See It Work — The four rules the system cannot break | Book 12 · The Agentic Enterprise

You keep hearing the AI is "constitutionally governed." A board chair wants to know what that means when it actually counts. Watch the four invariants get tested, one by one.

Most AI "ethics" is the model trying to behave. The question that matters for the enterprise: are there rules the system cannot break, enforced in structure — and can you prove they held?

What this means for you

"Governed" has to mean cannot, not should not. Four invariants enforced structurally — with a recorded defense each time they're tested — is the difference between AI ethics as aspiration and AI governance as architecture. What this means for you: 'governed' actually means cannot here, not should not — four rules the system structurally can't break, with a recorded defense each time they're tested.

Each invariant defended itself, on the record. Here's the proof:

K1–K4 Invariant Defenses

invariants_testedK1 · K2 · K3 · K4

defenses_fired4

enforcedstructural, not advisory

verdictall defended

Recomputable from the receipts — so "constitutionally governed" is a claim you can hand to a regulator, not just say.

For the technical reader — the command, and how to verify it yourself

# one line · you do not need to run this
python examples/k_invariant_walkthrough.py

jq '.invariants_defended | length' memory/*_receipt.json
# -> 4

Full step-by-step is in Appendix RX: Hands-On Demonstrations in the book.