Law without law: from observer states to physics via algorithmic information theory

This is a temporary placeholder with a couple of links:
* A talk at Perimeter Institute, April 2018 (notation has changed somewhat since then):
* Full version of my paper, will be revised soon:
What is going to be revised? Version 3 will be shorter and simpler -- one can (and probably should) do without observer graphs and without covariance. I was wrong that the latter notion is important (thinking more carefully about what probabilistic predictions really mean operationally). Now reading up on imprecise probability, large deviations, asymptotic equivalence and the like... more towards the end of 2018.
* Slides of a talk at Michigan Technological University, September 2018: ann_arbor.pdf

Less precise / to be revised / not really rigorous:
* The short version of my paper, don't read now -- will be updated soon and is subject to substantial revision:
* An essay that I've written for an FQXi contest (4th prize), more motivational than mathematical, perhaps fun but definitely not precise; updated version: essay.pdf