LVPMs can have hidden layers, these are trained by the action of betting.

Traders in prediction markets kinda earn their Shapley values right – via “selling the world”. So do agents in markets. So prediction markets really are markets for buying information from traders.

Come up with a model for markets * take the information analog (should get LVPMs) * see if the algorithm is similar to BayesNet updating * figure out how they “create new ideas/paths” * take the information analog (should get blackbox LVPMs, maybe like neural nets) * What is the role of “enumerating all agents””