Oceans and Outlaws, LLC · 4dteaching platform

THE OUTLAW
AI LAB

Students here don’t learn to code AI. They learn to govern it.

A decision-intelligence lab built on a chartered framework: AI that cites every claim to a registered document, refuses when data is missing, and leaves every decision with a qualified human. The lab is where students and teachers put their own field’s knowledge on that footing.

Phase 0 — What this is

AI you can put your name on

Most AI answers fast and sources nothing. That is useless the moment a decision matters — in a weld shop, a feed lot, a market forecast, a classroom. This lab runs on a different architecture: decision cells. Each cell handles one narrowly defined decision, on a registered document corpus, with deterministic calculation, and it earns the right to operate by passing a frozen benchmark it can never edit.

The first cell — structural-weld visual acceptance to AWS D1.1 — is chartered and in production. The framework it proved is what the lab licenses to build on.

Rule

No verdicts

The machine renders findings with citations. Accept, reject, buy, treat — those words belong to people.

Rule

Cites everything

Every criterion traces to a registered document and edition. “From memory” is not a basis here.

Rule

Refuses honestly

Missing data produces a named list of what’s missing — not a guess wearing a confident tone.

Rule

Human rules

Every cell ends at a qualified person. If a user proceeds over an open finding, the record shows it — informed, and on paper.

Phase 1 — How students climb

The ladder

No level requires writing code. Every level requires judgment — which is the thing your field already taught you.

L0Use

Use a cell

Run the live station. Watch AI cite, compare, and refuse. Ten minutes that recalibrate what you expect a machine to be allowed to say.

Produces → a corrected expectation

L1Break

Break a cell

Take real scenarios and make the platform wrong. Find the input it mishandles, the refusal it shouldn’t have made, the citation that doesn’t hold. Confirmed findings enter the benchmark’s amendment record — with your name on them.

Produces → benchmark findings

L2Author

Author on a cell

Write scenario packets and teaching cases from your own domain knowledge — livestock handling, shop safety, equipment, whatever you actually know. Content, criteria, and cases; the machinery is given.

Produces → curriculum & test cases

L3Build

Build a cell

A team picks a decision it has real authority over — feed rations, withdrawal times, market pricing, lockout procedures — and walks the full pipeline: register the corpus, define what facts the cell may demand, author its benchmark, and face the charter gate. AI drafts. Humans rule. The benchmark judges.

Produces → a chartered decision cell

Phase 2 — The proof

This isn’t a pitch. It’s a chartered method.

The framework’s first cell was stamped from licensed code books, verified line-by-line against the source documents, frozen, and then cold-tested on cases its builders never saw — before a single student touched it. Every artifact in the chain is hash-sealed in a custody registry. Pull any thread; there’s a hash at the end of it.

AIP-CHARTER-WELDVT-001 · SIGNED 2026-08-02 · sha256 7c78d8d7958675d4522063fc2147c35f5d1f7484c2b1898dbad1c168f312bad8

Phase 3 — How a partnership works

Three ground rules, stated up front

  1. The boundary is written before the first workshop. The framework, doctrine, and platforms are Oceans and Outlaws property, licensed. The cells your lab builds on them are yours.
  2. Governance travels with the license. Cells charter only through the acceptance gate — frozen benchmarks, held-out testing, cited criteria, human-ruled outputs. That’s what makes lab output publishable, and it isn’t negotiable per project.
  3. Domain expertise stays with the domain expert. The framework never substitutes for a professor’s market model or a teacher’s classroom judgment. It makes both auditable.