I'm not surprised. I started using Windsurf when it came out because I liked its UX better than Cursor's.
However, while Cursor and GH Copilot improved, Windsurf went in the opposite direction. On each update, I started to get more and more issues. The agent often tried to run shell commands, and it hung up, or I found minor UI bugs. One day, I decided to give GH Copilot another chance, and I was surprised by how it evolved, to the point that it worked better than Windsurf for my usage. I don’t know what happened internally at Windsurf, but I notice the degradation as a user. If my case indicates what happened to other users, maybe OpenAI saw declining subscriptions and canceled the deal.