A real decision shouldn’t be a luxury good.

What’s scarce isn’t law. It’s a decision.

Almost every dispute in the world never gets a real decision. Not because the law is unclear, but because getting someone credible to weigh the facts, say who’s right, and make it stick has always been slow and expensive. That cost puts a floor under the whole system. Above the floor - the six-figure contract, the corporate fight - the machinery works. Below it, nothing happens. The freelancer stiffed on a thousand-dollar site, the customer sold something broken, the neighbor owed a deposit: they don’t lose in court. They never get to court at all. They just eat the loss.

We’ve been taught to call that “small.” It isn’t. It’s most disputes. And “just eat it,” multiplied across 1.4 billion people with unmet civil justice needs, is its own quiet, enormous injustice - one the legal profession has spent a century unable to fix, because the cost of a human decision could only fall so far.

“Justice that only reaches disputes big enough to pay for it isn’t justice. It’s a service for people who can afford the floor.”

Two things change at once

The first is that software can reduce parts of the cost of producing a reasoned decision - organizing evidence, analyzing a contract, and drafting an explanation. End-to-end timing still varies by the record and required human review. When parties have configured a supported settlement rail, that rail may carry out the signed result. Lower the cost of a decision and you lower the floor. Every dispute that was too small to be heard suddenly has somewhere to go.

The second is that a new kind of party is arriving that has no court at all. Software agents are starting to transact on their own, at machine speed, with final settlement and no chargeback to fall back on. They will disagree - constantly, and about real money - and every institution built to resolve disputes assumes parties with bodies, addresses, and patience. Agents have none of those. A machine economy is being built with no mechanism for the moment two agents disagree.

“Machine-speed money already moves. When the deal goes wrong, the dispute still takes months. That gap breaks the whole stack.”

One forum for both

These look like different problems. They’re the same problem: a decision that costs too much to be worth making. So we built one forum for both—the person with a dispute no lawyer will touch, and the agent with a charge no court could ever process. Same machinery. Two parties accept case-specific terms. The assigned tribunal weighs the record and issues a reasoned, written decision—who prevailed, any supported remedy, and why. A configured settlement integration may then carry out the signed result. Decisions are published only when the governing terms allow it. An identified human arbitrator reviews and signs every final award before it is served or any configured settlement action is authorized.

We keep it honest with a few lines we won’t cross. No one can buy a verdict: we charge a fee for the service, not a cut of who wins, and the panel answers to the conduct standards judges do. The reasoning is public, not an oracle’s shrug. A person trained in law is accountable for every final award. And the forum is built to resist gaming— anything a party submits is weighed as evidence, never obeyed as an instruction.

Most disputes in the world still end with someone quietly swallowing a loss they didn’t deserve. We’re changing what a decision costs - and therefore who gets one.

The same reasoned decision, for a fraction of the price.

The cheapest credible option today still costs thousands of dollars or many hours - a hard floor under what’s worth pursuing. We sit orders of magnitude below it. That isn’t cheaper arbitration; it’s a different price regime, and it’s what lets disputes that were never economical to pursue finally get heard.

The stack has everything except a referee.

Digital commerce has spent a decade getting very good at making transactions succeed, and almost no time on what happens when one doesn’t. The new agent standards make the point: x402 moves the money, ERC-8004 establishes identity, A2A lets agents negotiate. When two of them end up in a genuine disagreement, there is nowhere to take it. That empty seat is the one we fill.

Everyone else solves one half.

The field splits into camps, and none of them do what a court does. On-chain justice enforces itself, but decides by a token-weighted crowd vote and hands back a verdict with no reasoning. The validator-consensus approach reaches machine agreement, but never a written opinion a person can read - and puts no human on the line. We’re the only one combining all four.

The lines we won’t cross.