On this page
I wanted a personal website that explained the substance of my work without resembling a conventional CV or a collection of marketing claims. It needed to connect research, professional projects, architectural decisions and earlier experiences. Above all, it needed to remain practical to maintain. An impressive site that requires another full-time job to populate would be a poor solution.
That made the website both a product-design exercise and an opportunity to observe a different way of developing software. The result is a working WordPress implementation, built and reviewed through repeated natural-language collaboration, code changes and inspection of the actual running pages.
The problem was larger than a visual design
The visible requirement was straightforward: a clean, restrained site with readable articles, accessible navigation and a coherent personal identity. The underlying content problem was less obvious. Research programmes needed to connect to capability studies, journal entries, evidence and earlier projects without becoming a flat list of unrelated posts.
Five distinct editorial collections were therefore established: Professional Projects, Research Programmes, Capability Studies, Research Journal and Foundations. A shared research-topic taxonomy and explicit related-work metadata allow different kinds of material to connect without being duplicated.
The site needed to support a serious research publication as well as an individual professional identity. That requirement shaped the implementation as much as the typography or colours did.
An architecture with clear responsibilities
The website uses WordPress, with presentation responsibilities in a dedicated theme and editorial functionality in a first-party must-use Core plugin. The theme handles layout, appearance modes, navigation and article presentation. The Core defines content types, metadata, search, evidence relationships, SEO and controlled importing.
That separation is not merely tidy organisation. It means changes to an article’s appearance do not require rewriting the rules that govern what the article is, how it relates to other publications or whether it is ready for public release.
Research figure 01
Research publishing as one connected system
A simplified view of the website’s responsibilities
- EDITORIAL INTENTResearch and professional workArticles begin with real subject matter and evidence
- CONTENT AUTHORITYWordPress CoreCollections, metadata, topics, relationships and draft status
- PRESENTATIONThe website themeArticle layout, navigation, accessibility and appearance modes
- REVIEWABLE OUTPUTDrafts and supporting materialPublished evidence is distinguished from internal sources
- HUMAN APPROVALReview before publicationThe final release decision remains with the author
Illustrative architecture, not a complete implementation trace.
The platform was developed locally using Local. The design direction became known as Midnight / Precision / Signal: midnight-blue navigation, mostly light editorial surfaces, clear typography, restrained accents and a two-column reading layout with contextual navigation. Light, Dark, System and High Contrast appearances were carried through the same design system.
Genuine content exposed the real requirements
Early demonstration articles were useful for exercising layouts, archives and relationships. They were deliberately treated as test material, not as evidence of past professional work. Once the publishing architecture was stable, they were removed and replaced with authentic research drawn from determinIQ documentation and an existing October 2026 research paper.
That exposed practical questions which a visually convincing mock-up would not have answered. Could a long article remain readable? Would a sidebar help rather than compete with the text? How should supporting evidence be presented? Would a second import overwrite the author’s changes? Should a new study link back to a programme that was still only a draft?
The response was to extend the existing WordPress content model rather than build a parallel publishing application. The first-party research importer creates reviewable drafts, reuses source documents where appropriate, records topics and explicit relationships, and leaves previously imported articles alone on subsequent runs.
The intention was to remove repetitive assembly work without removing editorial control.
Working with real pages changed the design
A series of actual browser screenshots helped settle the article design. We tried more visually prominent treatments, then removed them when they interrupted reading. The approved layout uses a restrained header, a comfortable main reading column, a compact contextual sidebar and supporting evidence that can occupy a wider section beneath the article.
This was iterative rather than a single design accepted unchanged. Important corrections concerned proportion, contrast, placement, unnecessary controls and the way long-form text behaved on screen. The current visual coherence is the result of those decisions, not simply a visual theme applied at the beginning.
The same principle applied to diagrams. Separate generated images initially looked like a plausible route to consistent architectural illustrations. In practice, the diagrams were mostly sequences of labelled panels. Native HTML and CSS reproduced them more simply, with selectable text and automatic adaptation to the website’s appearance modes. That decision is examined separately in a connected research-journal entry.
What can be demonstrated
By 14 November 2026 the supplied implementation included the dedicated theme, Core editorial plugin, authentic research draft collections, contextual sidebar, evidence display, native diagrams and individual SEO metadata. The source and the browser screenshots demonstrate those components, while the development record also includes several package and installation corrections.
A Lighthouse run on the local flagship research page reported scores of 100 for Performance, 100 for Accessibility, 78 for Best Practices and 100 for SEO. The same screenshot showed local HTTP and insecure-request findings contributing to the Best Practices result. Those scores describe one automated local browser test, not an independently audited production deployment or full accessibility certification.
The tests are useful nevertheless. They show that the site’s relatively substantial publishing functionality has not automatically required heavy front-end infrastructure. They also provide a concrete baseline against which later changes can be checked.
Where development went wrong
Not every update arrived in the right form. At one point the must-use Core files were incomplete in the installation location, and WordPress correctly reported that dependent editorial functionality was unavailable. The remedy was to deliver and verify a complete package with an unambiguous directory structure.
The original graphical-diagram pipeline was another useful error of judgement. It solved the drawing problem, but introduced a production process that the site’s own browser technologies could handle. Recognising that was more important than defending the work already done.
These are modest examples, but they matter because they show the role of acceptance and correction. A successful development process is not one in which every initial suggestion is right. It is one that can identify an unnecessary or faulty approach and arrive at a better result.
What the project establishes — and what it does not
The website is a working implementation of the agreed editorial architecture in a local WordPress environment. It demonstrates the integration of content structure, accessible presentation, research relationships, supporting evidence and a repeatable draft-production process.
The project was developed in a concentrated period that I found strikingly short compared with how I would previously have approached comparable work. That is a personal observation, not a measured productivity comparison. Nor does a clean Lighthouse score establish security, production readiness or conformance with every accessibility requirement.
Further work remains: publication review, verified historical material for Foundations, real-world link checking, a fuller manual accessibility assessment and production deployment checks. The important achievement at this stage is that the platform is substantial enough for real content to reveal what still needs improving.
Why this belongs in my research record
The website is not a commercial demonstration of determinIQ. It is a different piece of software with a different purpose, built using related principles of explicit responsibility, evidence and human-led acceptance.
That is what makes it useful as a case study. A personal publishing system may appear straightforward beside an analytics engine, but both require the same care in deciding what belongs to the system, what should be automated and what the human remains responsible for.
The most useful outcome is not the number of pages produced. It is a research website capable of growing from the work it documents, without becoming an obstacle to that work.
Project context
Research programme: From intention to implementation: human-directed AI-assisted engineering
Topic: AI-assisted engineering
Explore further
Further reading is listed below.