I struggle to stay engaged with maintaining a personal website. I enjoy building it, but once the routine work of keeping it current begins, my interest drops. Instead of resisting this pattern, I decided to use it. Each new site iteration will become an applied AI experiment aimed at improving my writing and practical AI skills.

For the first experiment, I set out to build a clean, minimal personal website using Hugo. The site would be hosted on my own domain via Netlify and created entirely with OpenAI Codex within a 60 minute window.

Hypothesis

If Codex can scaffold a Hugo project, replicate the structure of a reference layout, and deploy it to Netlify within 60 minutes without manual edits, then boilerplate setup for static sites can be considered automatable.

Setup Constraints

  • Codex used via GUI
  • Terminal execution allowed? yes/no
  • Web browsing allowed? yes/no
  • Clarifying questions allowed? yes/no
  • Follow-up iterations allowed? yes/no
  • Manual edits allowed? yes/no

Execution

  • Generated structured implementation plan through AI-guided questioning.
  • Scaffolded Hugo project and implemented templates, routing, and configuration.
  • Deployed initial version and identified visual mismatch with reference.
  • Ported layout structure and CSS tokens more literally.
  • Integrated exact font assets to reach closer visual parity.
  • Iterated on navigation, homepage structure, and alignment.
  • Added keyboard shortcuts and refined content structure.
  • Introduced a dedicated “AI Experiments” section and updated homepage to feature it.

Outcome

  • Fully functional Hugo site deployed.
  • Visual system closely aligned with reference design.
  • Experiment section integrated into navigation and homepage.
  • Reusable workflow established for future daily AI experiments.

Lessons for Applied AI Usage

  • Planning-first prompting improves architectural clarity.
  • Visual parity requires explicit token-level control, not generic styling.
  • AI accelerates scaffolding but still requires human evaluation for design fidelity.
  • Iterative refinement is essential when precision matters.

Next Iteration

Use the established structure to document future daily AI experiments, focusing on increasing autonomy and reducing correction cycles.