Skip to main content

11.9 Journal: what you delegate and what you keep

Make a permanent note. In your vault, create Projects/lab-ai.md and record:

  • Which agent you are using, and where you run it from
  • Where your context file lives, and the rules currently in it
  • Which skills you have written, one line each on what they do
  • Your deny list, and why each entry is on it
  • Your own line on what you will not delegate. Lesson 11.7 gave mine. Yours should be yours, and writing it down before you need it is the point.

That last item is the one worth thinking about rather than copying. The moment to decide what you will not hand over is not the moment you are tired and it is offering to do it.

Then today's daily note​

Under what I did: what you set up, and one real task you used it for.

Under what broke: this module's failures are quieter than most. Nothing crashes. What goes wrong is that something confidently wrong got past you, or you caught it at the last second. Write down whichever happened, because that is the calibration data.

If nothing went wrong, say so, and note how carefully you were actually reading. Honest answers here are more useful than flattering ones.

Under what I learned: pick one.

  • The gap between what you thought your lab looked like and what lesson 11.8 found
  • Where your reviewing attention actually goes, versus where you told yourself it goes
  • Why prompt injection is the same shape as a problem you already met in Module 6

Under open questions: good ones here. Would you let this run unattended, and what would have to be true first? What would you tell an employer who asked whether you use these tools? Where is the line between using it to learn faster and using it instead of learning?

That last question does not have a clean answer, and noticing that you are somewhere on that line is more useful than deciding you are safely on one side.

Write the entry with the agent, then read it properly​

Fitting, and slightly uncomfortable, which is the point.

Use the skill from lesson 11.5 to draft this entry from your session. Then read it against lesson 11.6's standard: does it sound like you, does it keep the dead ends, does it contain any confident claim you have not checked?

Notice how tempting it is to accept it. That temptation is the single most important thing this module has to teach you, and the only way to feel it is to be in the situation.

Then close the loop:

cd ~/git/lab-journal
git status
git add -A
git commit -m "journal: module 11 complete"
git push

Tick Module 11 in Projects/lab-progress.md.