tl;dr: AI scrapers are hammering git.kernel.org by crawling every commit URL across 922 forks of linux.git as HTML instead of just cloning the repos, consuming ~20% of total CPU capacity across 5 nodes purely to render commits for bots. Defenses like IP/ASN bans failed once crawlers shifted to residential proxy networks, and Anubis proof-of-work challenges (now at difficulty 5) are being solved by ~33% of bots. Legitimate traffic is estimated at just 2%, forcing kernel.org to disable features and gate expensive operations.