Yeah, I have no idea either, but it’s been around for more than a decade so it should be fairly easy to find a library that duplicates it.
I would be wary of AI-based solutions. There’s a risk of it picking up e.g. satirical/spoof sponsorships as actual ads, and perhaps not detecting unusual ads.
I’m slightly terrified of the day someone starts getting AI to reword and read out individual ads for each stream.
With software that misuses /tmp, I’m more worried about burning out my SSD endurance than running out of RAM.