Current: browser
The editor, local draft storage and HTML/CSS/JavaScript preview run on the visitor’s device. The website supplies static files; exported files remain separate from the published record.
RESEARCH · ENGINEERING · EXPLANATION
Independent investigations into AI systems, their engineering and their wider implications.
My research programme brings together 39 investigations into AI capabilities, limitations and opportunities for development. Each project connects source-code exploration with scientific evaluation, architectural explanation and ethical and societal analysis. The records distinguish planned investigations from documented findings.
Loading project records…
The project record follows the investigation from its research question and historical context through implementation, evaluation and interpretation. It provides a place to document decisions, discoveries, unsuccessful approaches and supporting evidence as the work develops.
Terminology notes record the definitions and amendments relevant to the investigation. Infrastructure as Code (IaC), cloud deployment and communications integrations are considered where they affect the system under study.
Each stage needs its own evidence of completion. Findings can send the investigation back to its question, design or implementation. Milestone statuses record project progress; completion remains linked to the supporting artefacts and checks.
The editor supports small HTML, CSS and JavaScript demonstrations. The initial example displays the project’s planned fictional test case; it illustrates the interface rather than demonstrating a completed AI implementation or research finding.
No model is running. Model APIs are not connected; the preview policy restricts API and resource requests.
The preview is intended for trusted experimental code. The frame restricts API/resource requests, forms and parent-page access; it is not a hardened untrusted-code service, and resource-heavy scripts can affect browser performance. It cannot run the Python starters or hosted models. Exported HTML retains the same content restrictions.
The editor, local draft storage and HTML/CSS/JavaScript preview run on the visitor’s device. The website supplies static files; exported files remain separate from the published record.
The linked folder holds source references, a research dossier, roadmap, evidence register and Python starter. GitHub stores the files; the Python starter requires execution in a suitable Python environment.
Local models, Python services, cloud GPUs and commercial APIs need a selected runtime, dependencies, authentication and cost controls. No backend or installed-app integration is connected here.
Project-specific architecture and sequence diagrams will identify components, data and protocols, distinguishing documented behaviour, unknown internals, proposed extensions and measured boundaries. Cloud infrastructure, Infrastructure as Code and unified communications belong in a project only when they affect its design.
A release fixes a reviewable version of the work: report, permitted code/data, figures, run instructions, evidence records, attribution, limitations and change notes. The public case study explains it; a commit/tag or archival record identifies the exact artefacts. Publication, independent review and scientific validation are separate statuses.
Read the shared methodology and release guide · Zenodo versioning · Citation metadata · NIST secure-development framework
Research direction and methodology: Grace R. Nettey. AI assistance supported the initial organisation, interface and draft framework. Project analyses, implementations and findings are recorded as they develop. Added 6 November 2026. Implementation record →