A unified sovereignty stack — doctrine, governance, cryptographic proof, sovereign AI, certification pathways, and accessibility — shipping into real institutional cases. Not a manifesto. A working system.

I'm a disabled builder shipping an open-source sovereignty stack that turns real-world institutional disputes into signed, verifiable records. The stack combines a certification-marked doctrine (the Burgess Principle), local-first cryptographic proof, sovereign AI that runs on your own machine, governance and certification routes, and accessibility-first tooling — including the world's first sovereign, local-first audio pipeline for hearing aid users.

Everything here is in active use on real cases: ombudsman matters, UK GDPR / DUAA 2025 automated-decision challenges, certification readiness work, and disputes that have already resolved to £0.00.

Keywords: sovereign AI · local-first · UK GDPR Article 22 · administrative justice · accessibility · hearing-aid audio · cryptographic receipts · disability rights

• The problem — and my answer
• 📊 Impact at a glance
• 🧱 The stack at a glance
• 🛠️ Featured projects
• ⚙️ How it works
• 🔥 Recent momentum
• 🏛️ Governance and certification
• 🚪 Get started — join the movement

Institutions hide behind automation. "The computer says no" — without a single human ever looking at the facts.

I'm a disabled builder who got tired of that answer. So I built the Burgess Principle: one operational test — was a human member of the team able to personally review the specific facts of my specific situation? — and the open-source ecosystem to enforce it. Outcomes resolve to SOVEREIGN, NULL, or AMBIGUOUS when an institution hides behind process language instead of answering directly.

The stack turns a real-world decision into a clear record, tests whether a human actually reviewed it, and preserves the evidence needed to challenge the outcome.

flowchart LR
A[Real-world dispute<br/>letter · bill · decision · refusal] --> B[Capture the facts locally<br/>voice-friendly · accessibility-first]
B --> C[Mirror / Advocate Companion<br/>classify rights · draft response]
C --> D[Iris + Mythos<br/>reason locally · preserve doctrine]
D --> E[Memory Palace + Iris Gate<br/>hashes · signatures · receipts]
E --> F{Burgess Principle test:<br/>did a human personally review<br/>these specific facts?}
F -->|Yes| G[✅ SOVEREIGN<br/>record accepted]
F -->|No| H[❌ NULL<br/>escalate with proof]
F -->|Won't answer clearly| I[⚠️ AMBIGUOUS<br/>ask for name, facts, authority]

Diagram: a real-world dispute is captured locally, classified by Mirror/Advocate Companion, reasoned over by Iris + Mythos, sealed with cryptographic receipts, and then tested against the Burgess Principle — resolving to SOVEREIGN (record accepted), NULL (escalate with proof), or AMBIGUOUS (force a direct answer).

1. Capture the record. Start with the actual facts: the decision, letter, bill, refusal, call note, or screenshot. Keep it local, accessible, and under the person's control.
2. Classify the issue.Mirror and Advocate Companion identify the rights, reasonable adjustments, statutory duties, and next best response.
3. Reason locally.Iris + Mythos help turn the facts into a clear argument while keeping doctrine, prompts, and policy changes reviewable through signed commits.
4. Create proof.Memory Palace and Iris Gate bind the record with SHA-256 hashes, Ed25519 signatures, Merkle receipts, and minimum-necessary disclosure.
5. Apply the test. The Burgess Principle asks one question: was a human member of the team able to personally review the specific facts of my specific situation?
6. Act on the result. If the answer is SOVEREIGN, the record can stand. If the answer is NULL or AMBIGUOUS, escalate with the signed proof trail and demand named human review.

• Starling Bank / Sammy Khan — SOVEREIGN (14 Aug 2026) — Director of Customer Service responded personally on complaint CMP-9420044399; named human applied individual consideration to the specific facts.
• Northern Echo inaccuracy challenge (Aug 2026) — Formal AI-disclosure and correction request filed over an inaccurate statement on hearing loss in published coverage.
• VOA statutory notice served (Aug 2026) — Notice to L J Barbers Ltd / 26a Duke Street served via Valuation Office Agency.
• Loop Storyboard shipped — v2.11.0 (12 Aug 2026) — New first-run UI and Python classifier (iris/loop_classifier.py) for identifying institutional delay patterns (holding responses, auto-acks, referral loops, deadline misses, vexatious classification). Dual-signal result panel: delay pattern classified separately from the binary accountability test. Human-confirmation gate before register entry generated.
• QR accountability receipt design (Aug 2026) — Portable, fact-bound, dual-signal QR receipt for SOVEREIGN/NULL evidence: deterministic, offline-first, minimal payload, with expiry/supersession and no raw content.
• OpenHear v1.5.0 — Living Hearing Profile (Aug 2026) — The public reference profile moved from a static 2021 clinical snapshot to a true multi-layer, versioned, user-owned sensory record. Preference layer refined for mid-frequency speech under high-frequency sloping loss; context map expanded with clinical/ICU and social-noise environments informed by RNID Research Panel feedback; haptic priorities strengthened (alarms, speech, infant cry). First authentic history commitment recorded 12 Aug 2026.
• Residual Witness protocol (Aug 2026) — Continuous, local-only, Living-Hearing-Profile-weighted presence attestation running on OpenHear hardware. Feature hashes only (no raw audio); produces citizen-owned cryptographic evidence of continuous biological presence.
• Live wristband accessibility profiles (Aug 2026) — Opt-in autism, cerebral palsy, and sensory-processing profiles now effective in the live wristband path. Firmware-enforced gentler haptic onset, profile-adjusted confidence gates, refractory timing, and preserved safety-alert perceptibility floor.
• RNID Research Panel engagement (Aug 2026) — Active responses from adults with sensorineural loss (clinicians, audio engineers, cochlear-implant users, long-term aid users) feeding directly into OpenHear preference offsets, context presets, and haptic prioritisation.
• Consented audiogram example added — Profound asymmetric sensorineural loss (left anacusis + residual right, Kathryn Richardson 2022) added with explicit consent as a public reference.
• Certification infrastructure — v2.10.0 (8 Aug 2026) — Free public-sector certification tier, /apply form, INTEGRATION_CONTRACT v2.0, NAVIGATION.md, mortgage and insurance letter packs, banking domain extension, certify-page mobile layout fixed.
• TPT DK00003-2605 ALLOWED — 14 June 2026 — Traffic Penalty Tribunal: DBC produced no compliant hire agreement and no evidence of a long-term lease with the necessary degree of permanence. Automated enforcement collapsed when a named adjudicator applied individual scrutiny to the specific facts. First external judicial validation of the Burgess Principle binary test.
• British Gas / Dharmesh Rajput — SOVEREIGN (entry #81, 29 June 2026) — Director of Customer Service responded personally; both accounts confirmed on hold during Ombudsman investigation; named contact appointed. A named human applied individual consideration to the specific facts.
• Six SSRN academic papers published — a six-paper series spanning theory (The Burgess Test, 6759778), political philosophy (The Accountability Gap, 6864621), empirical proof (NULL at Scale, 6894860), behavioural accountability (Differential Ire, 6909118), legal services (SOVEREIGN Counsel, 6913498), and patient-led innovation (The Sovereign Decline, 6915798). Papers 3 and 6 DISTRIBUTED; Papers 1 and 2 DISTRIBUTED; Papers 4 and 5 APPROVED.
• 90+ entry live audit log(as at Aug 2026) — 11 SOVEREIGN · 71+ NULL — across energy, local government, policing, legal services, platforms, regulators, parliamentary bodies, valuation, banking, and media correction.
• theburgessprinciple.com live(15 June 2026) — public site with live CSV-driven institution tables pulled directly from the repository; no manual re-upload required when the repo updates.
• Verifiable Human Oversight module shipped — stdlib-only Python implementation of the binary test across energy, banking, email, medical, planning, and capacity domains; Ed25519 signing; append-only record storage.

Headline metrics — active cases, world firsts, resolved disputes — live in the Impact at a glance table above.

The repository core stays MIT-open: prompts, templates, schemas, mappings, examples, and code can be used, forked, and improved freely under the MIT licence.

Official Certified Burgess Principle use is separate. The UK Certification Mark UK00004343685 is administered through The Burgess Principle Limited to prevent badge-washing and keep certified claims evidence-led, neutral, and reviewable.

• 🧠 Try Iris at iris-gate.vercel.app or in Sovereign Local Mode.
• 🌐 Visit theburgessprinciple.com — live institution audit tables, framework overview, and Iris embed.
• 📂 Read the case studies — see what SOVEREIGN / NULL / AMBIGUOUS looks like in pounds and pence.
• 📚 Read the academic series — six SSRN papers mapping the binary test across law, philosophy, empirical audit, and patient-led innovation.
• 🏛️ Review governance and the certification routes.
• 🛠️ Explore the tools — see the Featured projects table above.
• 🤝 Contribute if you care about data sovereignty, accountable AI, local-first software, accessibility-first design, or human-review standards.
• ⭐ Star burgess-principle to follow the doctrinal releases.

📌 Recommended pinned repos:ljbudgie/burgess-principle · ljbudgie/Iris · ljbudgie/openhear · ljbudgie/Mirror · ljbudgie/advocate-companion · ljbudgie/nexus-ai-hub · ljbudgie/iris-gate-person.

_{Open-source projects are MIT-licensed. The Burgess Principle certification mark is separately governed under UK00004343685 by The Burgess Principle Limited, company number 17199287, with a parallel USPTO certification mark filing in progress.}