A great question from @TheStalwart. The answer is that of course we have lots of open questions in economics, but none of them will get answered by a theorem. The question is not whether a statement is "true" but whether a model is useful to understand a given reality, and a model that may be useful to understand a reality may no longer be useful when the reality changes. That does not mean that usefulness is determined by fashion. Take AI. A pure task-exposure model says radiology can be done by machines, hence radiology is replaced. In fact demand went up as shown by the fact that both wages and quantities of radiologists increased in the meantime(notice: useful model here, both P and Q go up, the mark of a demand shift). If not the tasks, what? With Jin Li and Yanhui Wu we argue that markets price jobs, not tasks, and that the impact of AI depends on whether the job is a "strong bundle" (tasks are hard to separate) in which case AI helps the workers, or "weak bundles" in which case it splits the job, and makes it narrower. If both types of jobs react the same to AI, our model is useless and must be discarded. Another example. Last year in Silicon Continent I argued for the "r without g" impact of AI potentially causing a fiscal crisis in europe (Rstar goes up everywhere due to higher expected earnings and investment boom, g only where AI is adopted, i.e. not in Europe) on the basis of the Auclert et al calibration of AS and AD for assets in their cool Jackson Hole (2025) paper. You can quantify the argument, figure out the elasticities, and the implied increase in rstar due to the AI revolution and test whether that accounts for the fiscal window narrowing. Again, is this model "true"? Or just useful? Neither theory nor empirics settle the question by themselves (as Leamer argued). Economists argue about it, and the data offers the empirical discipline of the argument. No model comes out as "the truth". The models that survive are the ones that are able to keep explaining what happens.