I agree that's a major problem. It's not something I've solved yet. I suspect a web research sub-agent is likely what's needed, so it can pull in up-to-date docs for whatever library you need to work with.