Meeting notesCAN I VIBECODE
Supernormal?
BUILD FIRST
- One specific customer type and one primary use case
- A vertical slice that completes the main job without manual workarounds
- Visible limits, review steps, and a fallback when an external dependency fails
LEAVE FOR LATER
- Broad third-party integrations
- High-scale collaboration, syncing, and permissions
- Automation that can cause harm without human review
RESEARCH NOTES
What this tool actually does.
Supernormal is a bot-free AI meeting notetaker. The desktop app captures audio directly from the user's computer—no bot joins the call. Setup takes under two minutes: sign in with Google or work email, connect calendar, grant audio permissions, select language. During meetings, the app runs quietly in the background; users can add manual notes via a floating-panel notepad, which teaches the app priorities and appears under 'Your notes' in the post-meeting summary. After capture, AI agents generate deliverables from meeting context: follow-up emails, slide decks, creative briefs, one-pagers, proposals, calendar invites, and Instagram visuals. Users can upload files and connect email/calendar for extra context. Meetings can be grouped into projects so agents reference related history. Simple prompts like 'Draft the follow-up from yesterday's call' work without complex prompting. Templates
RESEARCHED BUILD PROMPT
Give this to your coding agent.
Build 'MeetingMind' — a functional replacement for Supernormal's core bot-free meeting capture and post-meeting AI deliverable workflow. The app is a local desktop application (Electron or Tauri) that records system audio during video calls without joining as a bot, then transcribes and summarizes meetings, and uses an LLM to generate follow-up deliverables from that context.
Workflow: (1) User signs in with Google OAuth and connects their calendar. (2) The app runs in the system tray and auto-starts capture when a calendar meeting begins (or via a 'New capture' button for ad-hoc conversations). (3) During the call, a floating panel shows a notepad for manual notes; these notes merge into the final summary. (4) After the meeting, the app transcribes audio (use Whisper locally or via API), generates a structured summary (action items, decisions, key points), and stores it with the transcript. (5) The user can then issue natural-language requests like 'Draft the follow-up from yesterday's call' — the agent pulls the relevant meeting context and produces a draft email, slide deck (markdown-to-PPTX via python-pptx), or one-pager. (6) Deliverables are exportable to Google Drive or as local files.
Technical requirements: local audio capture via Web Audio API or native OS loopback (e.g., blackhole on macOS, WASAPI loopback on Windows); calendar sync via Google Calendar API; an LLM-backed agent pipeline (OpenAI or local model) with retrieval over the user's meeting history; a simple SQLite store for meetings, projects, and notes; an MCP server exposing meeting context to other tools.
User-facing requirements: two-minute setup, system-tray operation, no bot visible in the call, manual note-taking during capture, grouping meetings into projects, and template-based deliverable generation.
Explicitly do not copy Supernormal's brand, logo, marketing copy, or UI layout. Design your own visual identity and interaction patterns.
Out of scope: cloud sync, team workspaces, mobile apps, native calendar booking, live captions during calls, and multi-language transcription beyond English.
Deliver a README covering setup steps (dependencies, environment variables, OAuth credentials, audio permissions per OS) and important limitations (system audio loopback requires OS-specific drivers, transcription accuracy varies with audio quality, LLM output quality depends on prompt context, and MCP server is experimental).