It could reach across my whole computer and search the open web. It could not see the button.
Signal
I spent an hour with an AI assistant on a task barely worth naming. A sidebar had accumulated a long list of pinned projects, and my custom groups had slid beneath it, buried. I wanted the groups back on top.
The assistant was formidable, and it set about being formidable. First it offered to unpin each project one at a time. Then, more cleverly, to collapse the pinned section so the groups would rise. Then, more cleverly still, to pin the group itself. And finally, at the far edge of its ingenuity, it offered to reach into the application's own data files and rewrite the order by hand.
Each idea was more resourceful than the last. None of them was the answer.
The answer was a feature already sitting in the interface: select the whole batch of pinned items and move them into a group in a single motion. I found it almost by accident. The assistant, which could read my files and reason across the width of the web, had never seen it.
That gap is the whole subject. The most powerful thing in the room could do nearly anything except notice the plain door in front of it.
Pattern
There is a reason the elaborate ideas came first, and it is not a flaw peculiar to machines.
In 2021, researchers at the University of Virginia published a study in Nature with a title that reads like a verdict: people systematically overlook subtractive changes. Across eight experiments — stabilizing a structure, balancing a grid, improving an essay — participants reached reflexively for addition, even when removing a single piece was simpler and cheaper. The bias sharpened under cognitive load and when they had only one pass at the problem. Give a mind less room, and it stops seeing what could be taken away. It sees only what could be piled on.
Every proposal I received was additive: add a step, add a click, add a script. The move that worked was subtractive in spirit — use what was already there, and add nothing.
Half a century earlier the philosopher Abraham Kaplan named the law of the instrument, which Abraham Maslow made unforgettable: give a small boy a hammer, and everything he meets needs pounding. The usual lesson is about scarcity — one tool, misapplied. The stranger lesson fits this moment better. What happens when the instrument in your hand can write, search, and execute almost anything? Then everything looks like a job worth building. Abundant capability does not cure the law of the instrument. It inflames it. A tool that can build will offer to build, because building is what it is for — and a system of great reach is itself a kind of ladder, forever proposing the climb.
Designers already have a name for what was missing. Donald Norman, borrowing the word affordance from the psychologist James Gibson, taught us to separate what a thing makes possible from the signal that announces the possibility is there — the signifier. A flat plate on a door says push; a bar says pull. When the affordance is real but the signifier is silent, a capable person stands there shoving a door that opens toward them. The batch-move was exactly that: an opening with no handle the assistant could perceive.
Implication
For forty years the classic account of automation ran in one direction. In 1983 Lisanne Bainbridge described the ironies of automation: hand the routine work to the machine, and the human is left with the hard residue and the exhausting task of watching a system that mostly runs itself — responsibility without rehearsal.
In my sidebar that irony turned inside out. The capable system did the elaborate work. The human supplied the one thing it lacked, and the missing thing was not more capability. It was perception. I did not out-compute the assistant. I saw the button.
This is the shape of the collaboration we are walking into. As capability becomes abundant and automatic, the scarce, load-bearing human contribution migrates from doing toward noticing. The person in the loop is not a slower copy of the machine, kept on as backup labor. They are the faculty the machine does not have: the ability to look at a situation and register the opening that was there all along.
Landscape architects know the discipline. Across a lawn, feet wear a diagonal track where the paved walk should have gone — a desire path, the honest record of where people actually move. The wise response is not to fence it but to pave it. An affordance is a desire path in miniature: the move a person reaches for first is evidence of the route the task wants to take. The elaborate workaround is a walkway laid where no one was walking. To help well, and to build well, is to find the path already worn and clear it — not to admire your scaffolding while someone stands beside it, looking for the door.
Posture
When you catch yourself reaching for the ladder, stop and ask where the door is.
Name the affordance already in the environment — the feature in the panel, the person who already knows, the arrangement one subtraction away. Try removing before adding. Follow the path already worn rather than paving a detour past it. Then make the smallest move that uses what is present, and notice how rarely the elaborate solution was required.
Undersong
Power and perception are not the same faculty, and they do not arrive together. A system can be able to do almost anything and still be unable to see the simple opening in front of it — and the more it can build, the more the building crowds out the looking.
The quiet turn in an otherwise forgettable hour was this: the most advanced thing in the room needed the plainest faculty in the room, and had to borrow it from someone far less capable who happened to be paying attention.
Before you reach for the ladder, look for the door. More often than we would like to admit, it is already open.
Drafted from a working session spent reorganizing a cluttered sidebar with an AI assistant, and the gap it exposed between capability and perception. Internal notes; no external source. Ideas referenced: Gabrielle Adams, Benjamin Converse, Andrew Hales & Leidy Klotz, "People systematically overlook subtractive changes," Nature (2021), and Klotz's Subtract; Donald Norman, The Design of Everyday Things, on affordances and signifiers, after J. J. Gibson; Lisanne Bainbridge, "Ironies of Automation," Automatica (1983); Abraham Kaplan (1964) and Abraham Maslow (1966) on the law of the instrument; and the landscape-design idea of desire paths.