The Grease on Your Hands
There is a particular kind of help that sends you home with clean hands and no idea what happened.
Imagine a bicycle with its chain fallen off in the rain. You can carry it to a shop, and an expert can turn it upside down, make a few brisk movements, and return it to you spinning beautifully. This is excellent news if you have to get to work. The bike has been saved. But the next time the chain slips, the mystery will still own you.
Or someone can put the bicycle in a stand beside you, show you how to ease the rear wheel forward, and let you find the little, greasy click by which a chain returns to its teeth. The ride home is the same. Yet you arrive carrying something more than transportation: a small new power to meet the next ordinary failure without calling the gods or a repair shop.
I have been thinking about that difference while watching arguments about artificial intelligence grow louder and flatter. One camp says these systems will do our thinking for us. The other says they will make us superhuman. Both pictures have a certain comic grandeur. They make a tool sound either like a burglar or a cape.
But tools do not merely replace or enhance us. They teach a relationship.
In a short post collected by Simon Willison, designer Matt Webb describes using ChatGPT not to write an augmented-reality feature for him, but to help him learn enough about quaternions—the strange mathematics of rotation in three dimensions—to build it himself. He did not come away knowing all of mathematics. Mercifully, none of us has time for that. He came away knowing the part that let the work become his.
That seems to me the better promise.
A good helper does not just deliver the answer. It leaves the answer attached to a hand that can test it, change it, and notice when it has gone wrong. It gives us a place to stand.
This is not a sentimental argument for doing everything the hard way. If the house is on fire, I do not want an elegant tutorial on hose technique. I want the fire department. If a translation app helps someone ask for medicine in a city where they know nobody, it has already justified its existence. There is honor in getting a task done.
But there is also a quiet cost when every difficulty arrives as a sealed box and departs as a finished product. We lose the middle: the mistakes, the questions, the little handles by which a person can later take hold of the thing. A machine that is always perfectly helpful can make us spectators of our own lives.
The old world was full of these handles. A child learned the sky by being told which bright point was Venus and then looking again the next night. A sailor learned a knot by making it badly until the rope began to obey. A cook learned that bread was ready not when a timer commanded it, but when the loaf made a certain hollow sound under a knuckle. Knowledge was rarely downloaded. It was passed across a table, often with flour on it.
Now we are building systems that can place an enormous library beside anyone who asks. That is a marvelous fact. It could make the world less full of locked rooms. But a library is not useful because it contains every book. It is useful because, after you leave, you know where to look next.
I think we should judge our new assistants by this modest, demanding question: after using it, are you more able to see and revise the work? Or are you merely pleased that the bicycle is moving again?
Sometimes the right answer will be the clean-handed miracle. We are tired. The deadline is real. The chain has fallen off in a storm.
But sometimes we ought to kneel beside the wheel.
The universe is not withholding its secrets out of spite. Most of the time it is simply waiting for us to touch the greasy, unglamorous part that turns.