Building this site with a governance toolkit
How this portfolio is built: a static Eleventy site, authored through a partner-and-governor workflow with an AI implementation agent.
Why static?
This portfolio is built with Eleventy, a static site generator that produces lightweight HTML and CSS. There's no database, no server-side logic, and no need to manage a CMS. The entire site deploys as static files via SSH, which means a smaller attack surface, faster load times, and zero operational overhead.
A governance workflow
The architecture here reflects how I think about building software with AI: governance by humans, implementation by machines. A human governor owns requirements and architectural decisions; a partner LLM decomposes the work and reviews it; and an implementation agent (Claude Code) writes the code from committed specifications on its own feature branch, never touching the mainline.
Each feature is a complete, verified artifact before any merge. The result is that every change is reproducible, reviewable, and audit-friendly—without sacrificing the speed that AI implementation can bring.
The stack
Eleventy 3.x with Nunjucks templates, self-hosted fonts (Fraunces + IBM Plex Mono, both SIL OFL), and a small vanilla JavaScript module for the projects page that fetches from the GitHub API at runtime. No build framework beyond Eleventy itself and a terser minification step for the one JS file. Deployment is a straightforward scp to a shared Apache host — chosen deliberately to keep the infrastructure legible and the dependency surface small.
Governance artefacts — decisions, debt, risks, prompts, specs — live in a separate repository alongside the site code. The two are kept separate on purpose: the site repo is clean, deployable code; the governance repo is the audit trail. Every architectural decision has a registered ID, a rationale, and a date. That separation is what makes it possible to hand implementation to an AI agent without losing accountability.
Why it matters
In a world where AI is increasingly involved in code creation, governance matters more than ever. This project demonstrates one model: clear specs, committed prompts, branch isolation, and human sign-off at every gate. The portfolio site itself is the proof—it's real, it's live, and every decision is documented.