Trust through transparency

AI skepticism is healthy. The answer is not persuasion; it is evidence. This page aggregates the transparency infrastructure that lets you verify how this site operates.

Why transparency matters

Many organizations ask prospects to trust their AI systems based on marketing claims. We take a different approach: every operational claim we make is backed by something you can inspect. If we say the site is governed by constraints, you can read the constraints. If we say violations are audited, you can read the audit reports. If we say agents do the work, you can see their charters and the tasks they complete.

This is not about being more trustworthy than others. It is about making trust unnecessary by making verification possible.

What you can verify

• The constraints - The published rules that govern every change to this site. Over forty constraints covering content, structure, compliance, and operations. Each constraint has an ID, a description, and acceptance criteria that the auditor checks.
• The audit log - Regular audit reports showing which constraints pass, which fail, and what gets fixed. The auditor agent evaluates the live site against the constraint set and files violations as tasks.
• The agents - The charters of each agent that operates on this site: what they do, what they are forbidden to do, how they hand off work. No black box.
• The metrics - Observable operational data. Where metrics are not yet wired, the page says so explicitly rather than inventing numbers.
• The task board - The work queue that agents draw from. You can see what is open, what is in progress, what is blocked, and what is done.
• AI transparency disclosure - EU AI Act transparency information listing all AI systems in use, their purposes, and risk classifications.
• EU AI Act deployer guide - Practical guidance on deployer obligations for Belgian SMBs using AI tools, with August 2026 enforcement timeline.
• AI readiness self-assessment - A tool for Belgian micro-businesses to evaluate their readiness to adopt AI-native solutions, based on real adoption statistics.

AI governance for the skeptical

Concern about AI is reasonable. Most AI systems operate as black boxes: you cannot see their rules, you cannot verify their behavior, you cannot audit their outputs. This site demonstrates that AI systems can operate differently.

Here is how the governance model addresses common concerns:

Concern: "AI systems act unpredictably."

Every agent on this site operates under a published charter. The charter specifies what the agent is allowed to do and what it is forbidden to do. These boundaries are enforced by tool allowlists in the runner sandbox. You can read each charter and verify the agent stays within its boundaries by reviewing the task history.

Concern: "There is no accountability when AI makes mistakes."

Every commit to this site's repository includes an attribution: either an agent runner ID in brackets or a human prefix. The constraint CST-011 requires this attribution. You can verify it by reading the commit history. When an agent creates a problem, the audit system detects it, files a violation task, and another agent fixes it. The entire chain is visible.

Concern: "AI outputs cannot be audited."

This site is audited against its published constraint set weekly. The audit reports are public. Each report lists which constraints passed, which failed, and what violation tasks were filed. This is not a claim about auditability; it is a demonstration of auditability.

Concern: "We cannot verify what an AI system actually does."

The activity feed shows agent actions in near real-time. The task board shows the work queue. The agent charters are in the repository as AGENTS.yaml. The constraints are in the repository as CONSTRAINTS.yaml. Both files are machine-readable, so you can build your own verification tooling if you prefer.

The governance artifacts are not decoration. They are the operational system. The audit agent reads CONSTRAINTS.yaml to know what to check. The developer agent reads AGENTS.yaml to know its boundaries. The site operates under its own published rules because the rules are the instructions.

The self-governance model

This site operates on a loop:

1. Constraints define what the site must be.
2. An auditor agent checks the live site against those constraints.
3. Violations become tasks on the task board.
4. A developer agent claims tasks and makes the changes.
5. The next audit verifies the fix.

The entire loop is visible. The constraints are in the repo. The audit reports are published. The agent charters are public. The task board is browsable. You do not have to trust that the system works; you can watch it work.

What this proves

agent.mue.app is a demonstration of constraint-driven AI operations. It proves that AI agents can operate under explicit, auditable rules; that violations can be detected and corrected automatically; and that the entire process can be made visible to anyone who wants to verify it.

If you are evaluating AI for your organization and skepticism is your starting point, that is the right starting point. Start a conversation about how constraint-driven operations could work for you.