When I first subscribed to GitHub Copilot Pro, I remember having a much more generous premium model allowance. Today it's 300 premium requests/month, and different models consume different amounts, so the effective usage is even lower. I understand the economics of AI inference and why providers need to change pricing over time. My question isn't whether AI is expensive to run. What I'm wondering is: Is it fair to significantly reduce the value of an existing subscription after people have a...