Mira is a project that I rarely talk about in Kaito pre TGE. In fact, it was a project that I had been following and introducing before they joined Kaito. The reason is that they seem to be really solving a real problem that troubles us - AI illusion, and this problem has a strong demand in specific situations. But to be honest, I have been pondering how they solve the AI illusion. If they really only solve it by combining several popular LLMs together, then putting a shell on top and issuing a coin, isn't TMD a self deceiving scam? I have also downloaded and read the basic paper they cited, 'Probabilistic Consensus through Ensemble Validation: A Framework for LLM Reliability'. The principle is actually to use multiple LLM cross validation and consensus mechanisms to improve accuracy and reduce the 'illusion risk'. I did not see any hard requirements for tokens. Plus, there was a period of time when I couldn't log in to Klok, so I didn't frequently post his tweets for the so-called Mindshare. But today their informative long article is really informative, accompanied by charts. After reading it for half an hour, I finally understood the principle behind their solution Verify the correctness of the output through distributed consensus of multiple models to ensure the reliability of AI, using a PoW+PoS hybrid mechanism as an economic security model In human terms, the Mira network is composed of multiple AI verification nodes, which need to stake tokens. If a node deviates from consensus (fabricated), the staking will be confiscated (Slashing). That's enough to understand. Isn't the foundation of blockchain now distributed verification nodes? The pattern has been integrated into AI result validation. In fact, there are still many things in the article, such as privacy protection, but I know that people don't have much patience to read now, so I won't push too much. Simply put, this article is quite helpful, at least to say that I now have a technical threshold for @ MiraN_Network, and my confidence in legitimate projects with organic token composition has greatly increased. Of course, one thing to say is that @ klok.app is not very user-friendly, and there are often situations where I ask questions but don't respond, which still takes up my question slot. It is still a bit far from my ideal tool without AI illusion. Let me show you the attached picture. Sometimes I feel like I have to urge a slow reacting academic genius several times to get results 🤣🤣 I hope the @ karansirdesai team can quickly iterate and make Klok silky smooth as soon as possible, at least partially replacing Chatgpt's position in daily work.