COUNTERPOINT

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for COUNTERPOINT

Inspiration

Most matching systems ask who is similar to you. COUNTERPOINT asks what another person might bring that you naturally lack. The idea is to translate an ancient relationship framework into a modern, transparent experience without turning culture into a verdict.

What it does

COUNTERPOINT generates a visible Five Elements pattern from a UTC+8 birth moment, explains why demonstration profiles may offer a counterpoint, and lets two people compare their patterns in a private browser session. Instead of a compatibility score, the product shows Person A, Person B, and the shared pattern. An optional AI layer translates only the supplied evidence into calm, non-deterministic language.

How we built it

We separated the system into three layers: deterministic chart calculation, inspectable complement evidence, and optional AI interpretation. React manages the private in-browser interaction, lunar-javascript calculates the Four Pillars, and DeepSeek explains a locked structured evidence object from a server route. Automated tests protect calculation boundaries and prevent the model layer from replacing evidence.

Challenges

The central challenge was building cultural intelligence without false certainty. Timezone, solar-term, and school conventions can change a Bazi chart, while generative models can make unsupported claims sound authoritative. The prototype makes its UTC+8 boundary visible and restricts AI to summarizing deterministic evidence.

Accomplishments

• Public end-to-end MVP
• Distinctive editorial relationship experience
• Visible personal, complementary, and shared elemental profiles
• Explainable recommendation evidence with no compatibility percentage
• Privacy-aware browser-local state
• Reproducible setup and automated regression tests

What we learned

A cultural system can support reflection without pretending to be a scientific predictor. The product becomes more credible when calculation, interpretation, uncertainty, and consent are visible parts of the experience.

What's next

Global timezone handling, independent chart benchmarks, user research, consent-first accounts, encrypted shareable sessions, deletion controls, and founder-owned subscription infrastructure. The planned business model includes a free personal profile, paid Circle sessions, and a Studio tier for facilitators. Accounts, persistent storage, and payments are not live in this prototype.

Built With

• codex
• deepseek
• react
• typescript
• vite