Oceans and Outlaws, LLC · 4dteaching platform
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
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.
The machine renders findings with citations. Accept, reject, buy, treat — those words belong to people.
Every criterion traces to a registered document and edition. “From memory” is not a basis here.
Missing data produces a named list of what’s missing — not a guess wearing a confident tone.
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
No level requires writing code. Every level requires judgment — which is the thing your field already taught you.
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
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
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
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
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