Vitalik proposed a novel idea of using options as a basis to construct synthetic assets instead of the traditional CDP+Oracle+liquidation model. It is elegant to use it on stablecoins, splitting ETH into a stablecoin and a leveraged token. The mathematical implementation of P+N=1 reminds me of the xy=k moment when AMM was proposed. The research is as follows: Let me explain first that synthetic assets here refer to, for example, building a stablecoin on the chain. In a decentralized way, you need an Oracle price and an ETH CDP collateral clearing system, which is the earliest form of MakerDAO. In terms of purity, this system still needs to rely on an external trust, which is Oracle. At the same time, a clearing system needs to be designed with sufficient risk control measures and buffering mechanisms, otherwise situations similar to 312 may occur. Vitalik's solution looks more elegant, replacing the debt system with options as the basis. Firstly, there is no risk of liquidation, specifically by dividing 1 ETH into a pair of complementary synthetic options: P: Put option (protecting against downturns) N: Call option (capturing the rise) At any time, 1 ETH can be split into (P, N) or merged back into 1 ETH. The mathematical implementation is P+N=1. (This relatively pure implementation is similar to AMM's xy=k) So options have expiration dates and exercise prices, so Oracle only needs to quote once at expiration, without the need for real-time quotes, greatly reducing Oracle's risk. If you are not familiar with options, you can simply think of splitting ETH into two parts, where P protects your insurance if it falls, and N earns more lottery tickets if it rises. No matter how the underlying ETH fluctuates, the sum of the two will always be equal to 1, never more or less. For example, For example, if the current ETH is $2500 and the exercise price is chosen as s=$1500. The relationship between P holder and N holder can be seen from the curve in the diagram. At this point, you hold 1 ETH and split it into P and N. What would happen if the price reaches x=$1800 at maturity? P=min (1, s/x) takes the minimum value of 1 ETH and 1500/1800=0.833 ETH, which is 0.833 ETH. At that time, the value of 0.833 ETH in US dollars was 1800 × 0.833 ≈ exactly 1500 US dollars. N=max (0,1-s/x) takes the maximum value of 0 ETH and 1-0.833=0.167 ETH, which is 0.167 ETH, worth $300 in US dollars. The sum of the two is exactly $1800, so P+N=1 is always balanced. So for stablecoin users, after splitting 1 ETH into P and N, they can directly sell N for P. But in fact, there is a key issue here, which is that once it falls below the exercise price s, although P can still obtain the full 1 ETH, it will lose money on the US dollar basis. Therefore, users must regularly adjust s and exchange old P for new P to make the exercise price s as safe as possible. The wear and slip points of this process are not very controllable, and users may be affected by MEV if they operate it themselves. Finally, V expressed his views on the current situation of DeFi. He believes that DeFi should not be just a replica of traditional TradFi (such as the CDP model, which actually moves the logic of traditional finance onto the chain), but should be completely different in architecture. Even if sacrificing a little convenience for higher security and decentralization is more worthwhile. It may be a bit difficult to understand, but I tried my best to integrate my own understanding to explain it. I think it is a good idea, and the advantage is that it eliminates the dependence on real-time Oracle. Even if there are problems, there is still a lot of time to intervene, and it can also give ETH more application value. Of course, the specific feasibility still needs to be verified through running data and models.