The first thing you learn at a potter's wheel is centering. You wet your hands, brace your elbows against your thighs, and press a lump of spinning clay until it stops fighting you. When it's centered, the mass goes still under your palms, no wobble, no drift. Get it wrong and everything you build on top of it pulls sideways and tears. Most people need months before the clay stays in place.

You can skip all of that. A slip mold pours liquid clay into a plaster form and releases a bowl in an hour, identical every time. The bowl is real. The potter is not.

That split between the bowl and the maker is the thing worth looking at, because our tools keep widening it.

A tool can hand you the output, or it can make you do the work that produces the output. These look the same. They aren't. The mold gives you a bowl. The wheel gives you a bowl and a person who can throw the next one without it.

I've watched a designer open a layout someone else generated and get stuck within minutes. Not because the file was bad. The spacing was even and the components were neatly named. But when the brief changed and a section had to move, there was nothing underneath to reason from. No record of why the grid was eight columns and not twelve, why this card carried a shadow and that one didn't. The decisions had been made and then thrown away. What was left was a finished thing with no inside.

Compare that to a system someone built by hand, badly at first, revising as they went. It might be messier. But every choice in it answers a question the maker can still remember asking. When the ground shifts, they know which parts were load-bearing.

This is the part tools tend to erase. Not the artifact, the reasons. A layout or a codebase carries two things: the result, and the trail of decisions that got there. The result is what ships. The trail is what lets anyone change it later without breaking it.

Strip the trail and you get something that works until the day it has to be different. Then nobody can touch it, because nobody knows what it's holding up. We call that legacy and blame the people who inherited it. The trouble started earlier, when a tool delivered completion and skipped the part where understanding accumulates.

Designing for craft instead of completion is mostly a question of what a tool refuses to do for you. The good ones leave the consequential decisions in your hands and take the tedium off them. The bad ones do it the other way around, automating the judgment and leaving you to tidy the result.

The tools that produce output fastest are the ones doing the most of your thinking for you. That's the trade, stated plainly. Every task a tool absorbs is a task you stop getting better at. Sometimes that's the right call — I don't want to hand-set type to send an email. But it's worth knowing what you're handing over, because the work you give away is often the work that was teaching you something. A bowl you never learned to make is fine right up until you need a different bowl.