Research & Debate · Research · intermediate
AI astrology, and better questions
A model can imitate a reading. It cannot take responsibility for one.
6 Sept 2026 · 1 min read

Language models can produce fluent Jyotish. Fluency is not a parampara. This archive uses machines as clerks — to file, to search, to remember a house list — not as oracles.
The danger is not that a model is 'wrong.' The danger is that it is smooth: a verdict without a badge, a time without a source, a diagnosis in a borrowed coat. CapSage would rather be slower and named.
Better questions outrank faster answers. The map is still a map.
Related

Research · Research
Confirmation bias in the chart
You will find the story you brought. Method is how you notice.
4 May 2026 · 1 min read

Research · Research
Eclipses are not verdicts
A nodal season is weather. Headlines prefer omens.
21 Mar 2026 · 1 min read

Research · Research
Birth-time reliability: five honest badges
Verified, highly rated, reported, disputed, unknown. Never present a guess as marble.
4 Sept 2026 · 1 min read
6 Sept 2026 · 1 min · CapSage Astro