The idea isn't bad, but the execution of the idea will make it or break it. If you code up a simple LLM wrapper, it will suck, because it will just hallucinate.

The idea isn't bad, but the execution of the idea will make it or break it. If you code up a simple LLM wrapper, it will suck, because it will just hallucinate. At the other extreme, if it just spits back URLs that look interesting, you aren't offering anything that can't be solved by an RSS reader. Finding the correct balance of applying some LLM assistance, but not giving it enough rope to turn into AI slop would be the secret sauce. Getting people to pay for it is a whole other question. We've built an online culture where people don't pay for content unless it is a streaming service of some kind. I think there is potential to change that, but you'd be fighting that battle at the same time as trying to nail the correct product. Not impossible, but this is probably a harder path to success that it would seem at first glance.