Writing
Notes from the workbench.
Small, specific things that cost us real time. Mostly bugs that looked like something else, and instruments that agreed with us while measuring the wrong thing.
The test passed because the code never ran
A regression test asserted that nothing bad happened. It passed for weeks — on a path that never executed the thing it was guarding.
Your section total is printing as a page subtotal
A table footer repeats on every printed page. If it holds a grand total, every page now shows a different-looking wrong number.
The date input that renders as yesterday
new Date('2026-03-01') is UTC midnight. Format it in Chicago and you get February 28. The bug only appears west of Greenwich.
The empty spreadsheet cell that ate the next three columns
One alternation regex over sheet XML, one self-closing tag, and a row silently loses data — for some rows only.