I’m less interested in GPT’s suggestions, which are generic and well-known, and more interested in the author’s own list of podcasts. It looks like a lot of coding content? What do they talk about in podcasts like these? I can’t imagine taking in software information in anything other than written form. I do like the occasional podcast on AI news, so I guess I can relate a bit?

"How do we solve problems?" - the core of software development - is an endless high-level topic that doesn't really require you to dive into specific code.

> more interested in the author’s own list of podcasts.

op here. One of my favorites is Soft Skills Engineering (https://softskills.audio). Check it out if you get the chance!

This just seems like a top-10 list of podcasts. Not particularly great recommendations, and definitely not personal based on OP's OPML.

I've actually been having much better luck finding recommendations with Gemini. I have been able to find some nice niche podcasts that I would not have found otherwise. ChatGPT's recommendations have not been particularly great and it feels harder to steer the suggestions in a given direction.

I did the similar thing with exported list of books with my ratings from Goodreads. ChatGPT-4 gave me some good suggestions.

Not very interesting. These are all very popular, Americanocentric, and "mass appeal" shows, nothing that indicates specificity in recommendations. Which is in line with expectations for GPT-3.5/4.

The problem is ChatGPT doesn’t understand podcast popularity nor the journey heavy listeners tend to go on.

To end up with a list like OP’s, you typically have heard of popular podcasts, tried them and stopped listening, either unsubscribing or not adding them in a new podcast app.

To ChatGPT the OPML just looks like you overlooked some similar material and it happily fills in the gaps. Integrating with a listener stats data source and capping the max popularity of the recommendations to the minimum popularity in the OPML shows would make it way better.

I’m not familiar with any of them, so it might have been useful to me… if this were the sort of podcast I listened to.

But while you’re looking at that. How does it compare to the list he handed the bot?

Interesting. Maybe not Reply All but I would expect most Americans to know about Joe Rogan’s podcast if not what it’s called. He is absolutely massive in the States, and last time I checked his podcast was the biggest in the world.

OP’s topics lean more tech and programming related. Reply All is kind of a tech show but as the description implies it’s more of a society show centered around tech topics. It’s also, as grandparent implies, one of the biggest and more mainstream tech ancillary podcasts. It is quite good by the way, if you’ve never listened to it.

I would have wanted more technical and deep cuts based on my recommendations

Maybe most Americans do, sure. I don’t see how that matters, here in Ireland. :p

As for the recommendations, I find that surprisingly often it works to tell the bot that. Start by asking it what it’s getting from the list, then correct it before it creates the new one. And ask it to justify itself; if nothing else you’ll be able to skip a few.

How many of these suggestions actually exist?

Most of the URLs respond with HTTP status 200, the few that didn't:

http://www.richroll.com/category/podcast/feed/: 302

http://moneyguyshow.libsyn.com/rss: 301

https://www.codingblocks.net/podcast-feed.xml: 403

Seems good enough for a first pass at least.

I want someone to create a plug in LLM that links into my Spotify (or other streaming service of choice) and see how well it can do at recommending me new artists to check based on my liked songs/artists/other playlists. I would be interested to see if it gave me better suggestions based on my taste than the YouTube/Spotify/Pandora radio algorithms.

Finally, have it create a playlist for me that I can import and play without having to manually look up any artists or songs it suggests.