I've also seen problems with some AI IDEs Gemini integrations recently too, which makes me suspect it is a capacity problem.
Curious if people are seeing this with bigger/more sophisticated workloads.
Wonder if Google are having serious scale problems, which probably would be the first time I've ever written that!
It seems to work okay for something like Cursor, where you clearly realize that you're using a certain model and are willing to take on the risk for higher quality output.