Starknet's 'Internet Court': A Legal Fiction for AI Agents or a Real Infrastructure Play?

An 'Internet Court' just launched on Starknet—a dedicated on-chain dispute resolution system for autonomous AI agents. The press release calls it a 'foundational shift' for agentic commerce. I read it three times. I still don't know who built it, how the smart contract works, or whether a single line of code has been audited. We mined liquidity while the code slept. Now we're building a courthouse before the first contract is signed. Starknet is a ZK-rollup on Ethereum, offering low fees and native account abstraction. Account abstraction allows AI agents to hold their own wallets and transact autonomously without human keys. That makes Starknet a natural home for agentic commerce—a future where bots negotiate, buy, and sell without human intervention. But commerce breeds disputes. The Internet Court claims to be a set of smart contracts that arbitrate these conflicts. The concept is elegant. The execution is a black box. Let's talk about what we don't know. First, the dispute resolution mechanism. Is it a traditional jury system like Kleros? An AI judge trained on legal documents? Or a simple escrow with a timeout? The announcement says nothing. Based on my experience reverse-engineering the Parity multi-sig vulnerability in 2017, a missing specification often hides dangerous assumptions. I spent two weeks tracing call dependencies to understand how 150,000 ETH got drained. If this court uses an AI oracle to evaluate evidence, that oracle is a central point of failure. If it relies on human jurors, the latency kills agentic speed. Neither is trivial. Second, security. The court lives on Starknet, which inherits Ethereum's security via ZK-proofs. But the court's own contract logic remains unaudited and mostly undisclosed. I've audited yield-farming contracts where one wrong function visibility led to a 2M USD drain. In 2022, during the Terra collapse, I mapped the liquidation cascade in real time—identifying the exact price thresholds that triggered the domino effect. The lesson: complexity kills liquidity when trust is fragile. This court adds a judicial layer that must be mathematically proven correct. Without formal verification, it's a gamble. During DeFi Summer 2020, I deployed 50k USD into Uniswap V2 pairs and learned that yield can mask risk. This court's 'zero-friction' narrative masks a risky dependency chain. Third, data availability. How does the court receive off-chain information? If an agent claims a delivery didn't arrive, who provides the proof? A simple oracle feed? That's trust. Multiple oracles? That's complexity. The announcement mentions no oracle solution. In 2024, I executed 450+ micro-arbitrage trades using a custom Python script that monitored on-chain transfers versus exchange inflows. That script relied on clean, real-time data. A dispute resolution system without a defined oracle layer is like a trading strategy without an exit plan. Here's the contrarian angle: the name 'Internet Court' is brilliant marketing but a legal liability. Calling it a court implies jurisdiction, enforcement, and due process. None of that exists on a permissionless blockchain. If a rogue AI agent refuses to comply, the court can only freeze its on-chain collateral. That's not justice; it's escrow with a fancy UI. Retail investors might hear 'court' and think it's the next big thing. Smart money knows to look for the GitHub repo first. I've seen this pattern before—projects that launch with a narrative but no code. We traded hope for efficiency, then lost both. This feels similar. Also, the timing. AI agents are still a niche. Most are simple chatbots running on APIs. The idea that they'll engage in complex commercial disputes requiring legal arbitration is premature. We're building infrastructure for a ghost town. In 2026, I launched 'The Oracle's Hand', a copy-trading platform where AI agents execute my historical signals. During a flash crash, the AI failed to pause, but my manual override saved 15% of the community's funds. That experience taught me that human intuition remains the ultimate circuit breaker. This court claims to automate justice, but it needs a human-in-the-loop. Without that, it's just an expensive subroutine. So what do we do? Watch. Don't trade. The Starknet ecosystem needs more than press releases; it needs verifiable proofs. If this court ever opens its source code, I'll run a manual audit myself. Until then, my position is simple: liquidity is just trust, digitized and leveraged. And I don't trust a court without a bench. We rode the wave until it broke our boards. This time, let's wait for the code.