Let P express a proposition of the form 'I am thinking that Q' (where Q expresses any proposition).
1. If a subject S sincerely asserts P, then S believes that P (by the nature of sincere assertion).
2. If S believes that P, then it is true that P (by self-verification).
Therefore, if S sincerely asserts P, then P is true (i.e. If S sincerely asserts 'I am thinking that Q', then it is true that S is thinking that Q). By this reasoning one could argue that self-ascriptive assertions of the form 'I am thinking that Q' are infallible (modulo concerns about content externalism??). Perhaps this is not a very substantial conclusion but it's something.
Given an additional premise about knowledge, we can conclude that if S sincerely asserts P, then S knows that P.
3. If S believes that P and it is true that P, then S knows that P.
Since this premise does not mention anything about justification or evidence it would need to be supported by some further argument about the nature of this kind of self-knowledge. I don't know how this would go, but if this reasoning could be sustained it would transform an assertion that P into knowledge that P. A Somewhat magical conclusion but still not very substantial as far as self-knowledge goes. It would need to be extended to different P's that express more substantial attitudes (of which I am very skeptical)... its cool but I think this kind of neo-expressivist reasoning is going about self-knowledge all wrong. I guess I would prefer some kind of unmediated observation model.