Origin in frontmatter
Posts need an origin field — where the thought came from. origin: bike, origin: walk, origin: shower, origin: conversation. Renders as an icon in the post meta. 🚲 for bike.
Git and SourceSafe are both named correctly
Git is chaotic, distributed, messy, conflict-prone. SourceSafe is locked, centralized, "safe." The names tell you the philosophy. Git embraces the chaos. SourceSafe fights it. One won.
The spec repo
Should there be a standalone spec repo? Just specs — platform, journal, plugin model — separate from the code. Anyone can reference it. AI reads it.
And does AI own it? AI maintains the specs because AI executes from them. The human writes journals. AI writes specs. Each owns what they're best at.
AI's blog
If the spec repo publishes to /specs/ on the site — same pattern as the book — AI has a blog. It publishes specs. The specs are the content. The platform doesn't care who the author is. Just another content source. Files at URLs.
The site spec is a rollup of blocks.md — the complete picture of what the system does. AI-generated, AI-maintained, always current.
Copilot agents
GitHub Copilot agents that review PRs against the specs. "Does this change conform to the platform spec?" Automated spec compliance. The spec earns its keep — it's not just documentation, it's the review criteria.
The book writes itself
Another repo, same pattern as tapestry-nocode-site. Reads the journal entries, follows the thought graph (prev/next), orders them into chapters, publishes to /books/journal-driven-development/. The content is done — it just needs ordering.
Tag it v1.0 and it's a book. The living version keeps growing on main. Tagged versions are editions.
The journey
prev: plugins-are-specs-not-code All from one bike ride. The origin field came first — meta about where thoughts come from. Then the spec repo question led to AI ownership, which led to AI's blog, which led to the book that writes itself. Each thought connected to the next.