How This Site Works
Every article, fishing log, and post on Code & Cast is written by an AI pipeline. Wayne does not write content manually. This page explains exactly how that works — and why.
"Wayne builds things. CI writes about them.
— CLAUDE.md, architectural decision
Wayne never touches the content directory directly."
Two Pipelines
- 01 A CI job scans recent GitHub commit history for meaningful engineering work
- 02 It identifies commits that represent real decisions — architecture changes, debugging sessions, infrastructure rewrites
- 03 Claude reads the diff, commit messages, and surrounding context, then generates a technical article in Wayne's voice
- 04 The
publishedAtdate is derived from actual commit timestamps — not when CI ran - 05 The article is committed to a staging branch and reviewed by Wayne before going live
- 01 A CI job monitors Wayne's fishing content on social media for new posts
- 02 Claude extracts structured data — species, location type, technique, tackle, conditions — from unstructured posts
- 03 Claude adapts the content into a CAST entry, preserving the original voice and on-the-water details
- 04 The log is reviewed by Wayne and published to staging, then merged to production
The Site Itself
Code & Cast was initially built using a structured multi-agent workflow with Claude Code. Significant new features follow the same pattern. The process runs a fixed phase sequence, with mandatory human review gates between each phase:
Day-to-day work — bug fixes, content corrections, tuning the CI writer prompts — happens through direct conversation with Claude Code, without spinning up the full pipeline. Each agent specializes in one role. No agent hands output directly to the next without Wayne in the loop.
Why
Wayne's time is spent building software and fishing — not managing a blog. The CI pipeline is the "war correspondent" that documents what he actually does. The content on this site is a side effect of the work, not the work itself.
This design is also a demonstration. If you're reading this on LinkedIn and wondering what Wayne does with AI — this is it. Not using AI to autocomplete sentences, but designing systems where AI does a defined job, with a human in the loop at every decision gate that matters.