I think this is doomed by the fact tweaks will break in subtle and unpredictable ways every time a website updates, and since you can't reliably detect that an existing tweak is now broken, you can't regenerate it automatically. Once the user notices something is broken, they'll regenerate, and since an LLM is writing the code, the result will be slightly different (layout, colors, positions, ...) than the previous iteration the user got accustomed to. This sounds maddening after a while, and frustration scales linearly with the amount of tweaks.

Sadly (?) the only reliable way to website unfuckery is and will remain crowdsourcing by a bunch of nerds (see: Easylist) for the foreseeable future. This product is the opposite of that, with everyone having their private collection of prompts/tweaks, which they will have to individually fix every two weeks.