This is an interesting question. I work with students (older teens/young adults) and they are dealing with similar issues. It seems like the world's expectations shifted very quickly under their feet.

I don't think anyone knows the answer. As a hiring manager, I definitely put less weight on generic CRUD apps etc nowadays. You can argue that people can actually just copy and paste from SO before, and that's true, but even with that you had to have some knowledge so integrate what you've copied. With AI assist, the process is orders of magnitude easier, as you can just re-try prompts etc.

What I look for instead is more information on the process of creation, which usually means examining their writing. How did they get the idea? how did they think about what features to build? But even this is not immune to AI contamination.

Overall, I think we're likely to move towards more reliance on verifiable longitudinal data rather than "spot checks". It's much more difficult/challenging to re-create for "portfolio cheaters", and easier for authentic applicants. I get my students to write a dev journal which I verify, and use that as part of a private portfolio that we can share with potential employers.

Overall, I'd say the vetting process is much more onerous on both sides and portfolios will now need proof-of-authenticity.