“None of this was even scraping my blog, it was all targeting at generating tarballs of every single commit of every single publicly listed git repo in my gitea instance.”
AI scrapers filled a one-person Gitea server with 20 to 30 GB of tarballs a day. Traffic went to more than double the normal rate and locked the owner out of his own git server. Amazon, OpenAI, Apple and Facebook announced themselves, and many others faked their user agents. The cost of their training data lands on hobbyists’ disks.