AI isn't for writing your sentences. It's for carrying the repetitive work.

You get home, open the laptop to finish one last thing. That last thing — a machine does it in five minutes.

It isn't hard work. It's repetitive work: stitch a few numbers together, paste them into the right cells, name the file, send it off. You did it yesterday, you'll do it again tomorrow, almost identically. It doesn't need your mind. It only needs your hands — and your hands cost far more than that.

Most newcomers to AI imagine it does the opposite: polishes a paragraph, dreams up a headline. Fine. But that's the smallest slice, and the one most likely to come out bland. The work it can genuinely take off your plate — and should — is that repetitive thing, the one you're still doing at seven in the evening after everyone else has gone home.

01You're using it through the wrong door

One misconception blocks most people from day one: treating AI as a clever pen. You type "write me a paragraph," it writes one. Nice — except you still have to read it, fix it, bend it into your own voice, and half the time it'd have been faster to write it yourself. Used that way, AI gives you a so-so writing assistant, and you conclude "it's overrated."

The right door is somewhere else. Not "write me something good," but "do the thing I have to do over and over." Those two doors lead to two completely different experiences:

Same tool, two doors. Most people walk through the left one and back out. The right door is less impressive — but it's where AI actually pays you back.

02Repetitive work belongs to the machine, not to you

There's a very ordinary line between two kinds of work. The first kind needs you: it takes judgment, it takes context, getting it wrong has real consequences — reviewing a contract, making a call, writing something that has to have soul. The second kind just needs getting done: fixed steps, the same every time, right-or-wrong obvious at a glance — assembling a report, reformatting, copying numbers from here to there.

Your whole day is a mix of the two. The trouble is that most people pour equal effort into both — sitting down to assemble a report with the same focus they'd give a serious decision. The machine can't draw that line for you, but you can. And once you've drawn it, the question changes entirely: this repetitive thing — why am I still doing it by hand?

The usual answer is "it only takes ten minutes." But those ten minutes don't stand alone. They're ten minutes multiplied by every time you redo them. The report you assemble each morning, twenty minutes a go, two hundred and fifty working days — that's over eighty hours a year on one mechanical motion. That isn't ten minutes. That's two weeks of your working life, evaporating into a job that never needed you.

The thing that can do that repetitive work — on command, tireless, without forgetting halfway — is what people call an agent. You don't need to understand how it runs, the way you don't need to understand an engine to drive. You only need to know one thing: it's something you delegate to, not a chat box to make small talk with. And the first job worth delegating is always the repetitive one.

03Eight familiar doorways

Almost anyone at a desk runs into at least a few of these. Each is a concrete pain — and each has a deliberate way through, not just "try a little harder."

These eight aren't separate — they're eight ways into one new craft: the craft of handing work to something very fast that needs you to know how to hand it.

Notice this: the first seven are pains, the eighth is a stuck. And that stuck — standing before the blank field with no idea what to type — is what keeps most people out of the game entirely. Not because they have no work to hand over. Because no one has shown them the first job worth handing.

04Don't ask "how do I learn AI" — ask "what should I hand over first"

This is where I watch people take the wrong turn at step one. They treat AI as a subject to study — find a course, collect tips, hoard good prompts. They study endlessly without touching real work, because real work is messy and tips are tidy and easy to collect. Six months later they know every tip and still haven't handed off a single real job.

The right direction is the opposite, and far simpler: don't study the tool first, look at your work week first. Find one repetitive job — small, low stakes, where a mistake hurts no one — the report you assemble by hand, the template email, the list you re-sort every morning. Hand over exactly one of those. Botch it, even — it's small. You'll learn more from one real handoff than from ten lists of tips, because the real handoff teaches what tips can't: how to describe it clearly, how to spot where it misread you, how to re-hand it cleanly.

And once you've handed off the first one, the other seven doorways open on their own. You'll hit the moment it answers wrong with confidence, the moment it forgets the context, the moment you wonder whether you should be babysitting. Each of those is a lesson waiting — and each has someone who paid for the lean way through it. You don't have to faceplant on each one yourself.

This whole garden is built around exactly those moments. Not to teach you "what AI is" — but so that when you trip over one of those eight, there's a deliberate path ready instead of you trying a little harder again. If you read just one more piece, start at the blank-field problem: which job this week is worth handing over first. It's the narrowest doorway — but it's also the only one that, once you're through it, widens all the rest.