There was a pretty funny story before about an AI programming project. The demo was super impressive, and they raised quite a bit of money. Later, it was exposed that the backend code was actually written in an old-school way by an Indian outsourcing team... But honestly, this kind of thing is pretty representative. For a lot of so-called AI projects, you have no idea if AI is actually doing the work. Especially in crypto, everyone’s talking about AI automating strategies, but you can’t verify any of this on-chain. At the end of the day, it’s still just a ‘trust me’ situation. Recently, I happened to see a project trying to solve this issue. It’s Flap working on an AI Oracle. What they’re doing is embedding AI prompts into smart contracts. When the contract is triggered, it calls an LLM (Large Language Model), and the result is sent back to execute. This way, the actual AI logic being executed is on-chain, and through a commit-reveal mechanism (which is essentially a decentralized verification method using blockchain hash functions—won’t go into details here), you can verify that the AI logic was genuinely executed. Compared to previous solutions, the advantages are: 1. By integrating AI prompts at the smart contract level, it’s more decentralized and blockchain-native, making it suitable for projects with high decentralization requirements. 2. The inference process is transparent, and the results can be verified on-chain. All in all, I feel like this kind of thing could be pretty useful in certain scenarios, like prediction markets or mechanisms that require fairness in design. At the very least, you can prove that decisions weren’t just made on a whim. From this perspective, I think AI Oracle might become a pretty essential tool and infrastructure in the future.