Ve33 is likely to exit the historical stage, and the new version that Aerodrome will launch in July has several targeted improvements to Ve33. The biggest problem with ve33 pointed out by Aerodrome is lag. Pool incentives are updated once a week, and the voting trend is based on historical performance. For example, if Pool A performed well last week, the majority of votes will still be concentrated in A next week, which is very unfriendly to new pools and the explosive potential of a new pool is basically imperceptible. The new design eliminates the vetoken and changes it to sAERO (which can be transferred at any time), and cancels the weekly voting setting for epochs. Instead, voting can be adjusted at any time. AERO rewards are also continuously distributed to Pool gauges in the form of streaming, while fees are also accumulated in real-time and returned to sAERO. Although it may seem like little change, in reality, the relationship between Pool, LP, and holder has undergone significant changes. The original design, voting, and incentive rewards all had a lag of one week. However, for veAERO, who holds the voting power and wants to obtain more fee sharing, they need to judge the pool that performs well in the future. In the old mechanism, there is basically no way but to choose the pool with good historical performance, because even if you judge that a new pool will explode next week, the cost-effectiveness of betting on it is extremely low. Choosing the wrong one during this one week window period will result in a loss of most of the profits. Basically, no one will make such forward-looking predictions, but keep chasing pools with good data. Once a pool becomes a "historical winner", it will continue to eat emissions, even if users start to doubt its future performance. Deterioration makes it difficult to switch quickly (with a one week hard cycle), which is clearly not friendly for new pools. After changing the epoch, the flexibility is greatly improved, the amplification effect is stronger when the prediction is correct, and the stop loss speed is faster when the prediction is incorrect, which will make the overall capital allocation efficiency higher. Further thinking, this real-time and dynamically adjustable design is naturally more suitable for AI agents to make automated decisions and frequently adjust warehouses. This is also a story that Aerodrome has been working on, such as the DEX model that is suitable for AI integration. The original VEToken model was difficult.