Recently, while watching the oracle on Solana, I truly realized how different "fast" really is. I used to think that 1.8 seconds and 3 milliseconds sounded fast, until I saw the real data: Pyth takes an average of 9 slots to block, and without priority fees, about 30% of updates are lost directly; The standard feed rate of Switchboard is about 250 ms on average, which is already much faster than the on chain path of the other party - while Surge is a separate product with an average delay of around 3 ms, and even in the worst case, it is faster than Pyth's average. More importantly, Switchboard has completely unlocked Surge - now anyone who wants to use a 3ms level feed can simply subscribe to SWTCH on the chain without waiting for approval or using API keys. Plug is free Pro 450 ms、Surge 3 ms， All on chain billing is transparent. This is what infrastructure should look like in my understanding: When you need the ultimate performance, it's there, it won't hold you back. In addition, they have put all of Kalshi's markets on the chain, with over 10000 real-time data sources readily available; It is also fully compatible with x402, and the on chain AI no longer needs to ask anyone to pull any data. The old oracle is still charging a brand premium, while the new oracle has already removed the threshold to the floor. Technological iteration is always faster than narrative, the market just hasn't caught up yet.