Ordo
overview
what this project was
Ordo is a covenant discipleship app for small groups, built around a single mechanic: simultaneous reveal. Members write their reflections independently within a cycle, and no one sees anyone else's until the window closes and all are revealed at once. The design solves a specific failure in group discipleship, where the first person to answer sets the tone and everyone after them writes a version of that answer.
The whole app is structured on the liturgical calendar rather than on arbitrary weeks, with cycles tied to seasons and scripture readings drawn from the date rather than generic content. Alongside the core reflection loop it carries a rule of life, a personal prayer book, an archive with seasonal recaps, and group configuration controlled by a designated steward.
The problem is one I have firsthand. I struggle with consistency in devotion and in Scripture, and Ordo is an attempt to build the structure I needed rather than the structure that would have been easiest to ship.
role
what you did
Founder and sole developer. Product design, architecture, full-stack implementation, AI integration, visual design, and the marketing site at startordo.com.
team size: solo project
execution
how it was built
React Native with Expo on the front end, Supabase for backend and auth, Drizzle ORM, and Anthropic's API for the AI layer. Roughly sixty edge functions in production, with the schema at migration 0055 and all migrations synced against remote.
The AI integration is deliberately restrained. Groups can opt into dynamic reflection prompts — off by default — which call Claude with a system prompt carrying the group's tradition, tradition weight, the current liturgical season, the cycle number, and the scripture reading looked up by the cycle's own opening date rather than today's. Any failure falls straight through to a curated static prompt bank, so the feature can never break the core loop. That opt-in default was a product decision as much as a technical one: a discipleship tool should not quietly hand formation over to a model.
Development ran in epics through Epic 28, covering the cycle lifecycle, group and steward permissions, the archive and recap system, formation reviews, and end-to-end test hardening.
The marketing site at startordo.com carries the same design language as the app and explains the simultaneous reveal mechanic and the seasonal structure directly, rather than leading with feature lists.
outcome
what happened
Functionally complete MVP through Epic 28, with all edge functions deployed, migrations synced, and zero TypeScript errors. Not yet publicly launched.
Current work is a visual direction shift toward a modern prayer book aesthetic — clean, seasonally expressive, and less component-heavy — alongside a set of UX fixes drawn from early user feedback: simplifying the read-reflections flow, restructuring the folders system, reworking the archive hierarchy, and replacing internal labels like "steward" and "waiting" with language that fits the register.
The next milestone is one real covenant group running a full season, which is the only honest test of whether the core mechanic works as designed.