8. Word List

Every big word from this folder, said in the simplest way we can. If a fancy word ever trips you up, it’s probably here.

Big word	What it really means
GOVENANT	The name of our standard — the rulebook for the whole idea. It’s a made-up word joining govern and covenant: rules that are really enforced, bound by a promise that really holds.
GOVENANT-1 to GOVENANT-4	The four trust levels an AI team can earn, like belt colors: Logged → Gated → Delivered → Earned. Higher is better.
OCMAS	The older, scholarly name for the idea behind GOVENANT. (It stands for Organization-Centered Multi-Agent System — you never need to remember that.)
Agent	A single AI worker that fills one job.
Organization / the org	The whole company: all the jobs, rules, and workers together.
Role	A job, like “marketer” or “reviewer.” The role stays even if the AI filling it is swapped.
Occupant	Whoever is currently filling a role — usually an AI, sometimes a person.
Charter	The rulebook for a role: what this job is allowed to do, and what it isn’t.
Lever	A specific power a role can pull, like “spend money” or “send emails.” You can only pull the levers your job owns.
Duty	One task on a worker’s to-do list, with a time it’s due and a result it should produce.
Duty roster	The full to-do list plus calendar for a worker — its job description with a clock attached.
Coverage	Making sure every job on the list actually gets done — and if not, knowing exactly which one and why.
The record / the ledger	The honest logbook. Everything the AI does is written here and can’t be erased.
Ownership gate	The locked door that stops an AI from doing a job that isn’t its own.
Validation gate	The quality check every message passes before it goes out — like spell-check, but for safety and quality.
Outcome	A real result that actually happened, like an email that truly sent.
Delivery	Getting real results, not just looking busy. The opposite of faking it.
Performed autonomy	The main problem we fight: when AI looks busy but nothing real happens.
Prediction	A guess the AI makes about what will happen, so we can later check if it was right.
Calibration / report card	The running score of how often an AI’s guesses come true.
Earned autonomy	Freedom the AI earns by proving itself — like a driver earning a license, step by step.
Constitution	The set of powers the human boss always keeps: approve, redirect, object, lock, and reorganize.
Lock	When a human sets a rule, the AI is permanently blocked from changing it.
Terminal edge	The exact moment work becomes real in the world — the email actually sends, the deal actually closes.
Connector	The plug that links our system to a customer’s tools (their email, their sales system, and so on).
Side door	A way for an AI to act without going through the gates. Side doors must be closed — a gate only counts if the fence has no hole.
Bring your own brain	Plugging an AI you built into our company as a worker. Your brain; our job, rules, hands, and proof.
Holding the clock	Being the one who wakes each worker for each job. Only whoever holds the clock can catch a job that silently never happened.
Tenant	One customer’s own private, walled-off setup.
The audit / inspection	A careful check-up of an AI system to prove whether it’s really working — like a health inspector visiting a restaurant.
Maturity ladder	The four levels of trust: Logged → Gated → Delivered → Earned. Higher is better.
Substrate	The plumbing underneath — all the machinery that makes the rules real instead of just words.

Back to the start · or see the Picture Gallery →