From the collaboration between Irys and Warden, it seems that Irys intends to become a benchmark for Chainlink oracle machines in the AI field, ultimately solving the problem of data unification. Chainlink aims to unify price data across different platforms, while Irys aims to unify interaction data across various AI models. From the ancient Filcoin, to Arweave, to CESS, although there are many data link projects, they only achieve the effect of a database. Adding decentralization actually increases the cost, resulting in most data links eventually becoming a more expensive database. The demand has not been accurately identified, and although AI+Web3 has been criticized, the pain point of Irys' AI model without memory is still quite strong. At present, these AI models are mostly one-time, which means that for example, when I chat with Deepseek and Grok, each round of conversation is a fresh start, and they cannot output content based on our previous interactions. Therefore, current AI models have almost no memory, which leads to a waste of computing resources on the one hand, as each round of conversation requires all the content to be regenerated. On the other hand, the user experience is also poor, especially for some AI companion products nowadays, if there is no long-term memory, the effect is very poor. Every time the AI model runs, it needs to output content based on the context of the interaction. If a huge amount of interaction data is stored, it would be too much, Irys, as a project that handles over 98% of Arweave's data submissions, has extensive experience in data storage. In addition, Irys is characterized by its dynamic programmability of data, rather than just serving as a static database. Therefore, it fits well with the role of AI model memory. For example, in this collaboration with Warden, interactive data that triggers model output content can be stored in Irys and marked as a condition for future triggering. When the model dialogue outputs content in the future, the previously stored data can be retrieved for relevance. Once triggered, it will be introduced into this conversation and can also be used by other models, just like telling your mom not to eat. Green pepper, then one day went to a restaurant and entered the dining scene again, so Irys Automatically retrieve and recognize the key memory of "not eating green peppers" and trigger a notification to the chef to remind them not to put green peppers. After this process enables the AI model to have memory ability, the efficiency and experience of using AI models and tools will be greatly improved, and Irys will continue to be closely monitored 🤔