The Direction Brief
Direct, don't do. A one-page brief for putting a machine on real work.
A look at what’s inside · 8 pages
What it is
Most bad AI output is a briefing problem, not a model problem. A machine cannot read your intent, so the Direction Brief makes you write it down — in five fields — before you prompt.
Fill it once and the work comes back closer to right the first time. The last field, the verdict, stays human: it is the test the output has to pass before you put your name on it.
What’s inside
- 01IntentWhat the work is for, and who it is for.
- 02InputsThe data, context, and examples it needs.
- 03ConstraintsWhat good looks like, and what it must avoid.
- 04ProcessHow it should work the problem, and where to check in.
- 05The verdictHow you will judge it — and own it.
How to use it
- Keep it beside you and fill it before any non-trivial AI task.
- Pairs with The P/HX Audit: brief the work you decided to delegate.
Keep going.
The P/HX Audit
Find the routine in your week, hand it to the machine, and reclaim the hours for what is human.
Read moreThe Output Verdict
Direct, don't do — and own what comes back. Five tests for judging AI work before you put your name on it.
Read moreThe Three Eras
Blue collar, white collar, blank collar — and the shift that defines the intelligence age.
Read moreFree at the door.
Tools are the first layer of the Codex. The programmes are where you put them to work, on your own role and your own company.