The following passage that I am quoting at length from Dag Prawitz ("Intuitionistic Logic: A Philosophical Challenge" in Logic and Philosophy edited by G. H. von Wright, Hague, Martinus Nijhoff Publishers, pp. 8-9) may be more illuminating:
Intuitionistic philosophers sometimes use true as synonymous with the
truth as known, but this is clearly a strange and unfortunate use. We
need a notion of truth where, without falling into absurdities, we may
say, e.g., that there are many truths that are not known today. But
do we need a notion of truth that allows truths which are even in
principle impossible to know?
The non-realistic concept of truth when at all reasonable agrees with
the platonistic or realistic concept of truth in the case of sentences
that are in principle decidable. Furthermore, the two concepts agree
(in contrast to the intuitionistic one mentioned above) in allowing
the existence of truths which in fact will never be known. What the
above non-realistic principle or truth rules out is the existence of
truths that are not even in principle possible to know.
The difference between the two principles boils down to this: on the
platonistic principle, a truth condition for a sentence obtains or
does not obtain independently of our means of recognizing that it
obtains or fails to obtain, and we are then forced to admit that there
may be truths that are in principle impossible to recognize (if we are
not to assert unwarrantably that all problems are in principle
solvable); on the non-realistic principle above, a truth is in
principle always possible to recognize, but we must then refrain from
asserting that a truth condition either obtains or does not obtain
(again, in order not to assert that everything is solvable). Both
principles respect the fact that we are not omniscient, but the
platonistic principle does this by introducing ideas the need of which
are not easily seen.
Martin-Löf and Prawitz seem to claim that, just as when the rules of a game is fixed and definite enough to state that a right move in the game is to be found out by necessity and recognisable as such, a truth in mathematics, according to their conception, (and presumably, other truths under similar conditions) takes on actual existence once a proof is devised, furthermore it has been already there as a potential truth. Hence, it could be inferred that we could describe not the criteria to identify potential truths, but the conditions under which truths could be potentially "awaiting".
Martin-Löf finds a metaphysical ground for their version of verificationism in Aristotle's views and especially, Thomas' reinterpretation of Aristotle's. I do not think that, however, the distinction is as closely related to these views as Martin-Löf puts emphasis on. Sure, the words 'potential' and 'actual' have core meanings that manifest themselves in any context that their employment appears meaningful. Nevertheless, relying almost only on this connection makes a weak argument. The most significant point that Martin-Löf and Thomas Aquinas could be said to share is that the epistemological status of the actual (which is present contra absent in a certain sense) is higher than the potential, and prior to it with regard to dependency relation. An explanation of this point is quite another subject and must be dwelled upon elsewhere. For the present question, I can say that it would have been less misdirecting if Martin-Löf and Prawitz had used 'latent' and 'effective' in place of 'potential' and 'actual'.