Oof, that's not a great premise to take as a requirement right out of the gate. More counterexamples than examples for that one.
> that uses cryptography to generate a deniable token that can't be cross-correlated but proves your humanity/age
If it's actually deniable/anonymous then how would it work for rate limiting? If you can't correlate their activity then you don't know if the million requests are a million people or one bot with a million connections. If you can correlate their activity then it's not anonymous.
Moreover, it's a false dichotomy that we should be doing either of these things. The better alternative to corporate surveillance isn't government IDs, it's no surveillance.
