Why I still write by hand, sometimes

I have practically all the technology I need to avoid writing by hand. I can take notes on my computer, store documents in the cloud, find any information in seconds, and use artificial intelligence to summarize, translate, explain concepts, or help me develop software, as everyone does nowadays.

And I use all those tools. I have for a long time, and I’m still amazed by how much they help us with these kinds of tasks.

Even so, I still keep notebooks on my desk and continue to buy printed books, although I use my Kindle a lot to read, of course. When I study mathematics (for everything related to quantum computing), I often work through calculations step by step on a sheet of paper. And although I use AI assistants to program, I also try to set aside time to write code without any help—that is, just me and the IDE, perhaps with the occasional help from Google or GitHub Copilot.

Let me explain why I think this matters.

It’s not nostalgia. I don’t think writing with a pen is better than typing on a keyboard, or that programming without artificial intelligence necessarily makes us better programmers. In a few years, we’ll probably see the opposite.

My concern is something else.

We have more and more tools that can do things for us, and that’s fine—of course it is. The question is what happens to our ability to do those things when we no longer need to.

A tool can amplify an ability or replace it

What I see around me, and what I read from one person and another, suggests that we’re moving more toward the latter.

Technology has always allowed us to delegate some of our work.

A calculator saves us from doing certain operations by hand. A GPS lets us reach a place without knowing the way beforehand. A search engine saves us from remembering where we found some information. An IDE catches errors that we once had to find ourselves.

None of this is necessarily bad—unless you think like an Amish person and live hidden away in a village in the backwoods of rural America. Even that I can understand, and I don’t see it as a bad thing either.

Delegating tasks lets us focus on different and often more important problems. Much of technological progress is precisely about this: building tools that help us do more with less time and effort.

Generative artificial intelligence has greatly accelerated this process because we’re no longer automating only mechanical tasks. We can delegate activities that, until recently, we directly associated with intellectual abilities: writing, summarizing, translating, reasoning about a problem, generating code, or explaining an idea. And it works well—extraordinarily and surprisingly well.

There is, however, an important distinction between using a tool to amplify an ability and using it to replace that ability.

There isn’t always a clear boundary between these two situations. In fact, it is probably a matter of degree.

I say this because after decades of observing the impact of the so-called information society, what I see is precisely people who are more misinformed, more like sheep, and less thoughtful than before.

I can use an AI assistant to explore alternatives to a solution I already understand. But I can also ask it directly for a solution to a problem I don’t yet know how to solve. The result may look similar, but intellectually these are very different situations.

In the first case, the tool expands what I can do.

In the second, there’s a risk it will keep me from practicing precisely what I need to learn.

Being more productive doesn’t always mean learning more

This distinction becomes especially important when we confuse productivity with learning.

If my goal is to complete a task, it makes sense to use whatever tools let me do it as efficiently as possible. It would be strange to systematically reject a calculator, compiler, search engine, or AI assistant just because I could do some of its work manually.

But learning has a different goal.

When I’m learning, effort isn’t necessarily a problem to eliminate.

Sometimes it’s part of the process.

Take a mathematical calculation. If all I need is the answer, using a tool to get it immediately may be the right choice. But if I’m studying which transformations lead to that answer, getting only the final number solves a different need.

There is an important difference between getting an answer and understanding why that answer is correct.

The same thing happens in programming.

We can ask an assistant to implement an algorithm, use the generated code, and check that it works. This can be extraordinarily productive. But writing the algorithm ourselves requires us to keep track of decisions, constraints, and mental models that partly disappear when we’re handed an implementation directly.

One activity aims to get the work done.

The other aims to exercise an ability.

Both make sense, but they don’t necessarily have the same purpose.

Why I still write by hand

Writing with pen and paper is much less efficient than typing on a keyboard.

I can’t easily rearrange a paragraph. There’s no Ctrl+Z. Finding something I wrote months ago can be difficult, and retrieving an idea depends much more on how well I’ve organized things.

Even so, sometimes I prefer it.

Especially when I want to think.

The slower pace stops being just a disadvantage. Writing more slowly also means I have to decide what I want to write before I do it. I can’t produce text as easily, and that changes my relationship with what I’m writing.

I don’t use paper for everything. That would be absurd in the way I usually work. I use it deliberately at certain times because I want to keep practicing writing as an intellectual activity, not just as the process of entering text into a tool.

Reading is similar for me.

I use digital devices and constantly read documentation and other content on a screen. But I still read printed books. For me, they offer a different kind of space, with fewer opportunities to switch context immediately and a more direct relationship with the reading.

I don’t think that automatically makes paper a superior medium. I simply find value in keeping both ways of reading.

Doing the calculations even when a machine can do them

Mathematics makes this even more obvious.

Today, getting the result of countless operations is trivial. We have calculators, computer algebra systems, notebooks, and AI systems capable of working through expressions and explaining their steps.

I use them too.

But when I’m studying a mathematical concept, I often still take out paper and work through the operations myself.

Not because I think it’s important to prove I can calculate something a machine can solve in a fraction of a second. The value is in the process itself: identifying what I’m looking at, choosing a transformation, making a mistake, catching it, going back, and understanding why one step leads to the next.

The mathematical result matters.

But when I’m learning, the path to the result is part of the result too.

Systematically removing that path may make us faster. I’m less sure it always makes us better at understanding what we’re doing.

Programming without an AI assistant

Programming is probably where this reflection feels most obvious to me right now.

AI-powered coding assistants are extraordinarily useful. They can generate code, complete functions, explain implementations, suggest tests, find bugs, or propose alternatives in seconds.

I use them, and I have no intention of stopping.

But precisely because they’re getting better all the time, I think it also makes sense for me to keep programming without them for part of my time.

When I write code without assistance, I have to maintain a mental model of the problem, remember the language’s features, think about data structures, decide how to divide up a solution, and confront my own mistakes directly.

It’s not always the most productive way to work.

That’s exactly the point.

If I’m solving a problem at work, productivity may be the priority, and I’ll use whichever tools I think are appropriate. If I’m learning Rust, studying an algorithm, or trying to understand a particular mechanism better, perhaps my goal isn’t to finish sooner.

Perhaps my goal is to make my brain work.

An assistant can help me enormously afterward: compare my solution with alternatives, point out a problem I hadn’t noticed, explain a language feature, or suggest optimizations. AI can then become an amplifier for learning.

But for it to amplify something, I want to make sure I’m still building an ability of my own first.

Intellectual effort can be trained too

There’s an analogy I find especially useful.

We have cars, elevators, and escalators. None of those inventions has made walking or running useless.

When we go for a run over several miles, we don’t do it because that’s the most efficient way to get somewhere. We do it, among other reasons, because the effort itself is valuable to our bodies.

I think we can apply a similar idea to some intellectual abilities.

Solving a problem without immediately looking up the answer. Trying to remember before searching. Working through a calculation. Reading for an extended period without constantly switching contexts. Writing an idea from scratch. Programming a solution before asking an AI to generate it.

These activities require cognitive effort.

And we have more and more tools that can save us from it.

The problem arises if we come to see any intellectual effort as an inefficiency that should be eliminated. Because in some situations, what looks like friction can be part of the training.

Don’t use less technology—use it consciously

I don’t think the answer is to adopt a kind of resistance to technology.

Much less to artificial intelligence.

That would mean giving up extraordinary tools and would also turn the issue into a false choice between technology and human abilities. We can have both.

For me, the point is to distinguish when I want to delegate a task and when I want to practice it.

If I need to be productive, I’ll automate whatever makes sense to automate. If I need to learn or want to preserve a particular skill, sometimes I deliberately choose not to.

That decision changes technology’s role. Instead of using it automatically whenever it’s available, I choose a tool according to my goal.

I don’t need to do all my calculations by hand to preserve my ability to reason mathematically. Nor do I need to write all my texts by hand or give up coding assistants to preserve my developer skills.

But I do think I need to keep practicing the things I consider important.

Because delegating a task and losing the ability to do it are not the same thing.

And I’d like to keep that distinction.

The abilities we want to preserve

Everyone probably has a different list.

Mine includes writing, reading attentively, reasoning through problems, working with mathematics, and continuing to program without assistance for at least some of my time.

Not because I think the future will need less technology. I think exactly the opposite.

My concern isn’t that technology will be able to do more and more things for us. I find that an extraordinarily interesting consequence of technological progress.

My concern is that we’ll no longer know how to do those things ourselves.

That’s why I still keep pen and paper close by. That’s why I still open a printed book when I could read it on a screen. That’s why I fill pages with calculations whose results I could get immediately. And that’s why, sometimes, I turn off the assistant and simply program.

I hope that if you’ve read this far, you’ve understood the metaphor of the pen and paper.

I’m not trying to be more efficient in those moments.

I’m training.

Perhaps the question we’ll need to ask ourselves more and more often isn’t how much we want to use technology. Nor how much work we’re willing to delegate to artificial intelligence.

The question I find truly interesting is a different one:

What abilities do we want to preserve when technology means we no longer need to use them?