What inspired me to build Dayou?It actually didn’t start as another “AI fortune-telling app.”My PhD is in computational immunology at NUS. Every day I deal with noisy biological signals: one marker can be a false positive, two might be coincidence. Only when multiple independent dimensions converge do you trust the signal.Then I saw how most AI metaphysics tools work: one LLM takes your Bazi/Ziwei and spits out a perfectly smooth story. It sounds complete, but you have no idea where the real signal ends and where the model is just filling in the blanks. That single-voice problem felt exactly like the noise I fight in research.So I asked: what if we treated Chinese metaphysics the same way we treat noisy biology? Not one mirror, but many — Bazi, Ziwei Doushu, Xiao Liuren, Liu Yao, Qimen, classical texts, and master-verified cases — all cross-checking each other in real time. Where they converge, where they diverge, and where we should honestly say “this part is uncertain.”That became Dayou: not another prediction machine, but a private, long-term AI mirror that speaks the language Chinese users already trust, for the real situations we actually face — career moves, relationships, timing, life decisions.The first paying user (annual Pro, 1,380 RMB) put it best: “我真的好想跟大有 AI 聊天。” She didn’t buy a chart. She bought an ongoing relationship with a trusted mirror.What problem were we solving? AI fortune-telling today is either rule-based calculators (cold, templated) or generic LLMs (hallucinating with zero grounding). Users want depth, continuity, and intellectual honesty — not entertainment. We’re solving for that missing “multi-signal consensus” layer.How did the process evolve? I’m a solo founder running everything with AI agents (Claude as supervisor + adversarial reviewer, Codex as executor). The workflow itself evolved massively: I started by building the sustaining system first (status tracking, parallel agent pipelines, daily review loops) instead of rushing features. Agents challenged my premises, turned repeated founder instincts (like reverse timing) into flagship features, and forced us to measure real user signals instead of building in a vacuum.A master practitioner reviewed every prompt template for traditional accuracy (only two commits of pure domain knowledge). We shipped, got real payments (and painful Stripe friction from China cards), listened, and kept iterating. The product today is the direct result of that human-AI operating system — agents execute fast, I keep judgment at the taste, trust, and truth boundaries.If you’ve ever wanted an AI that actually reflects with you using the frameworks you already live by, come try it: https://www.dayou.dev/Would love to hear your thoughts — especially if you’ve felt that “single voice” problem in other AI tools. What’s one life question you wish you could cross-check across multiple mirrors?

Dayou

Private AI companion grounded in the Chinese metaphysics

Private AI companion grounded in the Chinese metaphysics