AI + classical Jyotish: how answers stay grounded

Large language models are remarkable mimics. Ask one about your Saturn transit and it will produce something fluent, plausible, and quite possibly untethered from your actual chart. That fluency is exactly the problem.

The failure mode we’re avoiding

A generic “AI astrologer” works like this: you ask a question, the model recalls what Saturn transits generally mean, and it writes you a reading. The result sounds like astrology but isn’t your astrology — it never computed where your Saturn is, which houses it rules, or what period you’re running.

It’s the horoscope-column problem, just faster.

What grounding looks like

In TevaPatra the ordering is philosophical, not just technical: the chart exists before the answer does. Your placements, your running dasha, the day’s panchanga — computed, real, yours. The AI’s role is interpretation and language: turning “Saturn antardasha lord in your 10th, in debilitation, during Sade Sati” into plain language you can act on.

What it can’t do is freelance. If your chart doesn’t contain a claim, the answer doesn’t get written — the model is a narrator for computed truth, not a generator of it.

Why we think this is the honest way to build it

Astrology, done properly, is a system of correspondences with centuries of accumulated interpretation. An AI that respects that corpus — and stays inside what your chart supports — can be genuinely useful. One that freestyles is just a horoscope machine with better grammar.

We’d rather give you less certainty and more truth.