Hacker News Reader

Sounds like it would be susceptible to how they’d reverse engineer LLMs using distillation.

Connect the dots for me, I'm not seeing it. Unless you are talking about building a function approximator for a bunch of measured transfer characteristics, in which case you could as easily do that well before AI.

The several seconds of latency would be fun.

since human hearing is lossy you can get whatever sound you like. In the context of a moat there is none for a sfx

That's basically how convolution reverb has worked for 25 years.