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
Audit ledger
Verify any seal at auditproof.pro/verify30 questions. 8 domains. An independent operator assessment. Cryptographic seal at auditproof.pro.
Digital Sovereignty installs the hardware, local AI models, and Doctrine Engine that enable human oversight.
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
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
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.
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
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
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
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
PRIMARY MODEL
Gemma 3 12B
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
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.
Products per install (capacity)
Local AI models running in parallel
Ongoing software cost after installation
Client breakeven vs junior dev hire
Models Gemma 3 / Phi-4 mini / Qwen 3.5 · IP filed June 2026
THE SOLUTION
MONITOR
GEMMA 3 12B
"Watches everything"
EXECUTE
PHI-4 MINI 8B *
"Fixes what breaks"
* 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"
LEARN
DOCTRINE ENGINE
"Remembers everything"
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.
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
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 GitHubMIT license. Zero dependencies.
THE NUMBERS
Hours above are counted per product.
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
From first email to operational stack.
YOU EMAIL US
Day 0thomas@brainiaclimited.com. Tell us how many products you maintain and which tier interests you. Response within 24 hours.
EMAIL NOW →PRODUCT MAPPING CALL
Day 1-3We 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.
HARDWARE CONFIRMATION
Day 3-7You confirm hardware spec. We confirm compatibility. Doctrine template sent for your review.
INSTALLATION
Day 7-14Remote installation. Three models deployed. Doctrine configured. WhatsApp channel live. First anomaly report within 48 hours.
OPERATIONAL
Day 14Stack is live. You receive your first morning brief. The operator is no longer alone.
THE DOCTRINE ENGINE
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
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
PRO TIER
€6/MONTHCOMING SOONAlso in the intelligence network: Ghost OS Intel Drop (privacy/preparedness) · Intel (autonomous ops)
QUESTIONS
Hardware by Scale
Why this pays for itself: 45-minute install, €0 ongoing software cost, ~7-week breakeven vs. a junior ops hire.
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.
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.
Your stack will know more on day 30 than it did on day one. Without you doing anything except using it.
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.
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.
(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.
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.
30 questions. 8 domains. A cryptographic seal. An independent operator assessment.
GET ASSESSED →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.