Once upon a time, a person confronted with a problem would attempt to solve it. This was often unpleasant. There were manuals, false starts, and opportunities to discover that the crucial screw had rolled under the refrigerator. Today, an increasingly plausible first response is to open a chat window and describe the situation to something that cannot reach under the refrigerator but feels confident about the screw.
We are becoming problem describers. The hands still hover over the keyboard, the brow still furrows, and the coffee still goes cold. From a distance, civilization appears to be carrying on as usual. Look closer, however, and the work has changed. We are composing an exceptionally detailed account of what we would like somebody else to do.
This is a remarkable promotion. Yesterday you were wrestling with a spreadsheet. Today you are briefing an agent on the desired outcomes of a spreadsheet intervention. You have become management. Unfortunately, your entire department consists of a conversational system that occasionally invents a column and congratulates itself on the successful restructuring. You are now responsible for quality assurance, staff development, and explaining why imaginary columns are unacceptable.
The temptation is obvious. Solving a problem requires patience, knowledge, and the willingness to be wrong in private for quite some time. Describing one offers immediate dignity. “Please develop a robust solution” sounds considerably more executive than “I have been staring at this error since breakfast.” Add a few constraints and suddenly your confusion has a project charter.
Yet there is real intellectual work here. Anyone who has tried to explain a malfunction to technical support knows how quickly “it doesn’t work” becomes inadequate. What were you expecting? What happened instead? Under which conditions? A clear description forces assumptions into daylight. Sometimes, halfway through typing the question, you discover the answer. The machine has contributed nothing beyond an empty box and the impression of being available.
That can be useful. So can delegating tedious work while retaining enough understanding to judge the result. We already live comfortably among borrowed capabilities. Nobody insists that using a washing machine compromises the moral seriousness of laundry. A person who automates repetitive coding may have more attention left for architecture, just as a mathematician freed from arithmetic can concentrate on the argument.
The difficulty begins when the judgment is delegated too. An answer arrives, fluent and beautifully arranged, and we experience the warm sensation of a problem having left the building. Whether it has actually been solved becomes a secondary administrative matter. The formatting is excellent. There are bullet points. Surely somebody must have checked.
This is where “brain rot” enters the conversation, wearing the confident expression of a diagnosis made in a comments section. It is a wonderfully compact phrase for a much messier question: what happens to abilities we stop practising? The slogan gets ahead of the argument. A single afternoon of asking a chatbot to summarize documents cannot settle the future of human cognition.
Still, the worry deserves better than either panic or a cheerful subscription upgrade. Consider the beginner who uses an assistant to produce code that works. There is an immediate achievement and a possible missing lesson. If every obstacle disappears before the learner has examined it, where will the experience needed to recognize a bad solution come from? Expertise needs somewhere to grow.
The experienced user faces a different comedy. They can spot the mistakes, but must spend their afternoon explaining why those mistakes matter. “Preserve the existing behaviour.” “Do not delete the tests.” “When I said simplify, I did not mean remove the difficult part.” The advertised assistant has quietly created a vacancy for a full-time supervisor of enthusiastic misunderstanding.
Agents extend this arrangement. Instead of describing one answer, you describe an outcome, permissions, boundaries, exceptions, and the circumstances under which the system should stop. Your simple request begins acquiring the density of a minor constitution. By the time you have specified what “organize my files” permits, you may understand why societies invented lawyers.
There is also a less flattering possibility: some of our newly sophisticated prompting is a way of postponing decisions. We ask for ten options, then a comparison, then a refined comparison, then a recommendation adjusted to preferences we have yet to identify. The system keeps producing. We keep commissioning. Indecision has found an inexhaustible stationery budget.
But the same conversation can sharpen thought. Ask for objections to your argument. Try solving something first, then compare approaches. Make the assistant explain a step you can independently check. The value depends partly on what you continue doing yourself. A useful exchange leaves you better able to tackle the next problem; a soothing one may leave you merely ready to ask the next question.
Perhaps our new occupation will require a peculiar kind of discipline: knowing when to write a better brief, when to inspect the work, and when to close the chat window and struggle for a while. Friction can be wasteful. It can also be the place where understanding happens, usually without a progress bar.
For now, the grand evolutionary verdict can wait. We have acquired an extraordinary new way to get things done, along with an extraordinary new way to look busy while outsourcing our uncertainty. Whether we emerge wiser remains an open question.
We should probably think about it.
First, however, we need to describe exactly what we mean by “think.”
No comments yet