

The idea has some merit, however- one of the first rules of authentication is to not do it yourself, and rely on trusted libraries (or platforms, in the case of Auth0).
Why would any company choose a no-name AI-genned rando software as their auth stack that doesn’t offer any guarantees, or compensation in case of a breach, when you could instead… not do that and have a proper functioning tool with adequate support?

That depends on why you do it I’d say. Is it obvious for a human (or an LLM I guess) whether two aliases belong to the same person? Sure. But you can still separate which service an email you received belongs to, ie who leaked your data, or got breached, who sends you spam, etc.