ABOUT
What it is
A podcast tracker and learning journal: find real episodes via iTunes search (or paste a YouTube link), listen, capture timestamped notes and highlights, get an AI summary grounded in what was actually said, and export everything to Obsidian or Notion.
Why it exists
Podcast notes usually live nowhere, or get typed from memory hours later. PodMark timestamps notes as you listen and grounds its AI summary in a real transcript of the episode — not a guess based on the title — so the summary reflects what was actually said, not what a show with that title usually covers.
How it works
React/Vite frontend with Zustand state, deployed as a single Vercel project (static frontend + /api/* serverless functions, no CORS). Audio is transcribed via Groq Whisper (segment-level timestamps, per-error-code handling for rate limits, quota, and oversized files); the summary step runs through SumoPod's OpenAI-compatible gateway, grounding on that transcript over show notes when available, and falling back to a hedged general-knowledge guess from the title when neither exists rather than refusing to answer. Accounts are optional — signed-out users get local-only storage — and the AI summary endpoint requires a verified Supabase session specifically to protect API credits from public abuse.
PREVIEW