Ladies and gentlemen, have you ever encountered these annoying issues in DeFi? Lending platforms get liquidated inexplicably just because price data updates are a few seconds late; slippage on on-chain transactions becomes outrageous due to delayed price feeds. Behind these seemingly random risks, they all point to the same problem—the data push model of traditional oracles can no longer keep up with the speed requirements of DeFi.

However, recently, an interesting change has been happening: the entire industry is shifting from "continuous pushing" to "on-demand pulling." What does this shift mean? Why does it change the game? Let’s have a good discussion.

First, let’s talk about how the push model works. What is the current approach of most oracles? They regularly send data to the chain at fixed intervals, 24/7. It sounds good, and the data is always available. But the problem follows: first, whether you use it or not, you still have to pay gas fees—who ends up paying? Still the protocol and users; second, there’s a ceiling on push frequency, which can’t meet the needs of high-frequency applications that require millisecond-level real-time data.

Now, look at the pull model. The concept is simple: it’s the opposite—when an application needs data—such as when a user places an order or during liquidation checks—it actively requests the latest quote. This "fetch-on-demand" logic has obvious benefits. First, you don’t pay for idle time, significantly reducing on-chain costs; second, each request can be customized as needed, perfectly matching the specific requirements of the current scenario.

This is practically revolutionary for certain types of applications. Imagine a decentralized perpetual contract platform where every trade and position adjustment requires precise real-time prices. Using the pull model, it can obtain the most accurate data without burning money—something that was unimaginable in the push era.