Building in public
The story behind the product β pricing decisions, technical choices, and what I learn along the way.
The Math Behind Every Question: How Pricing Works in Astrology Bot
One question β up to ten AI calls. Claude tokens, 24% tax, Astrology API, Stripe fees, infrastructure. I break down every layer and explain the 2.0 multiplier.
From Analyst Agents to Energy Caching: How the Pipeline Evolved
The first version on GPT-5-mini was simpler and cheaper, but unstable. How a single insight about chart immutability turned the architecture upside down and led to four processing flows.
The General Overview Prompt: Why Synthesizing 12 Houses Is Harder Than It Looks
LLMs flatter, try to say everything at once, and start with the wrong thing. Three synthesis problems β and how an anti-compliment rule, topic limits, and identity-first solve them.
Energy Descriptions: One Analysis, Infinite Questions
The most labor-intensive part of the project isn't the code β it's a hundred prompt iterations. How the texts are generated that the system uses to assemble an answer to any question.
Why I Threw Out the Aspects From the Astrology API and Recalculated Everything Myself
The external API provides accurate astronomy, but its orbs and aspect set don't work for me. What we keep, what we discard, and what we recalculate from scratch.
What Kind of Astrology Does Astrology Bot Use
Modern Western school, tropical zodiac, Placidus. Ten planets, five aspects, custom orbs β and why fewer objects means less noise.
How My Own Chart Made Me Step Out of the Shadows
I never planned to become the face of this product. But my own system revealed a pattern I'd been blind to β one that had already played out twice in my life.