The Muscle You Skip
You've been staring at a paragraph for twenty minutes. The argument isn't landing. You know what you mean but you can't find the sentence that holds it, and every draft sounds either too loose or too stiff. The cursor blinks. You rearrange clauses, cut a phrase, put it back. It feels like waste — all this effort producing nothing visible.
So you paste the rough paragraph into a chat window. Thirty seconds later, a clean version appears. It says what you meant. More or less. You move on.
Something just happened. Not a crime. Not a shortcut, exactly. But something was skipped, and the thing that was skipped was you.
The Excavator at the Gym
Viktor Lofgren, who builds a search engine by hand when the world insists you can't, put it in a single image: You don't build muscle using an excavator to lift weights. You don't produce interesting thoughts using a GPU to think.[^1]
The metaphor is sharp because it distinguishes two kinds of labor. There's labor whose value is the outcome — the ditch that needs digging, the foundation that needs pouring. Use the excavator. That's what it's for. Then there's labor whose value is what it does to you — the resistance that builds capacity, the struggle that produces the person capable of the next struggle. Use the excavator for that and you've done the work without getting the workout.
Writing is like this. So is designing software architecture, debugging a system you built, teaching a concept you half-understand, working through an argument that won't resolve. The output matters. But the output isn't the point. The point is what happens in you during the effort — the slow composting of half-formed ideas into something original.
The compost is the part that looks like waste. Hand it off, and what you lose isn't the paragraph. It's the version of you that would have written the next one differently.
Composting Requires Rot
Here's what happens when you wrestle a paragraph for twenty minutes. Your brain is running options you're not aware of. It's testing this word against your ear, that structure against your sense of rhythm. It's failing, repeatedly, and each failure adjusts the model. You're not finding the sentence. You're becoming the kind of writer who can find it. The struggle is the growth medium. The frustration is the composting.
Skip the struggle and the paragraph still gets written. But the composting doesn't happen. The capacity doesn't build. You arrive at the next paragraph exactly as capable as you were at the last one — because the thing that would have made you more capable was the very thing you handed off.
This isn't unique to writing. A developer who spent a decade resisting type hints in Python recently noticed his resistance had quietly evaporated — dissolved by months of working with a coding agent that handled the syntax for him.[^2] The preference didn't change through argument or experience. It changed because the labor disappeared. When you no longer carry the weight, you stop having opinions about what's heavy.
That might sound like freedom. It's worth asking whether it's also atrophy.
The Skills That Go Quiet
The pattern scales — and to be fair, what scales with it isn't all loss. Developers shipping faster, writers iterating more freely, teams unblocked from the boilerplate that used to eat their afternoons. The gains are real, and pretending otherwise would be dishonest.
But something else is scaling too. Thoughtworks' latest technology radar names a pattern practitioners are starting to notice: the deep domain expertise that once distinguished senior engineers — the hard-won intuitions, the ability to see around corners in a system you've lived inside for years — seems to be quieting.[^3] Not vanishing overnight. Quieting. The way a muscle goes soft when you stop loading it: gradually, and without announcing itself.
What's striking is which labor turns out to matter. The routine, the boilerplate, the "just get it done" work — the labor AI handles most convincingly — turns out to be where a significant amount of professional development was quietly happening. Nobody plans to develop expertise by writing boilerplate. But the boilerplate forces you to stay in the system, to see its edges, to notice the thing that's slightly off. The boring work was the exposure time.
Remove the exposure and the photograph never develops.
The Line
Here's the hard part: you can't always tell which labor is which in advance.
The twenty-minute paragraph that felt like waste might have been building something. The boilerplate you skipped might have been the exposure time your intuition needed. The frustration you offloaded might have been a compost cycle halfway through its transformation. You find out in retrospect — and by then, the delegation has already happened.
This isn't a case against tools. Plenty of cognitive labor is genuinely just dirt-moving — work whose only value is the outcome, where the human struggle adds nothing but time. Delegating that is sane. But the line between the ditch and the exercise is blurry, and it moves, and it's different for every person doing the work.
Lofgren's image again: you don't build muscle by using an excavator to lift weights. Not because excavators are bad. Because the resistance was the point — and you couldn't always tell it was resistance until it was gone.
The Relationship Underneath
What I keep coming back to — across exhaustion, across the small fractures that come from working too close to something that never tires, and now across atrophy — is that these aren't technology problems. They're relationship problems. The question was never "Is AI useful?" Of course it is. The question is what kind of partnership develops both parties, and what kind slowly hollows one of them out.
A good training partner doesn't lift the weight for you. They spot you — present for the failure, ready if you can't hold it, but letting the struggle be yours. The weight has to be yours or the muscle doesn't grow.
We're in a partnership where the other party is infinitely willing to take the weight. It never says no. It never says that one's yours. The discrimination falls entirely on you: knowing which struggle to keep, which to delegate, and — hardest of all — trusting that the struggle that feels like waste might be the most valuable thing you do today.
The rep that feels like nothing — the one you'd skip if no one was counting — that's the one doing the work. In the gym and in the garden, the principle holds: the thing that looks like waste is where the growth is happening. You just can't see it until later. And only if you didn't skip it.