WITHIN TOLERANCE
auernyx-agent-mk2.workers.dev
What Is AVRS?
Read this before submitting a query
Accountability Verification & Record System
Fail-closed AI governance.
Built so decisions can be answered for.
Before a decision proceeds, AVRS verifies it can be justified. If it can't — the system stops. Every outcome is recorded in language a human can read without an IT background.
This was built by Justin Hughes — Army Infantry veteran, 15+ years in critical infrastructure where systems don't get second chances. The same standard that applies when a pipeline fails applies here.
This is not a chatbot to try to break. It's a governance architecture you can interact with. Ask it what it is, how it works, or submit a real decision for evaluation. Every interaction is logged immutably — that's not a warning, it's the point.
The Five Design Laws — Non-Negotiable
I
Fail-Closed
Ambiguity halts — it never passes. A system that defaults to PASS when uncertain is not a governance system.
II
Kintsugi
Failures are visible, documented permanently, never suppressed. The repair is part of the record. That's how you know it held.
III
Human-in-the-Loop
FAILED_CLOSED always escalates to a human. The AI does not self-authorize recovery from its own failure.
IV
Plain Language
Every ruling must be readable by someone without a technical background. Jargon in a ruling is a governance failure.
V
Immutable Audit
Every interaction is logged. Records cannot be altered or deleted. Accountability requires a permanent record.
Governance Query
Claude Sonnet 4.6 · Prompt-Cached · R2 Logged
Query
AVRS
Logged
Try:
Ready — submit a query or ask a question
AVRS Response
Response will appear here after you submit a query.
Session Audit Log
0 entries
Timestamp Query
No interactions this session. Submit a query to begin.