I'm interested to know how this methodology is supposed to diverge from existing methods of science in any useful way. We've had a hypothesis-driven science since (at least) Popper and this particular approach seems to offer no practical alternative; at best we're substituting one preferred type of non-empirical criteria for selection (in this case, a presumed ability to calculate he metaprobabilities that we'd have to assume for establishing likelihoods of given hypotheses, a matter which is itself controversial) for the currently-existing criteria, along the lines of elegance, simplicity, and related aesthetic criteria which currently encompass inference to the best explanation.

With regards to the above metaprobabilities, how are we to justify these assumptions with regards to epistemic uncertainty? Popper and Hume have already shown us that, for all intents and purposes, we can't justify induction and Solomonoff is no realizable way around that. I don't see how substituting one mode of inference, based on a set of assumptions that are given no justification whatsoever, provides a superior or even a reasonably-supportable alternative to current methods of science.

This is undoubtedly a useful heuristic for establishing confirmation of what has been observed, but then again so is the research structure we already possess; to claim that such an unsupported and unjustified methodology as outlined in this article is in any way establishing "truth" above and beyond current methods of confirming hypotheses is to overstate the case to such a degree that it cannot be taken seriously.