The alignment layer for human-AI collaboration
Create with confidence.
x-Arbiter governs the collaboration, not the model. It carries your team’s intent into every AI session, measures the work against it, and tells you both when it drifts.
For businesses building with AI. Works with any model.
The problem isn’t the model. It’s the collaboration.
Every project has its own vision. Every developer directs the AI differently. The model arrives knowing neither.
66%
of developers get AI solutions that are “almost right, but not quite.”
46% vs 33%
More developers distrust the accuracy of AI tools than trust it.
65%
say the AI misses relevant context when refactoring.
“An agent cannot be expected to fully police its own behavior.”
What changes with x-Arbiter
It starts with your project and the person directing it.
The AI arrives knowing neither the project’s vision nor how this developer works.
Before it writes, x-Arbiter gives it what your project is for — its direction, its standards, its history — and how the person directing it works.
The model starts where your team is, not at the average.
It lets the model work, then measures the result.
Fencing a model in wastes what it can do, and trusting it blindly ships “almost right”.
x-Arbiter doesn’t restrict the model. It measures every commit against your project, and when the work drifts, says where, to the AI and to you, so the next step comes back to what you asked for.
Output aligned with what this person, on this project, is trying to do.
Proof you can hand over.
When a client, an auditor or your own team asks how the AI’s work was checked, a chat log is not an answer.
Every measurement, decision and ruling is kept in the project’s repository and shown across all your projects on the dashboard. The judgment itself is tested on your own work and checked independently, so you know how far to trust it.
A record you can hand to anyone who asks. We’ll publish the judge’s real-work accuracy once there are enough rulings to mean something.
The proof so far
17,331
governance events recorded
822
of 896 findings resolved
279
decision records
Totals across 4 projects that share their governance data with x-Arbiter. Each project counts its own work, with test runs left out. As of October 9, 2026.
- x-Arbiter and this dashboard are built under x-Arbiter, every day. The totals above are counted from that work and from every project that shares its figures.
- Every model that judges here is tested on the project’s own work before its verdicts count, and one that fails is reported as failing, whichever company made it.
- Every measurement, decision and ruling is kept in the project’s own repository, where anyone with access can read it.
How the measuring works, and why a model can judge a model →
What it doesn’t do
- It doesn’t restrict the model or decide for you. It guides, measures and reports; you rule.
- It steers at each commit today, not in the middle of a task.
- Its judge is a model. We measure how often it is right, and your rulings count over its verdicts.
- It doesn’t run your code, your builds or your CI.
Bring it to your team.
x-Arbiter is licensed directly to businesses. Tell us about your team and what you’re building, and we’ll show you what it measures on one of your projects.
Not yet publicly released; licensed directly.