The fact of the matter is that the higher the staked amount the higher the rewards and the more likely inequality to grow, but why are the rewards higher in the first place? Because it is normal to reward at a higher level the increased liquidity offered, it is, after all, a service for the network. But, if we consider that in other conditions of greater equality, that liquidity would not be missing, but would be decentralized, it would come from someone else, then we could think that there is no point in rewarding accumulation, because it will not bring us superior results. Hence, we could have a reward model that accounts just for that. Instead of simply more is better, a model that accounts for the fact that more equals decentralized and therefore is not better. So rewards could be in fact solved by machine learning with the purpose of dynamically adjusting growth in order to satisfy the conditions: growth but not linear or exponential and at the same time increasing decentralization by decelerating growth if analyzed metrics are not satisfactory. This means that in the long run, it will dynamically move in an interval and will reach an equilibrium. The higher the growth in rewards the more decentralized, then rewards drop inverse proportionally with the size of capital, causing a lack of incentive for the rich and more incentive for the medium. Now the question would be what would be the reasons for which a system that slows dramatically rewards (for the actors that prevent decentralization) would not be a functional system? I hope what I said made sense, if not I apologize.