Most people fail at this by trying too hard on day one. They open an assistant, attempt something ambitious, get a mediocre result, and quietly conclude the whole category is overhyped. This plan is built to prevent that specific outcome. It is slow on purpose.
Day 1 — Ask it something you already know
Pick a subject you know well — your job, a hobby, a place you’ve lived. Ask it to explain that thing to you.
This feels pointless. It is not. You are calibrating. You will immediately notice where it’s sharp, where it’s vague, and where it’s confidently slightly wrong. That instinct is the most valuable thing you will build this week, and you cannot get it by reading about the technology.
Fifteen minutes. Then stop.
Day 2 — Make it shorter
Find something long and dull that you actually need to read. A contract, a policy document, a thread, a report. Paste it in and ask what it says.
Then ask a better question: “What in this affects me specifically?” or “What are the three things I’d regret not noticing?”
Because you have the original in front of you, you can check the answer. This is the safest and most immediately useful thing it does.
Day 3 — Write something you were dreading
Everyone has one. The awkward email. The description you keep putting off. The message you have rewritten four times.
Tell it the situation like you’d tell a friend: who this is for, what happened, what outcome you want, and what tone would be wrong. Then let it produce a draft.
You will probably rewrite half of it. That is a success, not a failure. The blank page was the expensive part, and it just went away.
Day 4 — Argue with it
Take yesterday’s draft, or make a new one, and push back properly. Too formal. Too long. You made up a detail I never gave you. Keep paragraph two, redo the rest. Less enthusiastic.
Most people never do this. They take the first answer as the verdict. Today is the day you learn that the second and third rounds are where the real quality lives, and that they cost you about ten seconds each.
Day 5 — Catch it being wrong
Ask it something specific and checkable. A date, a statistic, who said a particular thing, details of a small local business.
Then check independently. Sometimes it will be perfectly right. Sometimes it will produce something plausible, specific, and entirely invented — in exactly the same tone.
Seeing this once, yourself, is worth more than a hundred people warning you about it.
After today you will never fully trust it again, which is precisely the correct relationship to have with it.
Day 6 — Give it your voice
Paste in three or four things you have actually written — emails, posts, anything in your natural register. Say: “This is how I write. Match it.” Then ask for a new piece.
Left alone it defaults to a slightly eager, slightly corporate voice that flattens everything into the same shape. Feeding it your own writing is the single most effective correction, and almost nobody does it.
Day 7 — Choose your one thing
Look back over the week. One of those days will have produced something that felt genuinely useful rather than merely interesting. That is your one thing.
Now do only that, repeatedly, for a month. Do not go hunting for new uses. Let the one thing get fast and reliable and boring.
Boring is the goal. Boring means it has stopped being a novelty and become a tool you own.
If the week didn’t take
That happens, and it usually means one of two things: the tasks you chose were too big, or they were things you couldn’t judge the quality of. Both are fixable. Try again with something smaller and more familiar, or tell me what you tried and I’ll tell you what I’d change.