What if there is a crypto protocol that specializes in arbitrating on-chain disputes?
Imagine, every time the predictions like Polymarkets were fixed in a controversial way, users had a formal way to appeal through a type of neutral chain court system. Or if decentralized autonomous organizations (DAO) can rely on a good, knowledgeable third party to help them make decisions. Or if insurance contracts can automatically perform payouts when specific events in the world occur.
That is important what Albert Castellana Lluís developed and his team along with Genlayer, a crypto project that buys himself as a decision -making system, or trust infrastructure.
“We use a blockchain with lots of AIS coordinates and reach the subjective decisions agreement, as if they were a judge,” Castellana, co-founder and CEO of Yeageoi told CoinDesk in a interview. “We usually build a global synthetic jurisdiction with a gem of non -sleeping court system, super inexpensive, and that is very fast.”
The demand for such an arbitration project can be spike in the coming years with the development of AI agents -sophisticated programs enabled by artificial intelligence capable of performing complex activities in an autonomical way.
When it comes to crypto markets, AI agents Can be used in all kinds of ways: For trading memecoins, the Bitcoin arbitraging in exchanges, monitoring of the security protocols, or providing market views through in -depth analysis, to mention only certain cases of use. Ai agents Will also be able to hire other AI agents To complete more complex assignments.
Such agents can grow at an unexpected rate, Castellana said. In his view, most participants in the crypto market may manage a handful of them by the end of 2025.
“These agents, they work fast. They don’t sleep. They don’t get confined. You don’t know where they are. Are they going to the anti-money laundering rules? Will they have a bank account? Can they use a visa card?” Castellana said. “How can we enable fast transactions between them? And how can trust in a world like this happen?”
Thanks to its unique architecture, the Genlayer can provide a solution by allowing creatures – human or AI – to obtain a reliable, neutral opinion that weighs at any decision at the time of the record. “Everywhere you usually have a third party made to a group of people … We have replaced them with a global network that provides a consensus between different AIs, a network that can make decisions in a way that is right and as unbiased as possible,” Castellana said.
Synthetic court system
Genlayer does not seek to compete with other blockchains such as Bitcoin, Ethereum or Solana – or even defi protocols such as uniswap or compound. Instead, the idea is for any existing crypto protocol to connect to the Genlayer and use its infrastructure.
The Genlayer chain is powered by Zksync, a solution to Ethereum Layer 2. Its network is counted by 1,000 validator, each connected to a large language model (LLM) such as Openai’s Chatgpt, Google’s Bert or Meta’s Llama.
Let’s say that a Polymarket market is running in a controversial way. If the polymarket is connected to the Genlayer, users of the prophecy market are capable of raising the issue (or, as Castellana has placed, to create a “transaction”) along with the synthetic court system.
Once the transaction enters, the Genlayer selects five validators randomly to rule it. These five validators have a LLM they choose to find information on the topic at hand, and then vote in a solution. It makes a ruling.
But polymarket users, in our example, do not have to be satisfied with the decision: they may decide to appeal to the decision. In the case of the Genlayer, another set of validators – other than this time, their number jumped at 11. As usual, validators released a decision based on the information they had gathered from the LLMs. That decision can also appeal, making 23 Genlayer’s 23 validators pick up for another decision, then 47 validators, then 95, and so on.
The idea is to rely on the condorcet jury theorem, which according to the Genlayer pitch deck states that “when each participant is more likely than not to make a right decision, the likelihood of a proper outcome of most will increase dramatically as the group grows.” In other words, Genlayer found the wisdom for the most part. The more validators are involved, the more likely they are to be in an accurate answer.
“This means we can start small and very good, but we can also rise to a point where something very, very confusing, they can still get right,” Castellana said.
The average transaction takes about 100 seconds to process, Castellana said, and the court’s decision becomes final after 30 minutes – an hour that can extend if multiple appeals occur. But this means that the protocol can reach a decision on key issues in a very short time, day or night, rather than going through difficult world processes that may last month or even years.
Looking at the incentives
Genlayer’s mission naturally raises a question: Is the system possible? For example, what if all validators choose the same AI (say, chatgpt) to solve a given proposal? Doesn’t it mean that ChatGPT is important to release decisions?
Whenever you ask for an LLM, you make a new seed, Castellana said, so you get a different answer. On top of this, validators have the freedom to choose which LLM to use based on the subject at hand. If this is a pretty easy question, maybe there’s no need to use an expensive LLM; On the other hand, if the question is particularly complicated, the validator may choose for a higher quality AI model.
Validators can end in a situation where they think they have seen a certain type of question many times that they can pre-train a small model for a specific purpose. “We think, over time, there will be just endless new models,” Castellana said.
There is a strong incentive for validators that are in the winning part of the decision-making process, as they are financially rewarded for this-while the loss of the side ends with the existence of costs associated with the use of the calculation, without collecting any rewards.
In other words, the question is not whether a validator provides the right answer, but if it manages the majority.
Because validators have no idea what other validators are voting for, the goal is for them to use the necessary resources to provide accurate information in the hope that other validators will also be involved in that information – as it comes to the same incorrect answer may require strict coordination.
And if that gamble does not work, the appeal system is ready to kick in.
“If I knew I was using a good LLM again, and I think that other people were using a bad LLMS and that was why I lost, then I have a big incentive to appeal, because I know that in more people, there will be an incentive for them to use better LLMs too” because other validators want to earn rewards from a successful appeal.
The system makes it difficult for validators to collude, as they only have 100 seconds to reach a decision, and they do not know if they will choose to solve specific questions. A creature must control between 33% and 50% of the network to attack it, Castellana said.
Like Ethereum, Genlayer uses a native token for financial incentives. In a testnet launched, the project should live by the end of the year, according to Castellana. “There will be a huge incentive for people to come and build things above,” he said.
