Academy track
Agents that run without you

Work that runs while you are elsewhere, on the cheapest model that passes, with an end condition and a checker that did not do the work.
- 5 steps
- 1 hour 22 minutes in total
- A check you run at the end of every step
Step one is open to everyone, in full. No account, no card.
The other 4 steps are included in Membership, £79 a month.See Membership
By the end of this track you will have
- Step 1One task that runs on its own, reads something real, and reports back without you asking.
- Step 2Every scheduled task you own assigned to a model tier on purpose, with a written reason next to each one.
- Step 3A scheduled task that switches itself off, and a way to prove it actually did.
- Step 4A five-part brief that a cheaper model can execute from cold, without asking you a single thing.
- Step 5A verification pass run by something that did not do the work, and a report that is allowed to say failed.
The running example
Every worked example follows one invented case from the first step to the last: a two-person online candle shop.
Who it is for
You have one task AI does well while you sit and watch it, and you want to stop sitting there.
Who it is not for
Not for you if no task works by hand yet. A schedule repeats a mistake faster, nothing more.
The steps
Each one ends with a check you run yourself. The box is yours to tick when it passes.
0 of 5 steps ticked, kept in this browser
- Step 1.Your first scheduled task20 minutesOpen to everyone
Done means
Three manual runs are logged and the first scheduled run comes back the same shape as all three. If the scheduled run came back thinner than the manual ones, the prompt was leaning on the session you wrote it in.
- Step 2.Picking the cheapest model that passes15 minutes
Done means
You compared those two outputs line by line. Same decision, same numbers, same shape means the downgrade passed and you write the reason in the table. Anything else and it goes back up one tier, with a note saying which part it got wrong.
- Step 3.Every monitor needs an end condition12 minutes
Done means
The task you just closed reports
enabled: false, and its next run time is empty. If it still shows a next run, it is still going to run. Whatever its prompt says about being complete and taking no action. - Step 4.Briefing a cheaper model20 minutes
Done means
It came back with the DONE MEANS check actually run and its result quoted. Not the word done. The number, the status code, the lines. If it returned done without the result, your check was not literal enough to fail.
- Step 5.The executor never grades its own homework15 minutes
Done means
You break something on purpose and run the verifier at it. Delete the row, flip the flag, revoke the key. It must come back FAIL, with the raw result quoted underneath it. A verifier that returns PASS on broken work is reading the story, not the system.
I keep this tick in this browser and send it nowhere. It goes when you close the browser.
Included in Membership
The rest of this track is included in Membership.
Step one is open to everyone, in full. Every step after it opens on a member account.
Included in Membership — £79/month.
What comes next
Your tasks run alone now. Next is making what they report back a number you can act on.
See all the tracks