Truth theory sketch

"We can talk to each other", taken literally, must be either hopelessly ambiguous (not a statement at all) or true and undecidable.

This is because it requires a tacit conception of truth - 'using the language properly', so to speak, and also of what counts as a language.  Both of these must be reasonably reliable if the statement is to be true, but no algorithm written down in the language can demonstrate this reliability.

Also, no algorithm in a 'higher order' language would be translatable, so speculations of this sort are irrelevant (if any sense at all can be attached to them).

The reason the statement is true is not algorithmic:  it's falsehood would be unitelligible, a Moorean paradox.  If someone insisted on its literal falsehood, we would have to wonder what they meant (certainly not something related to the conventional uses of the words it contains).  They would be doing something at the same time as seeming to claim that it was impossible.  Either we'd be wrong about what they were doing, or we and they did not share an interpretation of 'impossible'.  This is not an algorithmic demonstration because no explanation to such an objector would have any weight - we could never be sure that there was enough mutual understanding to articulate an explanation.  We would either find a way of working out what they did mean - would find a way of continuing to talk with them which rendered their objection interpretable - or the conversation would disintegrate.

The tacitly implied predicate 'is true', here, is not grounded in Kripke's sense.  But it is still a kind of basis for grounding.