Why bother with Mastodon interop? Why bother building on top of a standard that doesn’t do what you want if you don’t think being part of the “Fediverse” is particularly interesting or a goal of the platform you’re building?
Something new is a better bet at this point than being anchored to or seeming like part of Mastodon IMO.
Because they pretend to want to be open, and it's sending a very clear signal that is not their goal if they're not even trying to work with the existing ecosystem.
If they just want to be a silo, that's fine. But in that case be honest about it.
No matter how open I wanted to be, if I was setting out to build a social network, I’d please precisely zero value on being connected to that “existing network”.
It is irrelevant to me, despite what a vocal minority might want me to believe.
ActivityPub is very generic; it can accommodate all kinds of changes. E.g. if you want to introduce activity / object types that have improvements over what Mastodon supports, you can do so. If you want to introduce vocabulary within existing object types that Mastodon wouldn't understand, you can do so without breaking federation with Mastodon or other Fediverse servers. I'd be a lot more sympathetic to anyone who decided to extend ActivityPub.
The minimum viable subset of ActivityPub is largely: Provide an endpoint that returns an Actor with a list of the required endpoints (inbox, outbox, follower, following etc.) - the spec requires a list of them, but you don't even need all of them for basic interop - and handle POST's to your announced inbox, and GET requests to the outbox. Add unique URL's as "id" fields in the JSON, and support GET to them. Follow the format of the JSON to provide at least the minimum set of fields to address the activity, provide a type, and provide the minimum fields for the given type.
For interop w/Mastodon you'd want to support Webfinger to find the Actor. Nothing stops you from also supporting other mechanisms, like BlueSky's domain validation.
Nothing stops you from supporting additional federation mechanisms. Nothing stops you from providing additional fields. Nothing stops you from storing data locally. Nothing stops you from adding additional ways of signing claims about individual objects or a whole repository of objects. Nothing stops you from providing additional mechanisms for distributed lookup of objects by id. Many of those things would be welcome if people wanted to do it on top of ActivityPub.