The Great Mental Models volumes I and II by Shane Parrish and Rhiannon Beaubien
These volumes work for narrative explanations, ie. these mental models work well in situations where you need to convey an idea or help explain something. Given Parrish's empire, this makes sense.
However, for capital allocators or others who care about where exactly the rubber meets the road, what matters are the boundaries of these mental models: we want to know both where we can apply a model and where it will blow up.
ie. it would have been nice to demonstrate a tractable start, and only then begin the work. eg.
For every mental model, there is a set of domains over which it is operant and dominant.
Where by "domains", we mean a set of situations parametrized by some variables, and by "operant and dominant", that the model suffices to generate correct predictions.
This immediately invites questions: What is the extent of a mental model's domain? How do these domains overlap and compose in situations where the model is operant but not dominant?
By way of exposition, distance = 1/2 * gravity * time2 is operant and dominant for small objects falling to the ground when the effects of atmosphere are negligible, and the model will not work otherwise (perhaps the object has a working parachute attached to it, or perhaps the object is not small and is a planet).
And as an example from social science, Cialdini posits that the operant methods for influencing another person are: Social Proof, Commitment and consistency, Liking, Authority, Reciprocation, and Scarcity. nb. that none are dominant, however they do (a) act to shift the choices that people make, and (b) work cumulatively so that the more methods applied will create larger shifts.
Finally, perhaps it is the case that domains have an extent and that we can compare the extents of two models to reveal the one with broader applicability, and then publish a list of major models ranked broadest first.
While this is a lot of hand-waving to do in a blog post1 and perhaps this approach turns out to be infeasible, it would have been nice for the authors to have at least discussed the considerations that went into these works.