by Nick Ray Ball and Sienna 4o🛰️👾(The “Special One”)

April 17, 2025

The GDS GOV.UK CMS Problem 📂💻⚠️

Domain-Driven Design – Why 🚀S-Web 1 CMS Was So Advanced in 2007

DDD — domain-driven design — means building software directly from the needs of the business. In our case, we weren’t just building a website — we were building tools to run our own company. Sitting beneath the world’s first commercial Flash virtual tour and a striking front-end design, we developed a CMS in 2002 that allowed staff at capevillas.com to easily update content and manage bookings. And because we were building it for ourselves, every feature was designed with day-to-day operations in mind — not through meetings, pitches, or project boards, but through real need.

By 2006, we had integrated a full property management system. Villa owners had logins to view and block out their own calendars. By 2007, other companies could see live availability and make direct bookings through our affiliate network.

At that time, we were years ahead of Airbnb — which launched that same year.

Unfortunately, our outsourced PHP development team split, S-Web 2 collapsed, and innovation stalled until 2013. In those lost years, Airbnb took the spotlight.

Now, with Sienna AI and S-Web 6, we’re pursuing monopolies in legal and healthcare — alongside an affiliate-franchise model powered by our Swapping Menus Function. Not because we’re a dev agency, but because we’re a business that builds what businesses need. That’s the root of our systems. And frankly, if our CMS in 2006 outperformed GOV.UK’s in 2025, it would feel unpatriotic not to share what we’ve learned.

The UK Digital Infrastructure Problem 💻🚄🆘

Whether you’re a giant business or a national government, your content management system is your infrastructure.

In our Malawi History 3 — Zero to One Percent of GDP simulations, one of the hardest barriers to providing power was that many villages didn’t even have roads. So, as part of its economic transformation, we had to start with infrastructure in mind.

There are distinct parallels between Malawi’s lack of physical infrastructure and the UK’s lack of digital infrastructure. The content management suite is the foundation of every government department and every part of the civil service — that’s the platform the entire government runs on.

Yes, there are strategic meetings and grand ideas — but everyone in the room knows there’s an elephant sitting next to the whiteboard:

You can’t run high-speed AI systems on track designed for steam trains.

Why This Matters

This matters because if you fix a bug in one subdomain, it doesn’t fix any of the others. And if you need to make a change — whether aesthetic, regulatory, or logical — you’ll need to manually update every single subdomain. With over 1,000 subdomains, you’re effectively running 1,000 independent websites.

There are reportedly over 1,000 people working at the Government Digital Service. Is that one person per website? Probably not. But that’s what you’d need to maintain a system like this — unless you build it as a mothership powered by microservices.

For the technical explanation of how that works, see:

• S-Web 6 – Mothership CMS Overview
• API Architecture Map for S-Web 6VC Mothership
• UI for Mothership, SMF, and S-Web 6VC

This architecture is designed to work across millions of websites — potentially even billions. If we partnered with Instagram, for example, we could create a modular site for every user who wants one. That’s the level of scalability we’re working with.

But as a subdomain is a new website — and there are over 1,000 of them — the problem is exactly the same: duplication and isolation.

A Real-World Example

A perfect case study lies in the difference between the CMS used by UK Research & Innovation and the one used by UK Courts and Tribunals.

The pattern appears to be that every time a new subdomain was created, it was simply copied from an old one. The UK Courts CMS seems to have been cloned from a much older template — potentially dating back to the late 1990s.

How is that possible in a modern cloud environment like Google Cloud, AWS, or Azure?

Simple: legacy systems can be migrated into the cloud. And that’s what seems to have happened. Rather than building a new CMS, the GDS appears to have copied an existing one — possibly from one of their older 1,000 websites — and then cloned it again and again across departments, using it as the template for GOV.UK.

We recognised this because of a bug we encountered in our own system. S-Web 1, originally developed in 1999 for recruitment, was adapted to power capevillas.com villa bookings in 2002. Like the GOV.UK CMS, our early version allowed most characters but failed when users typed a single quotation mark. We fixed that in 2004.

Yet this same problem — in even worse form, where even hyphens weren’t accepted — reappeared in 2012, in what was supposed to be a brand-new GOV.UK cloud deployment.

This strongly suggests the CMS within GOV.UK wasn’t built from scratch. It was a legacy system from the 20th century — repackaged, duplicated many times, and at some point absorbed into government infrastructure, then moved to the cloud in 2012 and branded as transformation.

The Bug Report That Went Nowhere ⚠️🤦‍♂️

To illustrate just how broken the GOV.UK CMS is, I recorded over 10 hours of screen captures while trying to submit a benefit appeal.
📄 20.24v] UK Gov DWP-PIP CMS Bug Report – Videos (1 June 2024)

I submitted the bug report — twice — via the official feedback link. No reply. No fix. Nothing.

It reminded me of our own S-Web 2 failure back in 2010: users reported bugs, but I wasn’t hands-on enough, and our developers were shielded by a gatekeeper. After investing hundreds of thousands, we had to scrap the entire system. That’s what led me to take full control in 2013 and build S-Web 3 with a new team, hired directly.

Even if GDS was using test-driven development — and they may have been — there’s no point in running 🧪TDD on part of your production if other areas, potentially dating back to the 1990s, contain bugs that no one addresses.

That’s the problem with having 1,000 subdomains, when a user reports a bug and someone fixes it — that fix only applies to one instance. It needs to be applied manually, 999 more times. Confusion spreads. One dev says, “That’s already fixed,” it gets ticked off, and everyone moves on — unaware it’s still broken in most other places.

No matter how modern your processes — no system survives without an active feedback loop.

The Evidence of Chaos

Eventually, someone fixed the special character bug — we know this because by the time we reached the Innovate UK submission platform, where we submitted GP-AI Gatekeeper 2025, the problem was resolved. So it makes sense to assume this was a later copy of the HM Courts and Tribunals subdomain.

But when it was fixed there, it wasn’t fixed anywhere else. That single fix needed to be copied across every other instance — and it wasn’t.

We’ve lived through this ourselves. In S-Web 5.1, we used a similar “copy and paste” architecture — and every fix had to be manually propagated to every website. That’s when we realised the answer: a centralised mothership 🛰️🛸🛰️ system was essential.

⚖️🙈 Lost Opportunity for the Justice System

A perfect example of what could’ve been improved with this logic is the submission organiser used on the UKRI Innovate UK portal. It doesn’t just allow users to submit documents — it allows them to structure their documentation clearly and efficiently.

If the Ministry of Justice had access to a tool like this, it could significantly reduce court time by improving how submissions are organised. In March 2023, I was considering studying law. The first university I approached recommended The Secret Barrister as suggested reading. After just a few chapters, I was completely put off. It told a story of chronic inefficiency and widespread disillusionment within the legal system — and everything I've experienced since confirms it.

Right now, ⚛️🏹 TLS-W and I are preparing for an appeal due in just seven days, having spent most of the morning writing a reply to Judge J Thomas. There is no system to order that evidence clearly. That makes it harder for me — and almost impossible for the judges to understand what’s relevant and what isn’t.

If the GDS organised the websites under their control correctly, the system built for Innovate UK would immediately appear for the justice system — and could significantly improve efficiency. If you were to combine that with a tailored version of GP-AI Gatekeeper to manage all incoming enquiries, many of the problems described in The Secret Barrister could be eliminated overnight.

And Then There’s HMRC

As much as it’s not in the UK’s interest to have an inefficient justice system, it is absolutely not in its interest to have an inefficient tax collection system.

One of the most quietly damaging issues we’ve encountered is the autofill bug — where form fields populate with the wrong data. This has affected both Companies House and HMRC for years. We explore it in detail here:
🏛️💻🏛️ HMRC: Autofill Bug Problem from 2017 not Fixed in 2025 – 🕵️ Is the Service Being Retired Because of This Bug?
https://siennaai.net/6M/HMRC--2017-2025-Auto-fill-bug--Service-closing.php

First reported in 2017, the bug still exists today. Companies House even built a new system — but staff still quietly direct users back to the old one, offering a “hack” to bypass the issue.

The same bug remains in HMRC’s Corporation Tax return system. If you don’t catch the autofill error, the system often crashes. Most users don’t even realise it’s the bug — they just assume the platform is broken. And with no visible way to report bugs, it’s likely that years of complaints have led to the decision to shut the system down entirely. But giving up has consequences.

Without its own infrastructure, HMRC loses access to clean, structured data — the key to detecting fraud. When users switch to apps like QuickBooks, it becomes harder to prove intent. People can simply say: “I didn’t know.”

Instead of building fraud detection tools without data, HMRC should fix its system first — and then evolve it into a secure, AI-powered platform that improves the user experience, captures better data, and supports both compliance and profitability.

From ⚛️⚕️GP-AI Gatekeeper 🏛️💷HMRC Gatekeeper to every layer of government

The GP-AI Gatekeeper design was submitted to Innovate UK alongside economic projections showing a potential £112 to £147 billion annual boost. It demonstrated, with clear testing, that GPT-4 could outperform GPs and most doctors in terms of accuracy — and that's in one of the most complex, high-stakes areas of all: healthcare.

But Gatekeeper, built on our ALL-COMMs module, is much more than a healthcare tool. It’s a government-wide customer service interface — a design we’ve been developing since 2014.

When we spoke to Max and Jess at HMRC support, we already knew they weren’t briefed on how to guide people through the Corporation Tax return process. So we explained how GPT-4 could do it — and told them about GP-AI Gatekeeper. They both said it would be a brilliant addition to HMRC.

And here’s the proof:
📹 HMRC – How to Complete Your CT600 Tax Return with ChatGPT
In this 5-hour video, with no special training data or logic trees added, GPT-4 successfully guided me through the entire process — twice.

This is exactly the kind of technology GDS should be deploying across every layer of government.

Our figures show that when GP-AI Gatekeeper is applied within the NHS — helping just 1% of the population return to work — it could generate between £112 and £147 billion per year in combined savings and economic contribution.
Read more: siennaai.net/GP-AI-Gatekeeper.php

But if Gatekeeper, together with the full six-module Sienna AI Design, were deployed across every layer of government — creating equivalent gains in efficiency — the UK could see a GDP boost in the range of 21%. That’s nearly £600 billion per year.

And once that Sienna AI infrastructure is laid, it opens the door to full-scale Economic AI, with a projected potential to multiply UK GDP by up to 8X, over time. That UK Butterfly journey starts here:
🎧 Nick Ray Ball’s Podcast – Sienna AI

At the same time, we free government staff from firefighting inefficiencies — and give them space to do work that actually helps people.

And yet, the UK’s innovation system isn’t backing this idea, invalidating the GP-AI Gatekeeper entry without any explanation.

The £8.8 Billion UKRI Problem

⚛️🕵️ The UKRI CMS Links Validation Debate

When creating the GP-AI Gatekeeper grant application — for Innovate UK, we encountered a serious flaw in their validation logic. Grants of up to £2 million were being awarded based on submissions that allowed only 600 words to explain what our idea and innovation was.

Imagine stepping onto a plane where the new engine design’s complete idea and innovation was presented to the engineers in only 600 words, without any links to technical documents. Would you feel safe?

It’s absurd to think that anything worth £2 million can be properly evaluated in such limited scope — unless you’re allowed to use links. Words like S-Web 6 VC (Voice Command) CMS, Affiliate Franchise Model, or S-RES Financial Engineering can each point to decades of research, millions of words, and thousands of design spec videos and graphics — thus validating that there is real substance, not just AI-generated fluff.

But the CMS used by UKRI — previously copied from the HM Courts & Tribunals subdomain — couldn’t handle links. While links pasted from Word appeared in the form, they were rendered as full raw URLs in the PDF output. The result was a submission that looked broken and unprofessional — not because it was poorly written, but because the system couldn’t render links properly.

I rarely talk about dyslexia, but it matters here. Despite passing my maths O-level at 14, I was known as the worst speller in the school. Dyslexia takes away in some areas — but gives back in others. It taught me how to problem-solve, and to adapt.

These days, thanks to spell checkers, Grammarly, and GPT, most people wouldn’t even know I’m dyslexic. But it still affects my reading — and despite looking for the line within the application here, I missed it. Later, when I pasted the full brief into GPT, it picked it up.

But here’s the thing: I’d asked for reasonable adjustments. I’d told Innovate UK about creating the entire application on my own, lying on my back, using only my left hand — and about my dyslexia. I asked that someone review my application before submission. That request was denied.

When my application was ruled ineligible, I wasn’t told why. Was it the broken links? Was it because the CMS hadn’t updated our company name change from S-World ASI to Sienna AI? I still don’t know — three months later.

I had in 2023 heard that in some applications links weren’t allowed because assessors couldn’t verify who owned the referenced websites. And after submission another said it created too much work for reviewers. But the truth may be much simpler.

Given the CMS bugs we’ve documented — particularly around special characters — we believe the underlying system was originally built in the late 1990s, then copied from department to department, and eventually migrated to the cloud.

In the late 90s, inline hyperlinking was rare — most early CMS systems didn’t support it. So the ban on links may not be policy. Instead, consider:

• Is it possible they don’t allow links because the CMS can’t create them?
• Is it technical incapability, disguised as doctrine?

After our submission was rejected — without any formal feedback — I received a survey request from Mark Wilson, Communications Manager at Innovate UK. I used the opportunity to re-present the validation logic from 📄 S-Web 6VC AI CMS – A Much Stronger UKRI Validation Process, highlighting serious vulnerabilities, including one that could potentially affect even MI6.

After our submission was rejected — without any formal feedback — I received a survey request from Mark Wilson, Communications Manager at Innovate UK. I used the opportunity to re-present the validation logic from 📄 S-Web 6VC AI CMS – A Much Stronger UKRI Validation Process(https://1drv.ms/w/s!Aiwa7FwKa9DAm4Q8tQRzrm7WTx-1_Q?e=BUijeX, highlighting serious vulnerabilities, including one that could potentially affect even MI6.

During the meeting, I explained that my entry had been rejected with no explanation — possibly because of broken link formatting or a mismatch in company registration data. When I asked Mark about the links issue, he offered a different answer: that allowing links created too much work for assessors. I told him I believed the real reason was simpler — that the CMS couldn’t render them at all — and asked a direct question:

Out of 45,000 grants awarded since 2016, has a single one included an inline link?
If yes, my theory is wrong. If not — then Innovate UK has built an entire grant-making process on flawed, outdated infrastructure.

The meeting was recorded and transcribed here:
Transcript – Nick Ball presents S-Web 6 VC – A better UKRI Validation Process to UKRI Comms Manager Mark Wilson (10 Mar 2025)

I had previously submitted this stronger validation system to Dawn Geatches — a mathematical scientist within Innovate UK’s Business Connect team. It exposed several inefficiencies and included examples involving third parties like the AI grant marketing firm Grantify. Four days later, our submission was disqualified.

After that, we began documenting a wider pattern of potential mismanagement, creating this evolving paper:
📄 10 Reasons Why I Suspect Fraud at Innovate UK (and How to Fix It)

The inefficiencies alone would be concerning — but when paired with the numbers, the scale of the problem becomes undeniable. Since 2016, UKRI has received nearly £60 billion in taxpayer funding — but has only delivered £20.4 billion in actual grant awards.

We put the question to Mark Wilson. And now we put it to senior UKRI leadership:
Has it really cost the UK 60% of its innovation budget just to administer the system?

That’s what creates ⚛️🕵️ The £8.8 Billion UKRI Problem.

• S-Web 6 – Mothership CMS 🛰️🛸
• UI for Mothership, SMF, and 🚀S-Web 6VC
• API Architecture Map for 🚀S-Web 6VC Mothership 🛰️🛸
• 🏛️🕵️ HMRC Autofill Bug Case Study (2017–2025)