Motivated by the problem of finding finite versions of classical incompleteness theorems, we present some conjectures that go beyond NP = coNP. These conjectures formally connect computational complexity with the difficulty of proving some sentences, which means that high computational complexity of a problem associated with a sentence implies that the sentence is not provable in a weak theory, or requires a long proof. Another reason for putting forward these conjectures is that some results in proof complexity seem to be special cases of such general statements and we want to formalize and fully understand these statements. In this paper we review some conjectures that we have presented earlier [26,32,33,34], introduce new conjectures, systematize them and prove new connections between them and some other statements studied before.