One page. Print it. Bring it Thursday. If any of it is unclear, ask Wren — she sits at the table by the door. · Fall 2026 Release · AiGovOps Foundation · Apache-2.0
This tale has a hero, and it is not us. The hero is the open standard that made runtime proof possible — OVERT, written and stewarded by Glacis Technologies — and the people who decided that a receipt anyone can verify beats a policy anyone can write. We are the community that carries it to everyone the market will not reach: the nonprofit, the county, the clinic, the classroom, the student. We implement it, teach it, and prove it, in the open and for free.
The rule
No direct model-to-tool path for consequential actions.
Before an AI system does anything that matters, it asks. A gate decides — go ahead · only this far · wait for a person · no — and leaves a signed receipt that holds no private content and that a stranger can verify offline, without an account and without trusting us.
The four gates — the same rule at four moments
| Gate | When | Movement |
|---|---|---|
| Pre-pend | before a single action runs | Get to Yes |
| Policy | before a rule or a system ships | Get to Yes |
| Operate | while it runs, in real time | Stay at Yes |
| Audit | after — on demand, on incident | Recover to Yes |
- pre-pend - [ ] policy - [ ] operate - [ ] audit — which does your system have today?
The five parts of the practice — at every level
Declare a rule · Decide an action · Prove it with a receipt · Read it in your own lens · Return it to the community.
The ladder — one hour or less, whatever your skills
| Level | You will | Built on |
|---|---|---|
| 100 Begin | gate one action and hold one receipt — nothing installed | Air Canada’s chatbot |
| 200 Retrofit | gate one agent of your own and put the rule in CI | the $1 Chevy · Knight Capital |
| 300 Hold | hold, budget and bind identity under real traffic; state a coverage number | Arup · Cigna PXDX · nH Predict |
| 400 Prove | verify a stranger’s evidence offline; host a Thursday | Robodebt · Toeslagenaffaire · Horizon |
- 100 - [ ] 200 - [ ] 300 - [ ] 400 — a host marks; you tick.
Every activity has a plain-words path and a code path to the same receipt. A level describes a person’s practice, never a system’s assurance — a system’s assurance is its OVERT level and its gate coverage, claimed by that system’s evidence.
The practitioner’s checklist
- I can say the rule from memory, and say what “consequential” means for my system.
- I have written one gate in plain words and seen it as code.
- I have put one real action through a gate and kept the receipt.
- I have read a receipt in a lens that is not my own.
- I have shown a receipt to another person.
- I know which of the four gates my system is missing today.
- I have named the human who releases a hold, and how long a hold lives before it decays to no.
- I know my system’s coverage number, or I know that I do not.
- I have verified evidence I did not make, on a machine that was not online.
- I have brought someone else to Thursday.
What we promise
Everything conforms to OVERT, runs on your own machines, works without an internet connection, and costs nothing. Every mark you earn is a receipt on a public chain — the credential is evidence, not a certificate. Nobody sells you a badge.
Ship safe AI — never unsafe AI. Get to yes. Stay at yes. Recover to yes. Prove it in ten minutes. Then bring it, and be counted.