yep. If one could solve the discoverability problem for podcasts they'd make a mint.
The problems, as I see them:
- must be resistant to gaming. Many of the really popular podcasts are backed by marketing money, many of the best ones are buried
- must be able to account heavily for taste. I really love midst, magic tavern, darknet, zzyx (RIP), night vale. Lots of people aren't into fiction so the system would have to have market segmentation built in. like a music app recommendation engine.
- should provide mechanisms to link to other stuff for the podcast. A huge part of the problem for the podcasters is the friction there is to give them money. This is mostly solved by podcast metadata but any good tool needs to display these properly.
as it is, current podcast recommendation algorithms miss tons of good content. There should be more than enough data on me to recommend a lot of stuff that I've had to hunt down Internet 1.0 style by googling and reading blogs.
I wonder if an embedded cbatgpt like system could offer such a recommendation engine on demand for different sorts of media. It wouldn’t ever be the best recommendations of all time, but it might be able to handle scale better than humans. Eg, to get something YouTube algorithm like for other services like podcast apps
The problems, as I see them:
- must be resistant to gaming. Many of the really popular podcasts are backed by marketing money, many of the best ones are buried
- must be able to account heavily for taste. I really love midst, magic tavern, darknet, zzyx (RIP), night vale. Lots of people aren't into fiction so the system would have to have market segmentation built in. like a music app recommendation engine.
- should provide mechanisms to link to other stuff for the podcast. A huge part of the problem for the podcasters is the friction there is to give them money. This is mostly solved by podcast metadata but any good tool needs to display these properly.
as it is, current podcast recommendation algorithms miss tons of good content. There should be more than enough data on me to recommend a lot of stuff that I've had to hunt down Internet 1.0 style by googling and reading blogs.