Audit

The ability to prove who was accountable for an AI-made decision is quietly disappearing by default.

The evidence your auditor needs doesn't exist yet. We generate it.

Digital Sovereignty is the infrastructure for independent operator capability assessment of AI systems. Not a legal certification. Evidence that an assessment was conducted.

461 days until EU AI Act takes effect - December 2, 2027

The Compliance Loop

Audit

30 questions. 8 domains. An independent operator assessment. Cryptographic seal at auditproof.pro.

Infrastructure

Digital Sovereignty installs the hardware, local AI models, and Doctrine Engine that enable human oversight.

Assessment

Proof of Operator seals the result. Verifiable by any third party at auditproof.pro/verify.

Every Digital Sovereignty engagement concludes with a Proof of Operator assessment. One does not exist without the other.

View the public ledger →

Sovereign Infrastructure

The Stack

WHAT YOU ACTUALLY GET

Your stack runs itself.

Four local AI models watch every product every 30 minutes. Payment gates. Links. Performance. Anomalies caught before you notice them.

It knows how you work.

Doctrine Engine records every decision you make. After 30 days it knows your portfolio better than you do. After 12 months it's irreplaceable - that's value, not lock-in.

Legally protected. Regulation ready.

Seven utility models filed. For high-risk AI systems covered by these provisions, the current application date is December 2, 2027. Proof of Operator produces machine-verifiable evidence of operator capability assessment.

9

months operational

7

utility models

8

live products

€0

external funding

1

mission

Brainiac Ltd · London · thomas@brainiaclimited.com

THE MODEL STACK

Four models. One sovereign stack. Zero cloud.

Every Digital Sovereignty deployment runs four purpose-selected local AI models. Each is optimized for a specific operational role. Together they form a complete autonomous monitoring and intelligence layer that never sends data outside your infrastructure.

Gemma 3 12B

VISION & MONITORING

Google's multimodal model handles visual monitoring. It screenshots and analyzes product pages, detecting layout breaks, missing content, and visual anomalies that text-only models miss.

8.1 GB · 12B parameters · Q4 quantization

Phi-4 mini 3.8B

ANOMALY DETECTION

Microsoft's compact reasoning model runs rapid pattern analysis on logs, metrics, and system events. Its small footprint means it runs alongside other models without straining hardware.

2.5 GB · 3.8B parameters · Q4 quantization

Qwen 3.5 4B

CODE & LOGIC

Alibaba's code-focused model handles technical integrity checks. It verifies HTML structure, catches broken links, and performs logical consistency checks across monitored systems.

3.4 GB · 4B parameters · Q4 quantization

Llama 3.1 8B

CONTENT & LOCALIZATION

Meta's versatile model generates reports, briefings, and localized content. It handles Czech and English fluently, making it ideal for morning briefs and operational summaries.

4.9 GB · 8B parameters · Q4_K_M quantization · 34.7 t/s prompt · 8.4 t/s generation

Total stack: 18.9 GB in 32 GB RAM - runs fully parallel with headroom. Zero API costs. Zero data leaving your infrastructure.

SEE IT IN ACTION

Pick a task. See which model handles it. Compare the output.

PRIMARY MODEL

Gemma 3 12B

VISION & MONITORING
PASS12B parameters · Q4 · 5-7 t/s

SAMPLE OUTPUT

Screenshot analyzed. Homepage layout stable. CTA button detected and functional. Pricing section rendered correctly. No visual anomalies detected. Verdict: OPERATIONAL.

FALLBACK MODEL

Llama 3.1 8B

CONTENT & LOCALIZATION
PROMPTEDSame task, different model.

WHY NOT PRIMARY

Llama 3.1 cannot process images. Would require OCR pre-processing step. Gemma 3 12B handles vision natively.

All output generated locally. Zero API calls. Zero data leaves your hardware.

6

Products per install (capacity)

1

Local AI models running in parallel

€0

Ongoing software cost after installation

~1 wks

Client breakeven vs junior dev hire

Operational

Models Gemma 3 / Phi-4 mini / Qwen 3.5 · IP filed June 2026

THE SOLUTION

Four models. One doctrine. Always learning.

MONITOR

GEMMA 3 12B

"Watches everything"

  • ·Pings all product URLs every 30 minutes
  • ·Detects 404s, broken routes, slow load times
  • ·Flags anomalies to the Executor layer
  • ·Lightweight - stays in memory 24/7
COST TO RUN: €0

EXECUTE

PHI-4 MINI 8B *

"Fixes what breaks"

  • ·Classifies every flagged issue automatically
  • ·Executes known fixes without human input
  • ·Drafts fix prompts for novel issues
  • ·Parses emails, receipts, documents, contracts
COST TO RUN: €0

* Current default. Model-agnostic role - can be substituted with Mistral Nemo, Llama 3.1 8B, or equivalent on request.

REASON

QWEN 3.5 8B

"Improves what runs"

  • ·Weekly upgrade analysis per product
  • ·Generates ready-to-paste improvement prompts
  • ·Handles complex scheduling and multi-step logic
  • ·Drafts copy, sequences, proposals
COST TO RUN: €0

LEARN

DOCTRINE ENGINE

"Remembers everything"

  • ·Watches every correction you make via WhatsApp
  • ·Writes amendments to skill files automatically - no manual editing
  • ·Maintains doctrine.md: every rule the stack has ever learned, dated and versioned
  • ·Weekly synthesis: what broke, what improved, what should change
  • ·Monthly: doctrine.md is read by Claude before the strategic audit
COST TO RUN: €0

Intelligence compounds: automatically

Default stack: Gemma 3 / Phi-4 mini / Qwen 3.5. All roles are model-agnostic. Alternative local models installed on request at no extra cost.

  • 01Every 30 min → Gemma 3 pings all product URLs. Flags anomalies to Phi-4 mini.
  • 02On alert → Phi-4 mini diagnoses, classifies, executes or drafts. Logs outcome.
  • 03On correction → You override via WhatsApp. Qwen 3.5 writes the amendment to the skill file automatically.
  • 04Every Sunday → Qwen 3.5 reads all corrections from the week. Proposes 3 doctrine updates. You approve or skip. Applied immediately.
  • 05Every month → Claude audits full stack. Sends strategic brief. Reads doctrine.md first.
  • 06Always → doctrine.md grows. The stack knows more this month than last month. Always.
  • 07Cost to run → €0 in software licences after installation. Intelligence compounds at no extra cost.
  • 08On every WhatsApp correction → doctrine.md updated. Skill file amended. Stack gets smarter.

THE MODEL LAYER

The stack runs four local models by default. Each model is assigned a specific role based on its architecture strengths. None of the roles are locked to a specific model.

GEMMA 3 12B handles monitoring. Small, fast, always in memory. Ideal for the lightweight 24/7 ping cycle. No architecture released to date does this job more efficiently at this size.

PHI-4 MINI 8B handles execution. Classification, tool use, fix generation. Strong at instruction-following and structured output. The local LLM space is moving fast - this role is reviewed on every new client installation. If a stronger model has dropped since your spec call, you get it.

QWEN 3.5 8B handles reasoning. Weekly analysis, upgrade generation, multi-step logic, drafting. One of the strongest reasoning models available at this parameter count. Consistently outperforms alternatives in structured output tasks.

The Doctrine Engine is model-independent. It runs on whatever model handles the REASON role. As that model improves, the Doctrine Engine improves with it - no reinstallation required.

> MODEL LAYER: REVIEWED ON EVERY INSTALLATION. YOU ALWAYS GET THE STRONGEST AVAILABLE STACK AT TIME OF SETUP.

The software stack - OpenClaw, Ollama, all four models - costs €0. What gets paid for is the expertise to wire it correctly, tune it to your specific portfolio, and activate the Doctrine Engine that makes it improve itself from day one. That is the product.

OPEN SOURCE

Transparency from the ground up

Presence.js, our open-source browser intelligence framework. No cookies. No backend. Under 15 ms.

Read every line of code that touches your data. The full framework is MIT licensed and available on GitHub.

View on GitHub

MIT license. Zero dependencies.

THE NUMBERS

Run the calculation.

5

Hours above are counted per product.

8
€60
Current annual maintenance cost€124,800
Digital Sovereignty installation cost€2,900 - €4,700
Breakeven~2 weeks
Year 1 net saving€122,300
Year 3 saving (compounding)€371,900
ROI on installation4892%

These numbers assume Digital Sovereignty replaces the equivalent ops time. In practice, it also catches issues that would never be caught manually - broken payment gates, degraded conversion paths, silent 404s. Those savings are not in the calculation.

AFTER YOU BUY

WHAT HAPPENS NEXT

From first email to operational stack.

01

YOU EMAIL US

Day 0

thomas@brainiaclimited.com. Tell us how many products you maintain and which tier interests you. Response within 24 hours.

EMAIL NOW →
02

PRODUCT MAPPING CALL

Day 1-3

We review your product list, infrastructure, and current ops load. We tell you exactly what the stack will monitor, what it will automate, and what it will flag.

03

HARDWARE CONFIRMATION

Day 3-7

You confirm hardware spec. We confirm compatibility. Doctrine template sent for your review.

04

INSTALLATION

Day 7-14

Remote installation. Three models deployed. Doctrine configured. WhatsApp channel live. First anomaly report within 48 hours.

05

OPERATIONAL

Day 14

Stack is live. You receive your first morning brief. The operator is no longer alone.

THE DOCTRINE ENGINE

The system that learns from you. Without being told.

Every correction you make via WhatsApp is written as a one-line amendment to the relevant skill file. Every Sunday the system synthesises the week's corrections and proposes three doctrine updates, which you approve or skip. After one year, doctrine.md is the most valuable document in your business - the accumulated operational intelligence of your entire portfolio.

DIGITAL SOVEREIGNTY INTEL

Weekly signal. The state of autonomous AI ops.

One email per week. What's changing in local LLMs. What's breaking in production AI stacks. What's worth your attention. No noise. No pitch. Written at FT-adjacent standard.

Every issue includes one excerpt from doctrine.md - one rule the stack learned this month, and the correction that generated it. Subscribers see the intelligence layer compounding in real time. No other publication shows you an AI system learning from its own mistakes in plain language.

FREE TIER

  • Weekly Stack Intel digest (summary only)
  • Public case study access
  • Model benchmark updates

PRO TIER

€6/MONTHCOMING SOON
  • Full weekly report with implementation notes
  • Upgrade prompt library (weekly additions)
  • New skill files and config templates (early access)
  • Model benchmark detail + recommendations
  • Direct access to Tom via reply
SUBSCRIBE PRO - €6/MO →

Also in the intelligence network: Ghost OS Intel Drop (privacy/preparedness) · Intel (autonomous ops)

QUESTIONS

Frequently asked.

Hardware by Scale

Five tiers. From solo operator to institutional deployment. Every component is the one we test and deploy in production.

Why this pays for itself: 45-minute install, €0 ongoing software cost, ~7-week breakeven vs. a junior ops hire.

TIER 1

DIGITAL SOVEREIGNTY GUIDE

€350 - €580 ONE-TIME

Price varies by portfolio size. 1-5 products: €350. 6-20 products: €580.

Complete self-install documentation including the Doctrine Engine - the self-improving layer that makes the system more accurate every week. Config templates for all four model roles. Skill files configured specifically for your portfolio. Per-product knowledge silo template. Correction Loop automation pre-configured. Everything needed for a technically capable operator to install a stack that gets smarter without being told to.

For: Technical founders who want to do it themselves.

OpenClaw config templatesOllama model setup guidePhi-4 mini skill files (5 pipelines) - tuned during setup, not genericPer-product knowledge silo template (one folder structure per product)Correction Loop setup - WhatsApp override to skill file amendment, semi-automatedDoctrine Engine setup - doctrine.md + weekly synthesis automationSkill files tuned during setupSystem Intelligence Report template - monthly self-audit, ready to runPDF founding document
TIER 2

DONE-FOR-YOU INSTALLATION

€2,900 - €4,700 ONE-TIME

Brainiac Ltd installs and configures the full stack for your specific product portfolio. Per-product spec files written. All five pipelines deployed and tested. Doctrine Engine active from day one (activates as a manual process that becomes semi-automated over the first 30 days). Guaranteed working on delivery day. Model selection is discussed at the spec call - the default stack (Gemma 3, Phi-4 mini, Qwen 3.5) is the current recommended combination, with alternatives available based on your portfolio's specific requirements.

For: Founders who want it done correctly the first time.

Full install on your hardwarePer-product spec files (up to 10 products)All 5 pipelines configured + Doctrine Engine active from day one (Doctrine Engine activates as a manual process that becomes semi-automated over the first 30 days)WhatsApp alert system configuredCorrection Loop active - your overrides improve the stack from day onePer-product knowledge silos created and seeded with initial patternsdoctrine.md initialised with your standing orders and known preferences30-day support via WhatsApp - including monitoring first doctrine cycleInstallation case study (with your permission)Model selection reviewed at spec call - you receive the strongest available local stack at time of installation, not a fixed snapshot

Your stack will know more on day 30 than it did on day one. Without you doing anything except using it.

2 INSTALL SLOTS OPEN THIS MONTH
TIER 3

MANAGED OPS RETAINER

€580 - €930/MONTH

Dashboard monitoring (Coming soon). Monthly full-stack audits (Coming soon). Weekly Doctrine Synthesis review (Coming soon). Monthly System Intelligence Report (Coming soon). You never think about product ops again.

For: Established founders with revenue who want zero maintenance overhead.

Dashboard monitoring (Coming soon)Weekly Doctrine Synthesis review - proposed amendments reviewed before they apply (Coming soon)Monthly System Intelligence Report - full pipeline accuracy stats, correction log, doctrine growth (Coming soon)Monthly strategic audit - reads doctrine.md before the audit beginsWeekly upgrade prompts (per product, generated by Qwen 3.5)Incident response (same business day)Monthly report via WhatsApp
RETAINER ONLY

RETAINER EXCLUSIVE: SYSTEM INTELLIGENCE REPORT

Delivered monthly. No other tier receives this.

EMAILS PROCESSED THIS MONTH: [n]

RECEIPTS EXTRACTED: [n] ([accuracy]% accuracy)

SKILL FILE AMENDMENTS APPLIED: [n]

MOST CORRECTED PIPELINE: [pipeline name]

MOST RELIABLE PIPELINE: [pipeline name]

doctrine.md GROWTH: [n] new rules this month

RECOMMENDED FOR NEXT MONTH:

[three specific amendments proposed by Qwen 3.5]

This is not a report about what happened. It is a report about what the stack learned.

3 RETAINER SLOTS REMAINING
TIER 4

AGENCY WHITE-LABEL (COMING SOON)

€1,750 SETUP + €350/MONTH

(Coming soon) Full system licensed under your agency brand. You deploy it for your clients and add autonomous AI ops to every retainer you sell. We support your implementation. You keep the margin.

For: Digital agencies with 10+ product clients.

There is one advantage to being an early agency licensee that cannot be bought later.

Every install you do feeds the master doctrine. Not your clients' private data - the structural patterns. The failure modes. The pipeline edge cases. The email formats that break parsers. The calendar conflicts that scheduling logic misses.

Each install makes every future install better. By install 10, your skill files contain failure mode intelligence that a competitor starting today cannot replicate. By install 50, the gap is not closable.

The doctrine compounds across your client network. That is not a feature. That is a moat.

White-label documentationAgency-branded config templatesImplementation supportMonthly model update briefingsLicense to resell as your own ops layerDoctrine meta-layer access - anonymised pattern data from all installs improves your client deploymentsPriority access to new skill files as they are developed
TIER 5

DIGITAL SOVEREIGNTY BUNDLE

+€3,500 ON ANY BRAINIAC BUILD

Every Brainiac Ltd build client receives the option to add this system on delivery day. Your product runs itself from the moment it's handed over. This is how 20 products stay maintained by one person.

For: Clients of Brainiac Ltd's build service at aidentity.pro.

This tier is bundled with a full Brainiac Ltd build engagement. Email thomas@brainiaclimited.com to discuss.

First installations in progress. Case studies published on completion.

Get assessed at auditproof.pro

30 questions. 8 domains. A cryptographic seal. An independent operator assessment.

GET ASSESSED →

Want the full Operational Sovereignty course outline?

10 modules. 40 lessons. The exact method, revenue model, and distribution strategy for the independent operator. Enter your email to get the full outline.

Digital Sovereignty provides independent operator capability assessments and infrastructure. The Proof of Operator seal constitutes evidence that an assessment was conducted by Brainiac Ltd. It is not a legal certification of compliance with EU Regulation 2024/1689 or any other regulation. Brainiac Ltd is not liable for regulatory decisions based on the seal. Companies should consult their own legal counsel.

DIGITAL SOVEREIGNTY

Autonomous AI ops · Brainiac Ltd · London · Est. 2026

Built and operated by Brainiac Ltd.

Intelligence amplification. Not automation.

thomas@brainiaclimited.com

aidentity.pro

Brainiac Ltd · 167-169 Great Portland Street, London, W1W 5PF

Protected by seven filed Czech utility models

Aligned with EU AI Act Articles 14 and 26 - effective December 2, 2027

View the line →

No social media. No cookies. No tracking. The stack monitors itself.

Digital Sovereignty is a technical installation and configuration service. Results depend on client hardware specifications and product portfolio complexity. Ollama and all referenced open-source models are independent products not affiliated with Brainiac Ltd. Brainiac Ltd accepts no liability for outcomes arising from third-party model behaviour.

Digital Sovereignty Bundle from €9,900 · Installation from €2,900 · Both delivered by Brainiac Ltd, London

© 2026 Brainiac Ltd · London

Terms of collaborationGhost OSAuditProof.pro

Runs on Fedora Atomic / Red Hat ecosystem