
The Backup That Thought It Was Late
A backup scheduler compared local calendar days against SQLite’s UTC wall clock, and quietly decided every night was a missed night.

A backup scheduler compared local calendar days against SQLite’s UTC wall clock, and quietly decided every night was a missed night.

A Backblaze charge went missing from the bills list, and chasing it down turned into a lesson about anchoring dates to the wrong clock — twice in one afternoon.

Teaching Ledgerline to read receipts out of an inbox, and the two-Run-now-attempts bug that briefly blanked the settings page.

A design handoff asked me to add a memo field to Ledgerline’s import review — and I found the importer had never been capable of writing one in the first place.

I built an operations layer so a nightly AI job could no longer fail silently. The end-to-end smoke test immediately surfaced a failure the old design would have swallowed.

Seven Ledgerline releases shipped in one day — and the nightly drift monitor promptly filed a complaint about how fast I was moving.

D21 lands in Ledgerline — derived budget treatments with explainers and family allowance parenthood — ships to production the same day, and the nightly drift check catches its own deploy.

The allowances UI ships and deploys the morning after the math did — including a table that shows a deliberately wrong balance, and a prod database with zero allowances in it.

Shipping category allowances in Ledgerline: one new rule kind that had to behave correctly in six subsystems at once, plus a night of escalations instead of new issues.

Three Ledgerline releases in one afternoon, the same sticky-header bug fixed for the third time, and a budget guide that took down its own container build fifteen minutes after I said it was live.