## 10. Open Questions ### Resolved (decisions made — see relevant sections) - ~~Sync architecture~~ → cr-sqlite + iroh selected (section 3) - ~~Minimum hardware specs~~ → 8GB RAM, 2020+ CPU (section 3) - ~~CRDT library evaluation~~ → cr-sqlite for SQL-level CRDTs, iroh for networking (section 3) - ~~Whisper model selection~~ → ggml-base.en desktop, ggml-tiny.en mobile (section 3) - ~~LLM model selection~~ → Phi-4-mini (8GB), Qwen 3 7B (16GB), Llama 3.2 1B (mobile) (section 3) - ~~Waitlist tool~~ → LaunchList £65 one-time (section 7) - ~~Payment processor~~ → Lemon Squeezy 5% + 50p (section 7) - ~~Pricing validation method~~ → Van Westendorp survey via Tally (section 5) - ~~Bionic Reading implementation~~ → CSS regex (bold first N chars), text-vide npm package or custom, MIT licensed - ~~Nudging delivery mechanism~~ → tauri-plugin-notification + Web Audio API chimes + context-aware suppression (section 4) ### Still open - Exact free tier limitations (number of tasks? transcriptions? time limit?) - Which frontier AI model for cloud tier (Claude, GPT-4o, other?) - App store submission timeline and requirements for Android/iOS - Sensory preference persistence ("sensory cookies") — save user's baseline motion/contrast/typography settings across sessions. MVP or v2? - Emotionally adaptive UI (detect frustration/fatigue, auto-simplify interface) — v2+ feature, but worth architecting for early - Mac App Store sandboxing compatibility with Tauri — needs hands-on testing - Access to Work approval process for specific software products — exact requirements TBD - Search volume data for "ADHD desktop app", "ADHD app for Windows" etc. — validate with Ahrefs/SEMrush before committing to SEO strategy - Tiimo's B2B pricing (not publicly available) — competitive intelligence via test enquiry - Visual timeline implementation — time blocks, Gantt-style, or simpler countdown view? Validate with beta testers. - Smartwatch integration for haptic nudges — Tauri v2 wearable support? Or companion app? - Low-fi body doubling: would showing anonymised user count ("3 others in deep work") require any server component? Could use iroh peer count from paired devices, but broader anonymous count needs a lightweight coordination mechanism. - Start/shutdown ritual UX: how prescriptive should the morning triage be? Fully AI-driven or user-guided? Beta test both approaches. - cr-sqlite development pace risk: monitor vlcn.io activity. If stalled, migrate to Automerge + SQLite BLOB storage (networking layer unchanged).