The line everyone quotes is “we shape our tools and thereafter our tools shape us.” Marshall McLuhan gets the credit but he didn’t write it. His friend John Culkin did, in a 1967 essay about McLuhan, and pinned it on McLuhan. But the original idea came from Winston Churchill; when rebuilding the bombed House of Commons in 1943, he told Parliament, “we shape our buildings, and afterwards our buildings shape us.” Somewhere “buildings” became “tools.” The line about how tools reshape us got reshaped.

We reach for the back half as a warning. The tool is doing something to you and the something is bad: skills rot, attention frays, you forget how to do it by hand. That reading is everywhere with AI, and it’s half a thought. It counts only what’s taken. The quote doesn’t say the tool makes us worse. It says it makes us different, and different is not necessarily diminished.

I work in software; that’s where my examples come from.

Reinventing ourselves, accelerated

I can’t cross a city without my phone, do much arithmetic in my head, or remember a phone number. Go back far enough and the story repeats: a literate culture gave up the trained memory that oral cultures ran on. Socrates worried that writing itself would breed forgetfulness.

Each was a skill a tool took. None left me limping and diminished. I became someone who doesn’t need it, and the room it opened up is where I grew instead. Sometimes that happened willingly; other times it was inevitable. I cannot practically choose not to be literate. Something always goes. I care more about who it makes me.

The one new thing is speed. Writing reshaped us over centuries; a chisel shaped the hand over a career. Slow enough to keep up with. AI isn’t: hand a problem off for a few weeks and the skill is already going soft. Tools have always changed us; that part’s fine. The speed is what I’m unsure of: whether we can adapt fast enough.

Consider how experts are made. You became a senior engineer by being a junior one: fixing the boring bug, wiring up the part nobody wanted. That work built judgment. It’s also exactly what agents are best at now, so the rung you started on is the rung the machine stands on. No grunt work, no new experts? Maybe. But new tools redefine expertise instead of ending it: the navigator became the person who knows when the map is lying. I suspect that the next senior engineer is made the same way, by directing and judging instead of grinding through it.

But can we build that judgment without the grunt work that used to create it, and before the change overwhelms us? The danger is that the judgment you need to catch the machine is exactly what the tool takes over. Lean on it long enough and it fades, right when you still need it. The skill the tool erodes is the skill it needs you to keep.

We shape them too

We treat a shift this big as weather: too large to push on, adapt or get flattened. That’s only the back half. We shape our tools too, and that’s the half we forget.

Sometimes we do it without deciding to. A shortcut wears across the grass because enough people cut the same corner, and then it pulls everyone else onto it. What skills we let slide, what we keep by hand, what we come to call good work: it all drifts the same way, the sum of small choices no one coordinated. No one chose it; everyone did.

Other times it’s deliberate. Aviation has spent decades on exactly this. Hand the flying to the autopilot and the hands keep their touch; it’s the mind that drifts, the habit of staying ready to take over. When the automation quits and gives control back, the pilot is slow to catch up, right when it matters most.1 Rather than accept that as the price of the tools, they named it and built training and checklists to fight it. Nobody handed that down; people who refused to treat the situation as fixed put it there.

At least in software, none of this is all bad. But it won’t be gentle: the job ahead looks very little like the one behind, the disruption reaches past us to the companies built on the old roles, and I can’t promise we adapt fast enough.

Either way, what feels cast in stone mostly isn’t. We shaped these tools, and we are not powerless over how they shape us back.


  1. Stephen Casner and colleagues found that pilots’ manual control skills hold up well under automation; what fades is the cognitive habit of staying engaged and ready to take over, which depends on actively supervising rather than passively watching (Casner et al., “The Retention of Manual Flying Skills in the Automated Cockpit,” Human Factors, 2014). The same concern runs through the FAA’s work on automation dependency. ↩︎