I guess I failed to bridge the inferential gap here. You're right "Solomonoff Induction would not consider this a description of x as it cannot be used to compute x." The point I'm trying to make is that a correct formalization of induction/Occam's Razor ought (in order to satisfy our intuitions) to assign x some probability > 1/3^^^3 since it has a short (even if uncomputable) description, but a formalization based on this P would not, therefore it can't be a correct formalization. But this is the case no matter what P we use (with different x depending on P), hence the "unformalizability" problem.