Coding Is an Art, and No One Is Defending It

The fashionable position about LLMs and coding goes like this: writing software is translation work. You have an idea, you have a specification, you have a feature in your head - and the code is just the syntax a machine needs to execute it. The programmer is a translator standing between intent and execution. A better translator obsoletes the translator.

This framing is doing all the work in the current discourse, even when no one says it out loud. It's why "natural language is the new programming language" sounds like a profound observation instead of a category error. If code is translation, then of course the translator goes when the translation gets automated.

The same argument is now being aimed at everything that used to demand skill. Suno tells you music is translation - the sounds in your head into the file on your hard drive, and the years on an instrument were just mechanical overhead to bridge the gap. Likewise, Midjourney encourages that a pencil or paintbrush is a translation tool you no longer need.

Music and painting have loud defenders. Musicians defend musicianship. Illustrators defend illustration. The defenders are everywhere.

Coding has almost none.

The easy read on that silence is that coding simply isn't art - that the discipline never had an artistic soul to defend in the first place. I don't think that's true. I think the silence has a different cause, and it's a problem coders made for themselves.

Walk into any engineering org and ask what matters. You will hear: the product, the system, the user, the outcome. You will not hear: the code. The code is treated as cost - the price you pay to get the thing that matters. A senior engineer who obsesses over the shape of a function instead of the behavior it produces is gently steered back toward shipping.

This is a coherent set of values. It built most of the software you use. But it is not an artistic set of values, and it has been the dominant self-image of the field.

So when LLMs arrive and offer to remove the cost, the field has no defense ready. It already agreed the cost was the problem.

Meanwhile, every working developer has had the same experience. You open a pull request, you read a few lines, and you know who wrote it before you check the author. Not because of conventions or tooling preferences - those are surface. You recognize the thinking. The shape of how a problem was decomposed. The rhythm of how state moves through a function. The decisions about what to name, what to abstract, and what to leave alone.

That recognition is the same thing a literature professor does identifying a writer by a paragraph, or a musician identifying a composer by eight bars. It's signature. And signature only exists when the maker had choices and made them in a particular way.

A translator's job is to disappear. A good translation of a French novel into English is one where the English reader forgets a translator was involved at all. If that were the right model for code, then two competent engineers given the same specification should produce roughly the same code, the way two competent translators produce roughly the same translation. They don't. They produce code that is recognizably theirs, and the differences are not noise - they are the substance of the work.

But the people who code this way have always been a minority inside the discipline, and they have never been the ones writing the public narrative about what programming is.

Selling timelines, selling productivity, selling outcomes to people who pay for outcomes. Nobody pays for a beautifully shaped function. So nobody talks about beautifully shaped functions in public.

Most musicians and illustrators currently defending their crafts are not defending them on outcome grounds. They are not arguing that hand-drawn illustration produces a better illustration than Midjourney. They are arguing that the making matters - that the practice has value independent of the artifact it produces.

That argument is available to coders too. It has just never been made loudly, because the coders who would make it have spent their careers being told it was unprofessional to think that way.

The result is a discipline arriving at its own crisis without the vocabulary to defend itself. When the tools came for music, musicians had centuries of language about craft, voice, expression, and tradition ready to deploy. When the tools came for code, coders had decades of language about velocity, throughput, and shipping. One of those vocabularies is useful in this fight. The other is the fight.

I don't think coding is dying. The people who experience it as art will keep doing it, the way people still hand-bind books and throw pots and write graphics engines for fun. The practice survives the economics. It always has.

But the public idea of coding - the one that gets written about, hired for, invested in - is going to lose the artistic frame entirely, because the artistic frame was never really there in public to begin with.

That's the part worth being sad about. Not that the art is dead. That too few bothered to call it art while it was alive.