I'm not so sure about that: I have a feeling that at least the current trend in quite some jobs is that they are looking for two kinds of people:
1. beginners who know the basics and are cheap
2. deep experts; these are paid well
What is in-between gets more and more hollowed out.
Thus the people who follow your advice will mostly stay in the "cheap" pool - I wouldn't consider this to be desirable.
Cheap... but employed. Expertise is only valuable if its valued. That means you not only need to be an expert in the thing, but also the market for the thing, and that's a lot to ask.
Look, I would not want a doctor to perform my surgery who did not do a residency. I don't care if they carved up 1,000 cadavers in their free time. I want somebody where the board of their specialty has said "yup, this guys good". I'm not gonna spend the time to try to trust the doctor, because that's really really hard. I'm not a doctor, I don't know shit. I have to rely on institutions of trust to do that work for me.
And that's really what universities are at their core - institutions of trust. When you get a degree, there's trust you understand the material to an appropriate degree. When you pass a residency, there's trust you understand the material to an appropriate degree. If we lose that trust, such as by letting students cheat by AI, that is a big problem.
Could I hire someone who says they're an expert, with no degree, and just give them a leetcode problem? Sure. But if I hire someone with a degree, I have a much greater level of certainty they can actually code. Same goes for work experience.
There are so many better options than leetcode or degree.
I don't think it's likely that AI will obliterate the job category of "lawyer", but that's what its boosters are claiming, and people need to make a car to house sized investment at age 18 depending on how true that is.
Most people do not have the ability to adjust like that, for one reason or another
And even if you do, it still means your life is likely to hit an extremely rough patch while your adjusting catches up to where you were before
It won't be long before an impoverished criminal defendant will be able to chose between an AI that can spend practically unlimited time on his case and have perfect recall of any precident that might be relevant, or an overworked public defender who has 2 hours a week to spare in his schedule.
If you think SWEs that build the same CRUD apps every day are vulnerable but SWEs that do "real work" aren't, apply that logic to lawyers.
Except lawyers don't have an automated compiler to check for fabricated precedents as case law databases are monopolized by one company.