One theme I'm starting to converge on is the idea of a generalist as an "expert" at a) maintaining a wide range of mental models and skillsets and b) developing a sense of which type of problems to apply each to.
In other words, effective generalists become good at knowing which speciality or approach should be applied to a problem, even if they "only" grasp the basics of any one discipline.
Example:
A software engineer wants to develop deeper friendships. They may think that building an app that reminds them to keep in touch with friends will help.
Of course they think that... software is what they know best.
But a generalist may take a different angle and see that the root cause isn't an automation / information problem, it may be a human psychology issue.
"The real problem is that you don't believe you're worthy of love. If you work on that, you may feel confident enough to want to reach out more."
The next email is going to start outlining the most practical, effective mini-mental models that generalists can use to solve practical problems.