Governance
I spent a year cataloguing security threats. Now I'm building the governance layer — where I draw the line on what technical capabilities should never be allowed, and who gets to decide. These documents are open. They're incomplete. They're meant to be challenged.
Initial drafts are synthesized by the author with AI assistance. Final versions require human collaborators and reviewers to identify gaps, errors, and cognitive biases. See the Transparency Statement for a full audit trail.
View source on GitHub →Table of Contents
Core Governance Documents
Transparency Statement
EditAuditable record of Human-AI collaboration in QIF Framework development
Transparency Statement: Human-AI Collaboration in QIF Framework
This document is auto-generated from the QIF Derivation Log, which is the single source of truth for all framework decisions and AI collaboration records.
Last generated: 2026-06-20 at 00:10:07
Table of Contents
AI Collaboration Summary
| Metric | Value |
|---|---|
| Total derivation entries | 111 |
| Entries involving AI | 77 (69%) |
| Corrections / retractions | 🔴 5 |
| Quorum swarm reviews | 🟢 5 |
| Human-only entries | 0 |
| Human decision rate | 🟢 100% (Kevin Qi is Accountable on every entry) |
AI Contribution Breakdown
| Level | Count | Description |
|---|---|---|
| AI-assisted | 0 | Human directed; AI implemented or researched |
| AI-generated | 0 | AI produced first draft; human reviewed and approved |
| Human-only | 0 | No AI involvement |
| Quorum-reviewed | 0 | Multi-agent swarm validation |
| Unspecified | 111 | Legacy entries without AI Contribution Level metadata |
Core Principle
Every contribution is categorized by its cognitive origin. AI assistance is treated as a tool subject to human oversight, not a collaborator with independent judgment on ethical or novel technical matters. All final decisions are made by Kevin Qi (A = Accountable in every RACI entry).
Tools Used
| Tool | Role |
|---|---|
| Claude Opus 4.6 | Primary AI assistant — co-derivation, implementation, literature synthesis |
| Quorum (Claude plugin) | Multi-agent swarm review — structured dissent, fact-checking, security audit |
Documented Corrections
The derivation log contains explicit correction entries (Classification: CORRECTION) where previous claims were retracted or revised. This is by design — corrections are the most valuable entries because they demonstrate active oversight.
Total corrections: 5
Generated from 111 derivation log entries. Source: QIF-DERIVATION-LOG.md
Generated by Quorum governance pipeline