AxiomSeal

X-Ethics

What Traktat X is — and why we keep the math private

AxiomSeal's validation pipeline is grounded in a formal axiom system called Traktat X, written by Patrick Müller. Here is what it does, what you get from it — and why we will not publish the formulas.

12.06.20263 min read

Every page on this site mentions "formal axioms". This post explains what that actually means — in plain language, without publishing the mathematics. Both halves of that sentence are deliberate.

Where the axioms come from

The axiom system behind AxiomSeal is called Traktat X. It was written by Patrick Müller (millerlab.io) and is used in AxiomSeal under license. It is not a policy document, not a prompt, and not a set of guidelines someone can quietly rewrite. It is a small chain of formal statements about logic, truth and coherence — the kind you can check a claim against and get a yes, a no, or a measured score.

What it does inside AxiomSeal

You never interact with the axioms directly. They sit behind three things you can see and test:

  1. The 15 validation tests. Every memory written through AxiomSeal runs through 15 checks — coherence, contradiction, hallucination markers, epistemic honesty, technical depth and more. Several of these tests are direct operationalizations of Traktat X axioms. A claim that contradicts itself, or asserts certainty it cannot have, loses score before it is stored.

  2. The X-alignment score. Each memory carries a score that reflects how well it holds up against the axiom set. You see the number and the per-test reasoning; the scoring rules derive from the axioms.

  3. The quarantine. When the validator concludes that a memory conflicts with what is already established, it is flagged or quarantined instead of silently stored. There is deliberately no override button — a governance verdict you can overrule at will is an opinion, not a verdict.

In short: the axioms are the reason AxiomSeal can say why a memory was accepted, graded down, or rejected — instead of pointing at a vibe.

Why we do not publish the formulas

Three honest reasons:

  • It is not ours to publish. Traktat X is Patrick Müller's work. We license its application; the work itself remains his.

  • Black-box by design is part of the product. Customers get scores, verdicts and per-test reasoning — fully auditable outputs. The derivation stays server-side, the same way a credit-scoring company shows you the score and the factors, not the model weights.

  • Publishing would not make you safer. What protects your data is that every write is tested and sealed (SHA-512 chain), and that you can verify the seal yourself at any time. That verification is public API. The mathematics behind the scoring adds nothing to your audit trail — the tamper-evident chain is the audit trail.

What you can verify yourself

You do not have to trust any of this on our word, which is rather the point:

  • Write a memory and read the per-test results in the API response.
  • Ask for any memory's seal via GET /memories/{id}/verify — the SHA-512 recomputation happens in front of you.
  • Run GET /chain-status and check the entire chain of your tenant.
  • Try to store a claim like "refunds never fail" and watch the pipeline reject it.

The axioms stay private. Everything they produce is inspectable. That is the deal — and we think it is the right one.

Questions about the framework? Book a call — or read the X-Ethics page for the longer version.