Friday, July 10, 2026
Don't guess someone's sky when you can compute it.
Made Ask Sage actually know your chart. Instead of guessing from retrieved text, it now routes your question, computes your real transits in your own zodiac frame, grounds the answer in the source corpus with citations, and replies like a teacher instead of an essayist.
Shipped
- Ask Sage, grounded in real math — added an intent router that sorts a question into three lanes (personal, educational, informational), and for the personal lane it grounds the answer in your actual computed transits — house-aware, in your own Vedic or Western frame — rather than guessing from retrieved snippets
- Cited, teacher-voiced answers — every answer now grounds in the source corpus and cites it, and renders as a short outline — a lede and a few bullets — in a concise teacher voice, instead of a wall-of-text essay
- The first Sage skill — shipped the weekly-transit skill, the first real capability the Sage agent can run on its own, forwarding the "coming weather" from a real sky-events calendar
- Full-page chat — rebuilt Ask Sage as a full-page chat that fills the viewport with a pinned composer and a lane picker of starter questions, and gated the ask to signed-in members
- The Almanac program board — built a socials program board that lays out post formats across the week — roundups, spotlights, by-sign notes, sky notes — and can compose X posts straight from ingested videos
Notes
The line I kept holding today: don't guess what you can compute. A lot of "AI astrology" is a model free-associating over retrieved text, and it sounds fine until it tells you about a transit you're not having. Computing the querent's real transits — in their own zodiac system, aware of which house it lands in — and then grounding the language in a cited corpus is the difference between a party trick and a tool. The teacher voice is part of that honesty too: an outline with a few clear points admits the limits of what it knows, where an eloquent essay just papers over them.