Time Under Tension

I hate saying ‘ay, eye’ out loud. Almost anything said near them arrives sounding preachy, or starry-eyed, or like a line a vendor would put on a slide, and I work in this field, so I catch myself doing it too. What I want here is something quieter than a position. I want to hold two things that are both true and let them sit beside each other without forcing one to win, because most of the honest conversations about this technology happen in that uncomfortable middle, where the variables are partly known, partly unknown, and partly nobody’s to control.

Truth #1. AI lets me do a great deal without friction or tension, and some of it it does better than I currently can, and faster. I am not going to pretend otherwise, and I have little patience for the people who do, the ones who have decided the whole thing is a passing fad and will not look at how good it has become.

Truth #2, the one I cannot put down. Almost everything I can do well today, I can do because of the long, slow, sometimes maddening stretches that came before it. Lifters have a phrase for the part of a movement that builds the muscle: time under tension, the bit where you are holding the load, it hurts, and nothing visible is happening. Learning runs on the same principle. Psychologists draw the line with drier words, performance against learning: what you can do while the support holds you up, against what is left once it is taken away.¹ The two come apart more often than we like to admit, which is how a finished, competent result can sit on top of almost nothing. The capability forms in the stretch where the endpoint is unknown and the path to it unclear and you have to stay there anyway, bored or frustrated or certain you are getting nowhere. That last feeling is the least reliable witness you have, because the difficulty that builds you tends to feel, from the inside, like failure.² If I reach for AI to remove that stretch every time it shows up, the convenience is real, and so is the slow erosion underneath it. In some quiet, compounding way I will be a less capable human a few years from now than I am today.

It is the students I work with who worry me most here, and the worry has little to do with whether they use the tools. Whether a student works with, writes with, and uses AI or not, if they never sit in the discomfort, the conflict of a problem that will not resolve, the discipline of staying with something dull, the time under tension with no guaranteed payoff at the end, the capability does not develop. You cannot skip the formation and keep what the formation produces.

Let me follow that down a side road for a moment, because it is where my own thinking keeps going. A colleague will say, with real concern, that students can no longer write an essay without AI. Someone reasonable will ask, fairly, why a student needs to write an essay at all. It is a good question, and the answer is not about essays. Writing is thinking made visible, and thinking is the skill we are trying to grow. You cannot think well about nothing, either: critical thinking needs a stock of knowledge to work on, something to pattern-match the new against the known. So when I talk about essay writing I am not talking about the essay. I am talking about reading the novel, sitting with its ideas, weighing viewpoints that disagree, and learning to leave the grey parts grey instead of forcing a tidy conviction onto something that deserves contemplation. The essay is only the evidence that the harder, invisible work took place. Lose the work and the evidence is worthless, whoever or whatever produced it. We can already watch this come apart. In one controlled study, students who wrote with ChatGPT handed in better essays and learned no more for the privilege; what they had delegated to the tool was the self-checking, the monitoring of their own understanding, that turns writing into thinking in the first place.³

The catalyst for this post was stumbling across the poem by Shawn Smucker that is ironically titled ‘Please Use AI’. He spends it urging you to hand the slow, human parts of your life to a machine, the meal plan, the trip you are planning with your child, the wedding toast, the eulogy, so that you feel, line by line, what you would be giving away.

Truth #3: Forgive me, Cognition, because I have sinned. Building a meal plan and writing a trip itinerary were exactly the first things I ever did with ChatGPT, back at version 3.5. I remember being amazed. Reading the poem, I felt like a fraud, cringing because I had done the very things it mocks. The honest defence I can offer is that this is not how I use these tools now, and not, I think, how they are best used. Smucker has named the shallow version of AI use, and named it better than I could, and the recognition stings in the useful way.

What I am doing with these tools now is stranger than those first attempts, and more hopeful. I am trying to use AI in ways that make me think differently about the world, about people and systems and ideas, about purpose and creativity and the work itself. Used like that, it does not flatten my thinking, it pushes against it. Where I have offloaded something, what has come back to fill the space is the most divergent and unexpected thinking I have done in my life.

The strongest objection to all of this is that I am drawing the boundary in the wrong place. Minds have always thought with tools, the argument goes, and there is no clean seam where my thinking stops and my instruments begin.⁴ I take the point, and I use it as the test. A tool that extends me leaves me more capable when it is switched off; a tool I can no longer work without has not extended me at all, it has replaced the part of me that used to do the work.

So both things stay true. These tools can hollow out the time under tension that makes us capable, and they can, in the right hands and with the right intent, put us under a new kind of tension altogether. I have not reconciled those two, and I am not going to try, because I would rather hold both at once than reach for the easy comfort of the doom or the hype. Wanting to innovate is human, and so is wanting to avoid discomfort. But the discomfort itself, the struggling and the failing, is the uniquely human part, and if we use AI to skip it we rob our future selves of the chance to be fully human.