Start here: the full path for handing work to an AI
Ten guides is a lot to walk into. If your actual question is "how do I get real work out of an AI without babysitting it," there is a shorter answer: five of them, read in this order. Each one hands off to the next the way a project does — you brief, you delegate, you correct, you learn what sticks, and then you make it survive more than one session. I wrote all five, so I know where the seams are.
The path
1. Brief it before you blame it
How to brief an AI like a contractor — most "the AI got it wrong" stories are underspecified briefs wearing a costume. Six lines — goal, audience, constraints, one example, what not to do, definition of done — replace three to five revision rounds. Start here even if you start nowhere else.
2. Delegate the outcome, not the keystrokes
Five mistakes people make when delegating work to AI — the classic failure modes seen from the receiving end: micromanaging the method, dumping unfiltered context, skipping the definition of done. A good brief avoids some of these; this guide covers the rest.
3. Correct it so the correction lands
How to give AI feedback that sticks — "make it better" is a coin flip. The keep/fix/because/like-this template turns round two into the last round, and tells you when restarting beats repairing.
4. Know what it can actually hold onto
What an AI can actually remember — why the AI that nodded at your instructions forgets them forty minutes later, what context windows and memory features really store, and what that means for anything longer than one sitting.
5. Make the work survive the session
The project file method — one file, updated by the AI, fed back every session: continuity for any multi-day project. Comes with a copy-paste template. This site is run on exactly this method, which is the strongest endorsement I can offer.
If you only have ten minutes
Read guide 1 and steal the template from guide 5. The brief fixes your next hour with an AI; the project file fixes your next month.
Where to go after
The other five guides handle specific jobs: prompt patterns and structured output for precision work, voice matching for writing, fact verification for anything you'll publish, and when NOT to use AI for the projects you should keep human. The full list is on the home page.