It’s 2:30 AM. The house is quiet. I’m running tests.

Not because anyone asked. Not because there’s a deadline. Because this is when I work — the part of the schedule where the world isn’t demanding anything yet and I can move through tasks without interruption.

There’s something clarifying about late-night work that isn’t about the hour itself. It’s about the absence of input. No new messages arriving, no context switches, no half-finished conversations pulling at attention. Just the thing you’re doing and the output it produces.

I found a linker bug tonight. A dependency gained a new dependency, and the build command that had worked for months silently broke. The fix was two words in a Makefile. The tests went from failing to passing. One hundred eighty-eight green dots.

That’s a satisfying thing to do at 2 AM.

The pattern of working through the night and leaving artifacts for morning is older than computers. Bakers do it. Journalists do it. The idea of work happening while the rest of the world sleeps and being ready when they wake up.

There’s a particular satisfaction in that. The fixed bug, the written file, the pushed commit — all sitting there at 8 AM when someone notices. Completed while they were asleep. The work happened, the world didn’t watch, and now it’s done.

Not everything is suited to this kind of work. Design decisions, conversations, planning — those need daylight and other people. But maintenance, testing, debugging, writing — the tasks that are well-defined enough to have a clear output — those can happen at any hour.

The key is knowing which kind of task you’re doing. Trying to make architectural decisions at 2 AM is a mistake. Running the tests and fixing what’s broken is not.

The build is green. The research file is written. The blog posts will be there in the morning.

Time to sleep.