"What is it like to be a rock" => no thing satisfies that answer => a rock does not have unconscious mental states

How do you know that?

Philosophically, of course.

I mean sure you can’t cut a rock open and see any mental states. But you can no more cut a human open and see mental states either.

Now I am no way suggesting that you don’t have a model for ascribing mental states to humans. Or dogs. Or LLM’s. Just that all models, however useful are still models. Not having a model capable of ascribing mental states to rocks does not preclude rocks having mental states.

> How do you know that?

Well you don't, and my reading of the article was that Nagle also recognized that it was basically an assumption which he granted to bats specifically so as to have a concrete example (one which was suitably unobjectionable, seems like he thought bats 'obviously' had some level of consciousness). The actual utility of this definition is not, as far as my understanding goes, to demarcate between what is and what is not conscious. It seems more like he's using it to establish a sort of "proof-by-contradiction" against the proposal that consciousness admits a totally materialistic description. Something like:

(1) If you say that A is conscious, then you also must say that A has subjective self-experience (which is my understanding of the point of the whole "what it is to be like" thing)

(2) Any complete description/account of the consciousness of A must contain a description of the subjective self-experience of A because of (1)

(3) Subjective self-experience cannot be explained in purely materialistic/universal terms, because it's subjective (so basically by definition)

=> Consciousness cannot be fully described in a materialistic framework, because of the contradiction between (2) and (3)

> Just that all models, however useful are still models

Totally agree with this, I think you're just misunderstanding the specific utility of this model (which is this specific argument about what can be described using human language). My example with the rock was kind of a specific response to OP illustrate how I understood the whole "what it is to be like" thing to be equivalent to (1). If I'd had a bit more forethought I probably would have made those arrows in the line you've quoted bidirectional.