Hi HN,

I have been actively listening to podcasts for many years. But I found myself typically going back to the same couple of shows every time, and even though I liked these shows the individual episodes had a big variety in how interesting I found them. That made me wonder, is it possible to always leave a podcast feeling inspired?

That is why I decided to build Qling. The idea has been in the back of my mind since 2022, however since I have a full time job, it was difficult to embark on such an ambitious side project.

Until I found GPT-4.

I consider myself a decent engineer, but the speedup I gained working with LLMs enabled me to take on a project like this one.

I built Qling as a platform that aims to find novel, inspiring and also new content tailored for each user. The platform scores podcast episodes by inspiration and merit. I recently also added previews to give bite sized highlight snippets that can be scrolled through in a TikTok/Reels-like fashion.

How it works:

– The ingestion engine ingests ~20k new episodes/week

– Transcripts: Whisper-large-v3-turbo and Meta-Llama-3-8B-Instruct on an A6000 GPU farm

– Scoring: episode embeddings + speaker credibility + insight + inspiration

– Stack: Django + Postgres + Redis on Heroku; SwiftUI client

Try it:

- iOS app: https://apps.apple.com/app/qling-podcast-player/id6740029729

- Simple landing page: https://www.qling.app/

Looking for feedback on:

1. Do highlight snippets help you decide faster?

2. Are the “inspiring” picks accurate or cheesy?

3. Any UX issues?

4. Do you find any new and unexpectedly great content?

I’ll be in the thread all day — ask me anything and I’ll ship fixes asap.

Thanks for checking it out!