrawal, and overwhelm has the profile of a firefighter. A cluster organized around old longing or shame has the profile of what those protectors are working around.
Only the strongest clusters surface as named parts. This is a deliberate choice about how much to say: a system could report thirty weak clusters and technically be more complete, and the result would be useless. A description you cannot hold in your head does not help you notice anything in real time.
The hard part of a reading is not what to include. It is what to leave out and still be honest.
Where this stays uncertain
Four genuine soft spots. None of them are secret, and none of them are resolved by adding more computation.
The ontology is a set of choices
Someone decided which psychological theme a given configuration maps to. That decision draws on a long interpretive tradition, but it is still a decision, and a different set of reasonable decisions would produce a somewhat different reading.
The weights are judgments
How much more an exact relationship should count than a loose one is not a fact waiting to be discovered. It is a tuning choice, made to produce output that is differentiated and legible, and it could sensibly have been made differently.
Clustering encodes a model
Grouping themes into manager-shaped, firefighter-shaped, and exile-shaped clusters assumes the IFS categories are the right containers. We think they are unusually good ones — they are legible, they map onto things people recognize immediately, and they carry a non-shaming stance toward behavior. But they are a lens, not a taxonomy of nature, and a different psychological frame would carve the same themes differently.
Language is lossy
The final step turns a scored cluster into a paragraph a person can read. Every such translation adds specificity that the underlying data does not strictly contain. That specificity is what makes a reading usable, and it is also where a description can overshoot.
Precision in the arithmetic does not transfer to the meaning. A number computed to four places is still a symbol underneath.
What a score is for
Given all of that, what should you actually do with a strongly scored part? Treat it as a well-constructed hypothesis about where to look, and then check it against your own life, which is the only evidence that counts here.
If a described protector matches something you recognize — the specific move you make when a conversation gets uncomfortable — the score did its job, which was to point. If it doesn’t match, the right response is to set it down. A reading is not a result about you, and it is not a diagnosis of anything. It is a structured place to start looking.
The score points. The recognition is yours, and it is the only part of this that verifies anything.
The full pipeline is described in how we calculate your reading, and the position on symbolic versus causal claims is set out in what astrology is and how we use it. Or you can run it on your own data and judge the output directly.