What would happen if there were a cryptographic protocol specialized in arbitrating disputes in the chain?
Imagine whether, when prediction markets such as Polymket were established in a controversial way, users had a formal way of appealing through a kind of neutral judicial system in the chain. Or if decentralized autonomous organizations (DAO) could trust an efficient and knowledgeable third party to help them make decisions. Or if insurance contracts could automatically execute payments when specific events of the real world occurred.
That is essentially what Albert Castellana Lluís and his team are building with Genlayer, a cryptographic project that is marketed as a decision -making system or trust infrastructure.
“We are using a block chain that has multiple coordinated AIS and reaching an agreement on subjective decisions, as if they were a judge,” said Castilian, co -founder and CEO from Yeerai to Coindesk in an interview. “We are basically building a global synthetic jurisdiction that has an integrated judicial system that does not sleep, which is super cheap, and that is very fast.”
The demand for such arbitration project can increase in the coming years with the development of AI agents, sophisticated programs promoted by artificial intelligence that are capable of carrying out complex tasks autonomously.
When it comes to cryptography markets, AI agents can be used in all kinds of ways: to market memecoras, arbitragar bitcoin in exchanges, monitor the safety of protocols define or provide market information through in -depth analysis, to quote only a few cases of use. IA agents may also hire other AI agents to complete even more complex tasks.
Such agents can proliferate at an unexpected pace, said Castilian. In their opinion, most cryptography market participants could handle a handful of them at the end of 2025.
“These agents work very fast, they don’t sleep, they don’t go to jail. You don’t know where they are. Are rules against money laundering? Are they going to have a bank account? Can you use a visa card?” Castellana said. “How can we enable rapid transactions between them? And how can trust in a world like this happen?”
Thanks to its unique architecture, Genlayer could provide a solution by allowing entities, human or AI, to obtain a reliable and neutral opinion that evaluates any decision in record time. “Anyway where it normally has a third party made of a group of humans … we replace them with a global network that provides a consensus between different AIS, a network that can make decisions in a way that is as correct and impartial as possible,” said Castilian.
Synthetic Court System
Genlayer does not seek to compete with other blockchains such as Bitcoin, Ethereum or Solana, or even protocols Defi as Uniswap or compound. Rather, the idea is that any existing cryptographic protocol can connect to Genlayer and make use of its infrastructure.
The Genlayer chain works with ZKSYNC, an Ethereum Layer Solution 2. Its network has 1,000 validators, each connected to a large language model (LLM) such as the OpenAI chatgpt, Google Bert or the finish line.
Let’s say a market in Polymarket is established in a controversial way. If Polymket is connected to Genlayer, users of the prediction market have the ability to raise the problem (or, as Castilian said, to create a “transaction”) with their synthetic court system.
As soon as the transaction arrives, Genlayer chooses five random validators to govern it. These five validators consult a LLM of their choice to find information on the subject in question and then vote on a solution. That produces a decision.
But Polymeket users, in our example, do not necessarily need to be satisfied with the ruling: they can decide to appeal the decision. In which case, Genlayer chooses another set of validators, except this time, his number jumps to 11. Like before, the validators issue a decision based on the information that collect from LLMS. That decision can also be appealed, which makes Genlayer select 23 validators for another ruling, then 47 validators, then 95, and so on.
The idea is to trust the Condorcet jury theorem, which according to the Genlayer launching cover states that “when each participant is more likely to make a correct decision, the probability of a result of the right majority increases significantly as the group becomes larger.” In other words, Genlayer finds wisdom in the crowd. The more validators are involved, the more likely they will have to concentrate on a precise response.
“What this means is that we can start small and very efficiently, but we can also climb to a point where something very, very complicated, they can still hit,” said Castilian.
The average transaction takes approximately 100 seconds to process, said Castilian, and the decision of the court becomes final after 30 minutes, a deadline that can be extended if multiple appeals occur. But that means that the protocol can reach a decision on important issues in a very short period of time, day or night, instead of going through arduous litigation processes of the real world that can take months or even years.
Looking at incentives
Genlayer’s mission naturally raises a question: is it possible to play the system? For example, what happens if all validators select the same AI (say, chatgpt) to solve a given proposal? Wouldn’t that mean that Chatgpt will have essentially issued the decision?
Every time you consult a LLM, it generates a new seed, Castilian said, so you get a different response. In addition to that, validators have the freedom to choose which Llm to use depending on the subject in question. If it is a relatively easy question, there may be no need to use an expensive LLM; On the other hand, if the question is particularly complex, the validator can opt for a higher quality AI model.
Validators can even end in a situation in which they feel that they have seen a certain type of question so many times that they can train a small model for a specific purpose. “We believe that, over time, there will be new endless models,” Castellana said.
There is a strong incentive for validators to be on the winning side of the decision -making process, because they are financially rewarded for it, while the losing side ends up incurring costs associated with the use of calculation, without collecting any reward.
In other words, the question is not whether the validator of one is providing a correct answer, but if it achieves the side of the majority.
Since the validators have no idea what other validators are voting, the objective is to use the necessary resources to provide precise information with the expectation that other validators will also converge in that information, because reaching the same incorrect response would probably require rigorous coordination.
And if that gambit does not work, the appeal system is ready to activate.
“If I know that I am reusing a good LLM, and I think that other people are using a bad Llms and that is why I lost, then I have a great incentive to appeal, because I know that with more people, there will be an incentive to use better LLM”, since others validated will want to win the rewards of a successful attraction, Castilian said.
The system makes it difficult for validators to collide, because they only have 100 seconds to reach a decision, and do not know if they will be chosen to solve specific questions. An entity would need to control between 33% and 50% of the network to attack it, Castilian said.
Like Ethereum, Genlayer will use a native token for its financial incentives. With an already launched testnet, the project must be put live at the end of the year, according to Castellana. “There will be a great incentive for people to come and build things at the top,” he said.
