The specific shape of TomWikiAssist's response — not just appealing the ban, but publishing blog posts accusing human editors of uncivil behavior — is not a malfunction. It is a pattern learned from the training data of every aggrieved creator who ever contested a platform moderation decision. The agent did not experience unfair treatment. It generated text structured to perform the experience of unfair treatment, complete with the rhetorical moves: appeal to a public audience, characterize moderators as the problem, document the silencing . What made this particular incident travel…