The Bluesky post that anchored this shift didn't argue for open-source models on capability grounds [12]. It argued on dependency grounds. That distinction matters because it marks a change in how people understand their relationship to AI infrastructure — not as consumers choosing between products of varying quality, but as participants deciding whose economic model they want to support. When the argument stops being "this model is good enough" and starts being "I refuse to fund the structure that makes the other model available," the conversation has left technical evaluation…