What's the nature of the difference?

I can only make inferences based on what you have described of yourself (for example 'post-rationalist' type descriptions) as well, obviously, as updates based on conclusions that have been reached. Given that the subject is personal I should say explicitly that nothing in this comment is intended to be insulting - I speak only as a matter of interest.

I figure we both have some sort of UDT-inspired framework for epistemology,

I think UDT dominates your epistemology more than it does mine. Roughly speaking UDT considerations don't form the framework of my epistemology but instead determine what part of the epistemology to use when decision making. This (probably among other things that I am not aware of) leads me to make less drastic conclusions about fundamental moralities and gods. Yet UDT considerations remain significant when deciding which things to bother even considering as probabilities in such a way that the diff of will/wedrifid's epistemology kernel almost certainly remains far smaller than wedrifid/average_philosopher.

bolstered in certain special cases by intuitions about algorithmic probability, and so any theoretical disagreements we have could presumably be resolved by recourse to such higher level principles.

Yes, most of our thinking is just a bunch of messy human crap that could be ironed out by such recourse.

Is our dispute more on the theoretical or pragmatic side?

A little of both I think? At least when I interpret that at the level of "theories about theorizing" and "pragmatic theorizing". Not much at all (from what I can see) with respect to actually being pragmatic.

But who knows? Modelling other humans internal models is hard enough even when you are modelling cookie cutter 'normal' ones.