For two years I tracked every working hour, tagged by project, and looked at the chart on Sunday evenings. The chart was always reassuring and almost never true. A forty-hour week and a twenty-hour week produced roughly the same amount of finished work, and the forty-hour one produced considerably more code I later deleted.
What the hours were actually measuring
Time tracking measures how long I sat down. It does not know the difference between the afternoon I spent choosing a database and the twenty minutes that afternoon in which I made the decision. Both are logged. Only one of them mattered.
Worse, the chart rewards the wrong instinct. On a Thursday with three logged hours, the honest move is often to stop and think. The chart says: open the editor.
One sentence on Friday
What I do now takes about four minutes a week. On Friday afternoon I write one sentence in a file, answering: what is true this week that was not true last week?
- “Halfmoon sends alerts to Slack, and two people are using it.”
- “I understand why the importer is slow, and it is not the database.”
- “Nothing. I read, and I fixed two bugs nobody reported.”
That last one is allowed, and writing it down is the whole exercise. Three of those in a row is a real signal — much realer than a chart showing thirty-eight diligent hours.
It changed what I start on Monday
The sentence is easier to write when the week contained one clear thing rather than five vague ones, so I have started arranging weeks that way. Not out of discipline. Out of not wanting to write “nothing” for a third week running.
The file is nineteen months old now. It is the only project record I have ever kept past the first month, and the only one I ever read back.