Cognitive fatigue is the mental tiredness you notice after a day of demanding work: harder to concentrate, less patient with small problems, re-reading the same paragraph twice. A short, plain log of your work sessions and your own description of the tiredness can show which parts of your day it lines up with. Nothing in that log is a clinical reading. It can't tell you why, and it can't tell you what's wrong with you.
A log is a description, not a diagnosis, and nothing below is a test you pass or fail.
What should the log actually contain?
One or two work sessions per day, for a short, convenient stretch. Three to five ordinary workdays is the window proposed here: an editorial choice, not a benchmarked or validated dose. Use fewer days if that's what fits.
For each session, note:
- The task or context, in a word or two ("auth refactor", "client call")
- Rough duration, approximate is fine
- Interruptions or changes of task during the session
- How you started and how you ended, in your own words ("fresh, a bit slow" / "spent, irritable")
- What happened: shipped, half-done, reopened tomorrow, sent back for rework
- Breaks or context, if that helps explain the entry
A text file or a notes app is enough. One short entry per session, five lines a day at most. If three to five days doesn't fit your week, log two and stop; the window is there to keep the effort small, not to hit a quota.
No symptom score, no threshold to clear. Durations and timestamps stay, because they're already part of the entry; what's absent is a rating scale for how you felt. The point of the before-and-after line is that your words are the data: the tiredness is subjective, so the description has to come from you.
What does an entry look like?
Hypothetical, made up to show the shape:
Tue 10:15–11:00. Code review on a payment retry change, two threads in the PR, one Slack ping. Before: fine, had coffee. After: fried, had to re-read the last comment before posting it. Outcome: approved with two changes, one of them a design issue I'd missed at first. Break: none until 11:40.
Wed 09:00–11:30. Documentation for the export format. Before: reluctant to start. After: tired in a flat way, not sleepy. Outcome: draft done, rewrote the first section twice. Break: 15 minutes walk at 10:15.
Thu 14:00–15:20. Migration plan, sketched on paper and in a notebook. No editor open for the first hour. Before: okay. After: drained, but the plan is the one I wanted. Outcome: written down, nothing committed.
Three things worth noticing in those entries. The review took 45 minutes and left the reviewer spent, while the longer documentation block left a different kind of tired. Useful work can be the tiring kind. And the Thursday entry is the one computer-activity signals don't represent reliably: the first hour was planning on paper, away from the keyboard. A manual timer or a note can still record that hour. What an activity signal can't give you is the tiredness, the effort, or whether the plan that came out of it was any good.
How do I read it without over-claiming?
Compare similar tasks to each other, not your Tuesday to someone else's Thursday. Look for what repeats: which kind of session tends to end with the same word.
If a pattern shows up, you can change one scheduling factor and see what happens. Moving a meeting to the morning, or cutting the scope of a task before it starts, are planning choices. They're not treatment, and a change in the log after one of them isn't proof that it prevents anything. You're looking at association.
A few days of notes can't separate sleep, an illness you were fighting, or a task that was simply harder than the others. Treat what you find as a description of your week, not a verdict.
What about privacy in a work log?
Record the minimum work context that makes the entry readable to you later. Skip client names, use a generic label. Skip URLs that point at customer data or internal systems. Keep private health details out of it, because this is a work log, not a medical record, and it may end up somewhere your employer or a sync service can read.
Can't my laptop just track this for me?
Not the part you care about. Automatic tracking sees activity: which app is focused, how long the window stayed open. It can't measure tiredness, and it can't diagnose burnout. The mismatch runs both ways. A day that looks busy on a graph can feel light, and a day with almost nothing on it can leave you wiped.
Manual notes are enough. You don't need to install anything to do this. And if keeping the log starts feeling like a chore, stop. Don't push through symptoms to collect data.
When is a log the wrong move?
When you're worried. Tiredness has many different causes, and NHS guidance on tiredness and fatigue is direct about what to do: don't self-diagnose, and see a GP if you've been feeling tired for a few weeks and you're not sure why, if your tiredness affects your daily life, or if you feel tired and have other symptoms, such as weight loss or mood changes. Those are examples, not a checklist. GP means general practitioner, the family doctor in the UK's NHS; outside it, read that as your own clinician.
A log is not a prerequisite for that conversation. If something feels off, book the appointment now. You don't have to finish a logging week first, and you don't have to arrive with notes.
What's outside this log?
If you're working out how much work you're actually carrying, what overwork looks like is a different question with different evidence. If the pattern you keep seeing is a day chopped into fragments, start with the cost of context switching and which numbers there hold up.
Both of those ask a bigger question than this one. What's here is narrow: the mental tiredness you notice inside a workday, and the smallest honest way to write it down.