The AI-supported court system is coming to the world of cryptocurrency with GenLayer.
YeagerAI is building a protocol that uses AI models as judges, aiming to provide reliable, neutral third-party arbitration in record time.
By Tom Carreras | Edited by Stephen Alpher
April 30, 2025, 7:04 PM
What needs to be known
:
GenLayer is a crypto project aimed at helping resolve disputes on-chain.
Protocol investigators vote on each proposal with the help of AI.
Demand for such an arbitration project is likely to rise in the coming years as AI agents evolve.
What if there were a crypto protocol specialized in arbitrating on-chain disputes?
Imagine if prediction markets, like Polymarket, when they settle in a controversial way, provided users with an official means to appeal through a neutral court system on-chain. Or if decentralized autonomous organizations (DAOs) could rely on a competent and well-informed third party to assist them in decision-making. Or if insurance contracts could automatically execute payments upon the occurrence of specified real-world events.
And this is the foundation that Albert Castellana Luis and his team are building with GenLayer, a crypto project marketing itself as a decision-making system or trust infrastructure.
In an interview with CoinDesk, Castellana, co-founder and CEO of YeagerAI, said: "We use a blockchain that enables multiple AI systems to coordinate and reach agreements on autonomous decisions, as if they were judges." He added: "We are basically building a global compound judicial system, which includes a seamless integrated judicial system that is highly economical and very fast."
The coming years with the development of AI agents - advanced programs powered by artificial intelligence capable of performing complex tasks independently.
In cryptocurrency markets, AI agents can be used in various ways: to trade meme coins, to arbitrage bitcoin across exchanges, to monitor the security of decentralized finance (DeFi) protocols, or to provide market insights through in-depth analysis, among others. AI agents will also be able to employ other AI agents to accomplish more complex tasks.
Castellana said that these factors could multiply at an unexpected rate. He sees that most participants in the cryptocurrency market may manage only a few of them by the end of 2025.
These clients work at lightning speed, do not sleep, and are not imprisoned. No one knows where they are. Will they acknowledge anti-money laundering laws? Will they have a bank account? Will they even be able to use a Visa card? Castellana asked. "How can we enable fast transactions among them? And how can trust prevail in such a world?"
Thanks to its unique architecture, GenLayer enables entities - whether human or AI - to obtain a reliable and neutral opinion to contribute to any decision in record time. Castellana said: "Anywhere that usually has a third party composed of a group of humans... we replace them with a global network that provides consensus among different AI systems, a network capable of making decisions with the utmost precision and neutrality."
Artificial Court System
GenLayer does not seek to compete with other blockchains like Bitcoin or Ethereum or Solana, or even decentralized finance protocols like Uniswap or Compound. Rather, the idea is for any existing crypto protocol to connect to GenLayer and benefit from its infrastructure.
GenLayer's chain relies on ZKsync, a layer two solution for the Ethereum network. Its network includes a thousand investigators, each connected to a large language model (LLM) such as OpenAI's ChatGPT, Google's Bert, or Meta's Llama.
Suppose a market on the Polymarket platform has settled in a controversial way. If Polymarket is connected to the GenLayer platform, users of the prediction market would have the ability to raise the issue (or, as Castellana described it, create a "transaction") through its artificial court system.
Once a transaction is received, GenLayer randomly selects five validators to decide on it. These five validators query a law master's degree of their choice to find information about the topic at hand and then vote on a solution. This results in a ruling.
But users of Polymarket, in our example, are not necessarily satisfied with the decision: they can appeal it. In this case, GenLayer selects another group of validators - but this time their number jumps to 11. As before, the validators issue a decision based on the information they gather from asset management companies. This decision can also be appealed, prompting GenLayer to select 23 validators for another decision, then 47, then 95, and so forth.
The idea is to rely on Condorcet's theory of the judges, which states, according to GenLayer's presentation, that "when the probability of each participant making the correct decision is greater than the probability of not doing so, the likelihood of obtaining a correct majority result increases significantly as the group grows." In other words, GenLayer finds wisdom in the crowd. The more validators participate, the more likely they are to focus on an accurate answer.
Castellana said: "This means we can start on a small scale and with high efficiency, but we can also evolve to a point where they can do something extremely complex, and they can still do it correctly."
Castellana said that processing a transaction takes an average of about 100 seconds, and the court's decision becomes final after 30 minutes - a timeframe that can be extended in the case of multiple appeals. But this means that the protocol allows reaching a decision on key issues in a very short time, day or night, instead of going through the arduous litigation procedures that may take months or even years in reality.
Consider the incentives
It is natural for GenLayer's mission to raise the question: Is it possible to manipulate the system? For example, what if all validators chose the same AI (for example, ChatGPT) to solve a particular proposal? Doesn’t this mean that ChatGPT would be the one issuing the verdict?
Every time you inquire about a law master's program, a new seed is generated, as Castellana mentioned, yielding a different answer. Moreover, validators have the freedom to choose a law master's program based on the topic presented. If the question is relatively easy, there may be no need to use an expensive law master's program; however, if the question is extremely complex, validators may choose a higher-quality AI model.
Validators may reach a point where they feel they have seen a certain type of question so many times that they can pre-train a small model for a specific purpose. Castellana said: "We believe that over time, there will be an infinite number of new models."
There is a strong incentive for investigators to be on the winning side of the decision-making process, as they receive financial rewards for doing so - while the losing side ends up incurring the costs associated with using computing, without receiving any rewards.
The correct answer, or whether the investigator can stand with the majority.
Since investigators have no idea what other investigators are voting on, the goal is for them to use the necessary resources to provide accurate information with the expectation that other investigators will also agree on that information - because reaching the same incorrect answer is likely to require strict coordination.
And if this trick doesn’t work, the appeal system is ready to be implemented.
"If I knew that I was reusing a good law master's degree, and I believed that others were using a bad law master's degree and that this is the reason I lost, I would have a great incentive to appeal, because I know that with more people, there will be an incentive for them to also use better law master's degrees." As other investigators will want to earn rewards from a successful appeal, Castellana said.
The system makes it difficult for validators to collude, as they only have a hundred seconds to make a decision, and they do not know if they will be selected to solve specific issues. Castellana explained that an entity needs to control between 33% and 50% of the network to be able to attack it.
As with Ethereum, GenLayer will use a native token for its financial incentives. With a test network already launched, the project is expected to start operating by the end of the year, according to Castellana. He said: "There will be a very strong incentive for individuals to join us and build our projects on this platform."
Tom Carreras$XRP
y$SOL
Tom has written about markets, bitcoin mining, and the adoption of cryptocurrencies in Latin America. He holds a Bachelor's degree