The tools that disappear

The best tool you use is probably not the one you talk about.

It is the one you reach for without thinking. The one that does not interrupt your attention to explain itself, surface its menus, or ask you to learn its vocabulary before it becomes useful.

It feels obvious.

That is the point.

Tools that disappear are rarely simple by accident. They take deliberate work to build. Someone had to decide what the tool should not ask of you. What it should not expose. What it has no business making you carry.

A good tool does its part, then gets out of the way.

Bad tools make themselves the subject

You can identify a bad tool by the amount of attention it asks for.

Not the attention your work requires. The attention the tool requires.

Its setup. Its structure. Its terminology. Its preferences. Its way of naming things. Its idea of how your work should be organized before you have even begun.

Bad tools are not always badly built. Some are technically impressive. Their failure is more specific: they optimize for the tool's internal logic instead of the user's actual work.

They make you adapt to them before they have earned the right.

Disappearing is earned

A tool earns the right to disappear by being correct more often than it requires correction.

By loading before you notice the wait. By organizing itself around how you already think. By failing without leaving a mess behind. By knowing when not to speak.

This is why great editors — for text, code, images, products, anything — feel less like software and more like extensions of thought.

The best moments with them are the moments when you forget the interface exists.

That forgetting is the product.

What this means for AI tools

AI tools are still in a phase where they often make themselves the subject.

They explain what they are doing. They surface uncertainty. They ask for confirmation on things that do not need it. They show process when what you needed was progress.

Some of that is useful. You should know when a tool is making a meaningful choice.

But much of it is the tool asking to be watched.

That is a cost.

The AI tools that last will be the ones that know when to disappear. The ones that surface decisions only when those decisions are genuinely yours. The ones that move fast enough, and correctly enough, that your primary relationship is still with the work.

Not the interface. Not the model. Not the system.

The work.

What the tool asks you to carry

When choosing a tool, the useful question is not only what it can do.

It is what it asks you to carry.

What do you have to remember? What does it make you translate? How often does it interrupt the thing you were actually trying to do?

The best tools have low carrying costs. They fit into the way you already work. They make the next step clearer. They remove weight without making a performance of it.

That kind of product is hard to build because the best parts are meant to become invisible.

Someone still has to care about them.

Someone still has to decide what should be hidden, what should be shown, and what should never become the user's problem.

Good tools do not disappear because there is nothing there.

They disappear because someone did the work.