The podcatcher app I've always wanted
I listen to a lot of podcasts. So: improving the quality of an average listening hour is a big deal. I also want to spend less time listening overall, while improving my ongoing situational awareness. So, I built a podcatcher app.
I’ve been using it for a month or so. It’s a big upgrade.
Episode summaries
The big thing: for every episode, the app gets a transcript1, then produces a summary:
- tl;dl (a couple of sentences)
- Key points (5-10 bullets)
- Notable quotes (3-5 verbatim quotes)
I usually read the summary. I can also listen to them, individually or as a playlist of all the queued episodes.
Result: I skip a lot of episodes I would otherwise have listened to in full, and get the key ideas from many more. The average insights per listening hour has gone way up. And I’m spending less time listening overall, which means more “shower time” and more music.

Smaller “quality of life” improvements
- Inbox view has two lists: “favourite podcasts” and “others”.
- The add-to-queue affordances I actually want.
- Prominent button to access the episode web page.
- Button to copy podcast summary, transcript and link for discussion with LLMs.
Ideas for soon
- Improve episode discovery: The search function should include topic search, with a taxonomy of topics that match my interests, or allow me to do ad hoc semantic search.
- “Follow on Twitter but for audio”: get all of someone’s interviews from any podcast (not just those I’m subscribed to).
- Retention support: integrate with Mochi SRS. Try forcing myself to type my takeaways into a text box at the end of each episode (blocking future episode listens until it’s done).
- Distinguish work and leisure: I listen to philosophy and history podcasts for fun, but most of my listening is for “work”. The UI should reflect this distinction.
Want to try it?
If there’s demand, I’ll make this available to others. Email me to join the alpha.
Footnotes
If the podcast publishes transcripts on their website, it uses that. Otherwise, it generates one using AssemblyAI. ↩
