I run an open source webradios webapp side project[0] that basically do this mass polling for song titles, using Elixir genservers. It all runs on an small oversized EX44 Hetzner server amongst other apps.

I thought of extending it to offer what Tuneping apparently does but couldn't imagine who would need that. I guess Tuneping (which I guess is OP's site) proves there's a market? ;)

[0] https://www.radio-addict.com

If you can't be bothered to take the time to clean up the incoherent Claudespeak, I'm not going to bother trying to read it.

> Two honest limits: the classification is a reading of the announced text, not of the audio.

Why do people do this? The author cared enough to make this, put it online, and submit it to HN. Can you not take five minutes and write some perfectly straightforward body text?

The mind boggles.

Mind the boggles.

I saw "because every idea below rests on it. " and immediately my "AI Slop" alarm went off.

But I'm curious now: why _do_ LLMs talk so much about "load bearing", "rests on it", etc.? I don't remember reading such phrasing growing up, so how did it get into their training set?