Elaine, Jereyme | Author
Sissi@TEDAO｜Translator
Translator’s Note:
This translation will introduce the innovative proposals that received funding from the Token Engineering Commons (TEC) in the spring of 2024. TEC is one of the important communities in the world that supports and promotes Token Engineering. It is committed to creating and maintaining a sustainable ecosystem and providing a support and collaboration platform for community members through its forums and other resources.
This project utilizes reinforcement learning and agent-based modeling and simulation technology to optimize the bonding curve mechanism in the token ecosystem. By exploring and responding to potential malicious strategies under different PAMM & SAMM bonding curve combinations, the project aims to significantly improve the economic security of the token system. In addition, the project is also committed to promoting the popularization and practice of Token Engineering so that more people can understand and participate in this cutting-edge technology, paving the way for building a more secure and sustainable token ecosystem.
1. Proposal Details
1.1 Background Overview
The bonding curve is an indispensable part of the token ecosystem. It plays a key role in controlling token price fluctuations, providing necessary liquidity, and dynamic token supply. By mathematizing the relationship between multiple elements in the token ecosystem, the bonding curve also opens the door to "engineering control" of the token ecosystem.
As early as 2018, the IncentiveAI team proposed the idea of using AI-agent for mechanism optimization. By observing the behavior of greedy Machine Learning agents, they identified the possible behavior of users after the system was deployed in the real environment, and continuously optimized the mechanism design by comparing the difference between the actual behavior and the expected behavior. They also applied this concept to the research on the bonding curve of the Ocean protocol. Unfortunately, the project was not implemented on a large scale, and no project code can be found for reference or operation.
Since 2023, BCRG (Bonding Curve Research Group) has conducted comprehensive research, development, education, and application of bonding curves, especially in the joint research of bonding curves of PAMM (Primary Automated Market Maker) and SAMM (Secondary Automated Market Maker). However, according to BCRG's description in Modeling & Simulating bonding curves, due to resource limitations, it has not yet entered into deeper research such as malicious strategy exploration, penetration testing, and hypothesis analysis.
Our team has long been focusing on the exploration of the field of Token Engineering, and is committed to using agent-based modeling and simulation to solve the design and optimization problems of complex systems.
1.2 Project Introduction
In this proposal, we aim to inherit the concept of Incentive AI, explore the malicious strategies of potential attackers under different PAMM and SAMM bonding curve combinations through AI-agents trained by reinforcement learning, and find relatively stable and high-quality bonding curve parameter combinations through further comparative analysis and behavior space exploration, so as to continuously optimize the mechanism design of the protocol, narrow the gap between expected behavior and actual behavior, and reduce the economic security risk of the token ecosystem.
Specifically, in the selection of PAMM bonding curve, we selected the four most common types of Linear, Exponential, Power and Sigmoid; in the selection of SAMM bonding curve, we selected the two most common types of constant product (e.g Uniswap) and hybrid (e.g Curve), which resulted in 8 combinations of PAMM and SAMM. We will use agent-based modeling and simulation methods to conduct experiments, use AI-agent to explore the potential malicious strategy set of each scheme and the probability of each occurrence, and intuitively display the consequences of malicious strategies on the system through simulation results, and try to explore relatively scientific malicious attack response strategies and bonding curve mechanism optimization solutions through experiments.
At the same time, we have applied for Holobit's premium account sponsorship, and will use this advanced modeling and simulation platform to make our model building details and the entire experimental process fully transparent.
Possible innovations
I. Introduce reinforcement learning into Token Engineering to form a set of protocol mechanism optimization methods based on AI-agent and agent-based modeling and simulation;
II. This method is universal, feasible, and reusable, and may be helpful to the economic security of the entire token ecosystem;
III. Thanks to the powerful tool Holobit, this model can be understood, used and verified by the public.
Short-term goals of the project
I. Use AI-agent to explore potential malicious strategies under different PAMM and SAMM bonding curve combinations, identify possible risks under various mechanism combinations, and explore corresponding risk response strategies and mechanism optimization solutions;
II. Provide a relatively scientific and rigorous research method for the development of bonding curve;
III. Based on the experimental results, several suggestions are put forward to improve the economic security of the token ecosystem from the perspective of bonding curve.
Long-term goals of the project
By combining the popularization of AI's agent-based modeling and simulation methods with the promotion of Token Engineering, everyone can become a Token Engineer, thus laying a solid foundation for "building a more anti-fragile and sustainable token ecosystem in a decentralized manner in a community-driven manner", further promoting Token Engineering, and accelerating its development in theory and practice.
2. Expected results
Using Holobit for agent-based modeling, the following outcomes are expected to be delivered:
A token economy off-chain simulation model that introduces AI-agent, including 8 experimental schemes of PAMM and SAMM combinations. At the same time, the model is completely transparent and can be understood, used and verified by everyone;
A research report on potential malicious attack strategies under different PAMM and SAMM bonding curve combinations explored by AI-agent (including modeling process, experimental content, vulnerability risks and optimization solutions).
3. Mission and Values Alignment
Convenience: Holobit supports public sharing, and the modeling logic is simple, visual and intuitive, ensuring that everyone can read, use and verify. Therefore, this model can be opened as a public good, and everyone can access and test it, such as the Terra/LUNA ecosystem case that has been given.
Education: Through detailed models and simulation tutorials, the project can help the public gain a deeper understanding of how bonding curves work and their key role in the token ecosystem; through agent-based modeling and simulation, the project can show the public how to analyze and deal with dynamic relationships and potential risks in complex systems. This skill is widely applicable and is also a key skill for studying Token Engineering. If this model can be used to popularize this set of methodologies and tools in the community, it can further promote the popularization, development and practical application of Token Engineering.
Transparency: Only when the public can understand it can it be considered truly transparent. This model does not involve code, and the modeling mechanism and experimental process are visualized through the Holobit tool. Through modeling and experiments, not only the mechanism of the model is transparent, but also the risks of the mechanism design are further transparent, and specific repair suggestions are given.
Community-driven: The community can fork this model to conduct various experiments, not limited to bonding curves, but also for governance, growth, and other research. More importantly, this set of methodologies and tools can be reused on other protocols. Everyone can publish their research results in the community, disclose the loopholes and optimizations of a token ecosystem, and truly achieve community-driven self-regulation.
Aligned with the Token Engineering principles: After mastering this set of methods and tools, everyone can use these skills to perform economic security audits of protocols. Therefore, "decentralized completion of token engineering" becomes possible, and we can pool the power of collective wisdom to build a more anti-fragile and sustainable token ecosystem.
