If they had the slightest bit of survival instinct they’d share a archive.org / Google-ish scraper and web cache infrastructure, and pull from those caches, and everything would just be scraped once, repeated only occasionally.

Instead they’re building maximally dumb (as in literally counterproductive and self harming) scrapers who don’t know what they’re interacting with.

At what point will people start to track down and sabotage AI datacenters IRL?