@nprofile1q... I think @nprofile1q... does this for the Doerfelverse. I think he had to install a plug in to properly serve .xml files or something like that.

@nprofile1q... @nprofile1q... @nprofile1q... @nprofile1q... @nprofile1q...

It's been a while since I did anything, but IIRC, with an lnaddress, my apps look up a keysend first, then fallback to lnurlp.

This was a design choice because I like that keysend makes sending the metadata easier, so it's my preference over sending to lnurlp when given both options.

@nprofile1q... @nprofile1q... @nprofile1q...

None of this is to say I dislike AI. I worry about what it will do to us, our ability to think new thoughts, our ability to trust ourselves. Like McLuhan says "The medium is the message." and I worry how this medium will change our minds. But I use AI every day, it's helped me learn new things, it's helped me quickly iterate ideas, it does grunt work for me, but I've also seen how it 'dumbs' some of my guys down, especially my junior guys, and I too have concerns.

@nprofile1q... @nprofile1q... @nprofile1q...

I drop in a PDF showing panel schedules, and ask it to tell me what electrical loads are higher than 30 amps, and it confidently gives me a list of loads with their ampacity, and it's all made up.

The problem I have with LLMs, is if you know enough to double check it, it's often quicker to just look it up yourself rather than babysit it, and if you don't know enough, then you don't know when it's making stuff up. I'm simply not confident enough in it to trust it... yet.

@nprofile1q... @nprofile1q... @nprofile1q...

The other problem I've found with being wrong 50% of the time, is if you aren't an expert in the topic, you don't know it's wrong, and I've found many people treat the LLM as infallible.

I've had to tell my estimators they cannot use LLMs for asking electrical code calculations. I've asked it to size wires, thinking it's all hard data that's in charts, and it confidently makes stuff up. If I wasn't already familiar with the charts, it wouldn't even raise a red flag.

@nprofile1q...

My thinking is if the podcaster wants his metadata, he can get some skin in the game and host it himself or pay for hosting instead of relying on the apps to run his business for him.

Maybe they should host with Fountain if they want their metadata.

And I think Strike has a Webhook api. I know Alby does.

@nprofile1q... @nprofile1q...

This whole business of having 20 different apps all hosting their own messages is stupid. The podcaster ends up needing a server anyway to go through all the invoices, fetch each message from each app, then display it in a user friendly way. If they're going to need a server to do that anyway, let's cut out the part where each app stores their metadata and the app can just send the metadata straight to their server instead.

@nprofile1q... @nprofile1q...

This isn't the responsibility of the Index or the apps. This is the responsibility of the podcaster. If they want their messages, they should figure out how to receive and host them or hire a host that does.

@nprofile1q... @nprofile1q...

Can I have a private relay that public apps can post to, or would they need some secret way to post to the private relay?

@nprofile1q... @nprofile1q... @nprofile1q... @nprofile1q... @nprofile1q...

The problem on the podcaster side is they lose the TLV record (comments, if the wallet allows are limited to 200 or so characters). I don't know how much of the TLV is needed, but stuff like remoteItem is usual for something like a music show.

There are solutions to this as well. The podcaster could have a TLV hosting service they use that the app sends the record to. This has other problems, but those are solvable too using Webhooks to verify payments.