Build #0: Outcome

LogicBoat is a working static-first technology publication and build lab, deployed at logicboat.app. This internal project is the first documented build: a foundation for publishing practical articles and searching them without operating a runtime backend.

Architecture

Next.js App Router generates the site, TypeScript checks the code, and Tailwind CSS styles the reading interface. Fumadocs MDX/Core compiles and loads content; Zod validates each collection's frontmatter. Pagefind generates a browser-search index from the exported HTML. GitHub stores the source, and Cloudflare Pages builds and deploys it when changes reach main.

The repository locks Next.js 16.3.8, Fumadocs Core 16.16.1, Fumadocs MDX 15.4.6, and Pagefind 1.5.2. These versions describe this foundation, not a promise that future versions behave identically.

Content Model and Publishing Pipeline

Five collections serve different purposes: Builds document projects; Learn explains concepts; Tools introduces practical software; Notes captures short findings; Blueprints holds implementation plans when they are ready to publish.

The publishing path is:

MDX → frontmatter validation and compilation → static routes
→ Next.js export → Pagefind postbuild → Cloudflare Pages

Each collection feeds its section index, detail routes, and sitemap automatically. npm run build -- --webpack exports the site to out/; npm postbuild runs pagefind --site out, creating out/pagefind/. Cloudflare deploys the whole artifact.

Content in published collection directories is public. Builds draft status labels a page; it does not hide it. Unfinished working material belongs outside those directories.

What Was Verified

As of October 6, 2026, the recorded foundation checks include:

  • ESLint passes; the production build passes TypeScript checking and static export.
  • Cloudflare Pages local serving accepts extensionless article URLs and returns a real HTTP 404 for missing routes.
  • Production route checks confirm working detail pages, metadata/canonicals, and missing-route behavior.
  • Pagefind production runtime checks return relevant results for static, n8n, and appointment against the then-published content. A nonsense query returns zero results; normalized result links return HTTP 200.

These checks establish the publishing and search foundation. They do not claim that every article or an external automation has been tested.

Constraints and Known Limits

The publication has no database, authentication, traditional CMS, or runtime backend. Search runs in the browser against generated static assets.

Browser/device and accessibility testing is not comprehensive. Interactive search testing, Safari, Firefox, physical mobile devices, and dedicated screen-reader checks remain outstanding. Fumadocs webpack cache-analysis warnings have been non-blocking in recorded builds.

Lessons

Static-first fits content that can be rebuilt when it changes. It avoided account management, database operations, and a server-backed publishing interface while keeping the content pipeline inspectable in Git. The static-first foundations article explains the tradeoffs; the constraint Note captures the editorial reasoning.

A publication can document tools without running them as part of its platform. The n8n guide is relevant to future automation work; n8n is not a dependency of this site's foundation.