It is costly to move ERC20 tokens around, because each transaction has a fee that is not proportional to the amount of the transaction, called gas. So you could move $10 worth of ETH for the same cost as moving $1M worth of ETH. Because of this, the idea of collecting a large group of tokens together and moving them in a block to capture the best returns, then sharing the returns among the token holders is very advantageous. You only have to pay one fee to move the whole block from one place to another, and instead of representing a significant cost for small investors, the single gas fee becomes insignificant in proportion to the gains achieved.
AI for Yield Aggregation
While it is possible to algorithmically move the asset pool around every time there is a better opportunity to get rewards, there are too many factors to predict successfully. An AI algorithm can improve its performance and determine when and where to move the assets to capture the best return.
Since changing the investment denomination is always costly – swapping tokens on Uniswap will cost at least 0.3 percent per transaction – the goal is to minimize the number of swaps used while maximizing return. Given the volatility of returns on most liquidity pools, blindly chasing the maximum APR will result in too frequent transactions that erode profits. Having an AI to solve this nonlinear minimax optimization problem will result in better and better return over time, as the learning gathers more data.