6.9
The Daily Loop
Why this exists
You now have every piece of the job separately. This lesson turns them into a routine that takes fifteen minutes and means nobody has to ask you whether things are alright.
It's the last lesson of the level because it's the shape of the job itself. Everything before this was a skill. This is the day.
The idea
Fifteen minutes, same order, every morning. The value is in it being boring.
A checklist you follow when you remember is not a checklist. The whole point is that it runs on the days you're busy, distracted, or half-asleep — which are precisely the days something will have gone wrong overnight.
A loop that works:
| Step | Minutes | What you're asking |
|---|---|---|
| 1 · Alerts | 2 | Did anything fire overnight? |
| 2 · The money number | 3 | Does yesterday look like a normal yesterday? |
| 3 · The jobs | 3 | Did the scheduled things run? |
| 4 · Errors | 4 | Anything new, or anything that's now frequent? |
| 5 · Say something | 3 | Post the status, even when it's fine |
Step 5 is the one that changes how people see you, and it's the one that feels most pointless. Posting "all clear, revenue looks normal, everything ran" every morning seems like noise — until the morning you post "sync didn't run, I'm on it", and everyone already trusts the source because they've been reading it for three weeks.
Silence is not information. If you only speak up when something's wrong, nobody knows whether quiet means healthy or means you didn't look.
Two things make the difference between a real loop and a nice idea.
Same order, every time. Not because order matters technically, but because a fixed sequence is what lets you notice something missing. Variation means you can't tell whether you skipped a step or it wasn't there.
Write down what "normal" is. Yesterday's number means nothing without a sense of the usual range. Note the range now, while things are calm, so at 7am you're comparing against something written down rather than a vague feeling about what looks right.
Then make it shorter. After two weeks you'll know which steps have never once found anything, and which found something twice. Cut the dead ones. A five-minute loop you run every day beats a twenty-minute one you run on Tuesdays.
And the thing to watch for over months: anything you find manually more than twice should become a check — that's 6.5. The loop should shrink as monitoring grows. If it isn't shrinking, you're doing by hand what a timer should do.
Do
- Write
daily-loop.mdwith your five steps in a fixed order and a time budget for each. - For each step, note what normal looks like — the usual range, the expected number, what "ran fine" means.
- Run it tomorrow morning. Actually run it, in order, at the start of the day.
- Time yourself. If it's over twenty minutes, it's too long to survive a busy week.
- Post the status even though nothing's wrong. That's the step you'll want to skip.
- Run it every day for a week, noting for each step whether it found anything.
- At the end of the week, cut every step that never found anything, and turn anything you found twice into a scheduled check.
- Write down what you cut and why.
Step 5 is the whole lesson. The habit only pays once it's established, and it only becomes established by being done on days it seems unnecessary.
Done when
That's Level 6. You can read a system you didn't build, diagnose it from a log, investigate without breaking it, ship a change and prove it, know before anyone tells you, decide what to escalate, know what not to touch, handle credentials properly, and run the day.
That's the job. Everything after this is leverage on top of it.
Never do
Never skip the loop because it's quiet. Quiet is the output, not the reason to stop. The morning you skip is drawn from the same hat as any other.
Never let it grow. Every step earns its place by finding something. A loop that only ever gets longer stops being run, and a loop that isn't run is worse than none, because everyone believes it's happening.