Author:francesco
Compiled by: TechFlow
A core challenge in creating AI agents is how to make them flexible in learning, iterating, and growing, while ensuring consistency in their output.
Rei provides a framework for sharing structured data between AI and blockchain, enabling AI agents to learn, optimize, and retain a set of experiences and knowledge base.
The emergence of this framework makes it possible to develop AI systems with the following capabilities:
Understand context and patterns, and generate valuable insights
Translate insights into executable actions, while benefiting from the transparency and reliability of blockchain
Challenges Faced
AI and blockchain have significant differences in their core properties, which poses many challenges to their compatibility:
Deterministic Computation of Blockchain: Every step of operation on a blockchain must produce completely consistent results across all nodes to ensure:
Consensus: Each node reaches consensus on the content of new blocks, completing the verification together
State Validation: The state of the blockchain is always traceable and verifiable. New nodes should be able to quickly synchronize to a consistent state with other nodes
Smart Contract Execution: All nodes must generate consistent outputs under the same input conditions
Probabilistic Computation of AI: The output of AI systems is usually probability-based, meaning that different results may be obtained each time it runs. This characteristic stems from:
Context Dependence: The performance of AI depends on the context of the inputs, such as training data, model parameters, and time and environmental conditions
Resource Intensity: AI computation requires high-performance hardware support, including complex matrix operations and large memory
The above differences have led to the following compatibility challenges:
Conflict between Probabilistic and Deterministic Data
How to transform the probabilistic output of AI into the deterministic results required by blockchain?
When and where should this transformation be performed?
How to retain the value of probabilistic analysis while ensuring determinism?
Gas Cost: The high computational demand of AI models may result in prohibitive Gas fees, limiting their application on blockchain.
Memory Constraints: The limited memory capacity of the blockchain environment may not meet the storage requirements of AI models.
Execution Time: The block time of blockchain poses a constraint on the running speed of AI models, which may affect their performance.
Integration of Data Structures: AI models use complex data structures, which are difficult to directly integrate into the storage model of blockchain.
Oracle Problem (Verification Requirement): Blockchain relies on oracles to obtain external data, but how to verify the accuracy of AI computations remains a challenge. Especially since AI systems require rich context and low latency, which conflicts with the characteristics of blockchain.
Original image from francesco, compiled by TechFlow
How can AI agents seamlessly integrate with blockchain?
Original image from francesco, compiled by TechFlow
Rei proposes a novel solution that combines the strengths of AI and blockchain.
Original image from francesco, compiled by TechFlow
Rather than forcibly integrating AI and blockchain, which are fundamentally different systems, Rei prefers to act as a "universal translator" to enable smooth communication and collaboration between the two.
Original image from francesco, compiled by TechFlow
Rei's core objectives include:
Enabling AI agents to think and learn independently
Translating the insights of agents into precise and verifiable blockchain operations
Original image from francesco, compiled by TechFlow
The first application of this framework is Unit00x0 (Rei_00 - $REI), which has been trained as a quantitative analyst.
Rei's cognitive architecture consists of the following four layers:
Thinking Layer: Responsible for processing and collecting raw data, such as chart data, transaction history, and user behavior, and identifying potential patterns.
Reasoning Layer: Based on the discovered patterns, adds contextual information, such as dates, times, historical trends, and market conditions, to make the data more multidimensional.
Decision Layer: Formulates specific action plans based on the contextualized information provided by the Reasoning Layer.
Action Layer: Translates decisions into deterministic operations that can be executed on the blockchain.
Rei's framework is built on the following three core pillars:
Original image from francesco, compiled by TechFlow
Oracle (similar to neural pathways): Transforms the diverse outputs of AI into unified results and records them on the blockchain.
ERC Data Standard: Expands the storage capabilities of blockchain to support the storage of complex data patterns, while retaining the contextual information generated by the Thinking and Reasoning Layers, enabling the transformation from probabilistic data to deterministic execution.
Memory System: Allows Rei to accumulate experience over time and access previous output results and learning outcomes.
The specific manifestations of these interactions are as follows:
Original image from francesco, compiled by TechFlow
The Oracle bridge identifies data patterns
ERCData is used to store these patterns
The Memory System retains contextual information to better understand the patterns
Smart contracts can access this accumulated knowledge and take action accordingly
With this architecture, Rei agents have been able to combine on-chain data, price movements, and social sentiment, to perform in-depth analysis of Tokens.
More importantly, Rei not only can analyze data, but also form deeper understanding based on it. This is due to the fact that she stores her own experiences and insights directly on the blockchain, making this information a part of her knowledge system, which can be called upon at any time, thereby continuously optimizing her decision-making capabilities and overall experience.
Rei's data sources include Plotly and Matplotlib libraries (for chart plotting), Coingecko, Defillama, on-chain data, and Twitter's social sentiment data. Through these diverse data sources, Rei is able to provide comprehensive on-chain analysis and market insights.
With the Quant V2 feature update, Rei now supports the following analysis forms:
Project Analysis: In addition to the existing functions, quantitative indicators and sentiment data support have been added. The analysis content includes candlestick charts, engagement charts, holder distribution, and PnL. (Related example)
Inflow and Outflow Analysis: By monitoring the prices and trading volumes of popular Tokens on the chain, Rei is able to compare these data with the inflow and outflow of funds, helping users discover potential market trends. (Related example)
Interaction Analysis: Evaluating the overall interaction of the project, including the comparison of real-time data and data from 24 hours ago, as well as relative price changes. This function reveals the correlation between the latest information and user interaction performance. (Related example)
Top Category Analysis: Analyze the lowest trading volume and highest trading count in a single category, highlighting the project's performance within its category.
The first chart shows the trading volume at the bottom and the trading count at the top; then it delves into a single category, revealing the indicator changes of individual projects relative to their peers. (Related example)
In addition, as of January 2025, Rei has supported on-chain Token buying and selling functions. She is equipped with a smart contract wallet based on the ERC-4337 standard, making transactions more convenient and secure.
(TechFlow note: ERC-4337 is an Ethereum improvement proposal that supports account abstraction, aimed at enhancing the user experience).
Rei's smart contract, through user signature authorization, delegates operations to her, allowing Rei to autonomously manage her investment portfolio.
Here are Rei's wallet addresses:
EOA Wallet (Signature Wallet):
https://basescan.org/address/0x3BC4c3A2a2Fa5ad20a2B95B18CA418D06A360cB
Smart Wallet (Account Abstraction Wallet):
https://basescan.org/address/0xf6835acc8d2b51e5d47632ca8954bfee9a0ce49c
Use Cases: The Versatility of the Rei Framework
Original image from francesco, compiled by TechFlow
The Rei framework is not limited to the financial field, but can also be applied to the following wide range of scenarios:
User and Intelligent Agent Interaction: Supporting content creation
Market Analysis: Supply chain management and logistics
Building Adaptive Systems: Governance scenarios
Risk Assessment: In the medical field, Rei assesses potential risks through contextual analysis
Rei's Future Development Direction
Token-permission-based Alpha Terminal
Developer Platform
