What's different is the classical permissions. Previously you could use the webRequest permission to execute custom JS functions in response to network activity. In MV3 you must now write a declarative rule instead using the declarativeNetRequest permission. By moving from an imperitive model to a declarative one, Google now has exacting control over what ad blockers can and can't do.
Gorhill's argument is that the stated reasons for MV3 do not align with the implementation. The example he gives is that Google claims the new API is more private. However you still need <all_urls> so the extension is as un-private as its MV2 equivalent. The only difference is that now Google controls what blocking you're allowed to declare and how much of it you're allowed to do.
There's a similar community discovery where MV3 implementation is provably counter to Google's claims of performance enhancement - MV3 extensions need to rehydrate state every 5 minutes as Chrome shuts down their service worker and for highly active extensions such as ad blockers this is actually less performant than the MV2 implementation with a long-lived background.
Basically, Google is full of shit.
