Note: the product source code is private. This repository is the program management layer of LingoDeck, a real product I lead end-to-end. It exists to show how I run a program: roadmap, specs, risks, decisions, and release gates, with AI agents automating the routine parts.
LingoDeck is a mobile flashcard app for learning American English with a focus on moving vocabulary into active memory (spaced repetition + active recall), not just passive recognition.
Key capabilities:
- Flashcards with IPA transcription and native audio (American English)
- Translations into the learner's native language (user-selectable)
- Phrases and collocations, not only single words
- Topic packs: real-life scenarios: doctor visit, dentist, therapist, astrology, professional vocabulary (TPM/PM/PMP), and more
| Artifact | Where | Cadence |
|---|---|---|
| Roadmap | ROADMAP.md | Reviewed monthly |
| Feature specs | specs/ + specs/INDEX.md | Per feature, AI-drafted → human-edited |
| Decisions (ADRs) | decisions/ | Per significant decision |
| Risk register | risks/RISKS.md | Reviewed bi-weekly |
| Release gates | releases/ | Per release, go/no-go checklist |
| Status reports | GitHub Issues (label status-report) |
Weekly, auto-generated |
Specs, risk drafts, and weekly status reports are produced by Claude Code subagents from my pm-agent-toolkit, then reviewed and owned by me. The agent proposes; the PM decides.
Backlog and milestones live in GitHub Projects on this repo.