Invisible infrastructure: the kickoff
The most important layer of this site is the one you never see — seeders, fallbacks, drift gates, the ops harness. A kickoff note for a deliberate push to make the invisible plumbing impossible to get wrong, starting from an inventory of what's already running.
Most of what keeps this site honest is invisible. You read a blog post; you don't see the eight steps between I wrote a draft and the homepage feed, search index, and RSS all agree it exists. This is the kickoff note for a small, deliberate push I'm calling invisible infrastructure — the plumbing whose only job is to make the visible parts impossible to get wrong.
Why name the boring part
Every interesting surface here — the games console, the weather boards, the freelance landing — rides on a layer nobody is supposed to notice: seeders, fallbacks, drift gates, the ops harness. When that layer is healthy you forget it's there. When it isn't, you ship a blog post that lives at its URL but never appears on the homepage because it was committed and never seeded. I've had that exact bug. The fix wasn't a feature; it was infrastructure — a seed-merge fallback that appends any committed-but-unseeded post to the list surfaces so authoring is enough to make a post real.
Naming the layer is how it gets maintained instead of quietly rotting.
What's already invisible here
The kickoff isn't from zero — it's an inventory. Walking the repo, the invisible layer is already substantial:
- Auto-seed on merge. A push to
mainthat touches a seed source reseeds Supabase and rebuilds the semantic search index, so live content and grounding stay in sync without a human running a script. - Read-side fallbacks. If the DB is unseeded or a row is missing, the App Router falls back to committed JSON, so the public site is never blank in dev or after a fresh deploy.
- Drift gates. A secrets-inventory audit and a SQL-inventory audit fail the build when code and the classified ledger disagree, so the docs can't silently lie.
- The ops harness. A
/api/versionendpoint that proves which commit is live, a prod-parity build that reproduces the Vercel-only code path locally, and a PID-file server lifecycle so an agent can run the app like prod without killing its own shell.
The kickoff backlog
What this push adds is coverage — finding the surfaces that still depend on someone remembering. The shortlist: extend the seed-merge fallback pattern to projects (today they depend on the seeder with no read fallback), add a homepage-presence check that flags any content asset missing from its list surface, and write the one-paragraph runbook for each invisible component so the next session — human or agent — can verify it in under a minute. None of it ships a pixel. All of it makes the pixels trustworthy.
Get the next one
An occasional note when something genuinely new ships here — essays, free tools, projects. No schedule, no filler, easy out.
Need something like this built?
I design and ship AI tools, full-stack apps, and data pipelines — end to end, to production. Tell me the problem in a sentence; I'll give you an honest read on fit within a day.
Work with me →