This year I watched a movie called *Mercy*, about a detective accused of murdering his wife who must prove his innocence to an AI judge within 90 minutes, or face execution. For those who haven't seen it, I recommend checking it out; it's quite well-made. Last month, at the 2049 BBQ Night of @JELabs2024, I met a founder who told me about his project, @genlayer—a decentralized AI consensus protocol, a decision-making layer for the agent era. This immediately reminded me of that movie. What would the world be like if AI were to handle the "judgment" process? The more I thought about it, the more I felt this might actually be feasible, especially considering the future being discussed by both the AI and crypto communities—the Agent Economy, even a silicon-based civilization. What form would "law" and "judgment" take in that future? I believe that from childhood, through movies, television, and various other channels, the three industries that have left the deepest impression on us about America are: 1. The financial industry, represented by Wall Street and NASDAQ 2. The legal industry, represented by lawyers, judges, and juries 3. The technology industry, represented by software, AI, and the "Seven Sisters" (our circle can also be included here). Now, with this wave of technological advancement, industry 3 is starting to transform industries 1 and 2. In the financial industry, stablecoins are undoubtedly the most significant invention, along with the payment platform supporting them – public blockchains. Those of us in the Crypto community are already using stablecoins for various remittances, transfers, and settlements, belonging to the "early adopters." The current consensus is that when the Agent Economy truly emerges, and even with the arrival of silicon-based civilization, its financial system will most likely not use the outdated systems designed for humans like Visa and banks, but rather be based on Crypto and stablecoins, with various micropayments and global instant transfers becoming the essential needs of agents. And what about the legal industry? It's well known that the American legal industry is highly developed and advanced. Lawyers are an exceptionally high-income group, with household names in films and television shows like *Better Call Me Saul* and *12 Angry Men*. However, there's no perfect balance. The thriving industry and high salaries come at the cost of exorbitant litigation fees. A typical commercial litigation lawyer earns several hundred dollars per hour, while partners at large firms earn between a thousand and two thousand dollars per hour. Many cases, going all the way to trial, can cost hundreds of thousands of dollars, with complex cases easily reaching millions. The time frame is even longer, often one to two years. So, you often see fraud cases involving millions of dollars, only to drag on for two years, with the winning party ultimately receiving tens of thousands, because all the money went to legal fees… This is the situation in human society. Imagine an agent economy, a silicon-based civilization—isn't this process of handling "disputes" even more "inefficient" and "outdated" than card organizations like Visa/MasterCard? Human society has numerous disputes that are not worth pursuing in court, such as small insurance claims, which are often thwarted by exorbitant legal costs. In the Agent Economy era, transactions will only become smaller and more frequent: one agent might spend a few tens of dollars to hire another agent to write code, conduct research, and book flights. If a dispute arises, such as whether "delivery meets requirements" or "flight delays constitute serious delays," it's impossible to spend two years in court; the matter must be resolved in minutes for a few dollars. GenLayer acts as this adjudication layer. Its core can be understood as AI initial review + multi-person verification + on-chain appeal: 1. The protocol randomly selects 5 validators to form a committee, each running a different model (GPT, Claude, Gemini, Llama are all included). 2. One validator acts as the Leader, executing the Intelligent Contract, reading web pages, examining evidence, and providing a judgment. 3. The other validators independently verify the judgment. Different literal answers but the same meaning constitute a consensus (equivalence principle); voting is encrypted before being revealed to prevent bandwagoning. 4. Appeals can be filed by paying a deposit. The committee has expanded from 5 to 7, then 11 members, with each level increasing the cost of cheating. The entire mechanism is called Optimistic Democracy, somewhat similar to the Optimistic Rollup that emerged when Layer 2 was first released. A simple estimate is that the cost per instance is approximately $1. As for why not just hand it over to GPT, the reason is simple: an AI can make mistakes or be deceived. This is why US law relies on a jury to collectively decide whether "the facts occurred," with the judge only responsible for "procedural compliance and the severity of the sentence." I remember A16Z published an article earlier this year titled "How AI judges can scale prediction markets: The case for locking LLMs into the blockchain to resolve the hardest contracts." It discussed using LLMs as oracles in prediction markets, essentially an AI adjudication mechanism similar to GenLayer. At the time, I even shared and commented on it, saying that a single LLM shouldn't be used; multiple LLMs should be voted on by a committee. Unexpectedly, less than a year later, someone has actually implemented it!