Code is cheap, Show me the Art !

<TLDR; GenAI tools are compressing the craft phase of software engineering, the tedious labor of translating ideas into code, into conversations. What remains feels closer to art: intent, direction, judgment and taste.>

---

I still reminisce about my first-ever program, written in Java, to draw Pascal’s triangle for a class in high school. It was magical and a proud moment when everyone else wanted to know how I solved the problem. The craft of writing software was so endearing that I gravitated toward it.

Fast forward 25 years: a quote from Rick Rubin’s book The Creative Act: A Way of Being made me reevaluate what I think about my favorite craft.

… Craft phase is one of the least glamorous parts of the
artist’s job. There is creativity involved, but it often carries less of the magic
of exploration and more of the labor of brick-laying.
This is the point in the journey where some struggle to carry on. For now,
we need to look away from the open field and turn toward a winding staircase
a hundred stories tall. A long, precarious climb lies ahead.

Rubin describes how artists encounter real-world seed ideas, experiment with them, find what’s most moving, and then arrive at the most laborious part: the craft phase.

I don’t usually consider myself an artist, but I can’t ignore the parallel with my personal projects and I can connect a lot of my failures to that “staircase”. You might have heard of the Google graveyard. I have my own graveyard of software and hardware projects: overflowing private GitHub repositories, half-finished hardware boards, and folders organized in N different ways across M servers (home and cloud).

Almost every project follows the same arc:

• It starts with a seed.
• I experiment and explore.
• Then it stalls as soon as I hit the craft phase.

The craft phase is where I slip into perfectionist mode. It gets serious. It takes energy and time. And it literally drains the fun out of the project until it feels like a second job. At that point I decide I don’t want two jobs. I’m not good at striking the right balance between hobbyist vs pro-dev.

In the past two years, the way I write software has fundamentally changed. Not because I suddenly became more disciplined, but because the craft, this manual labor of translating ideas into code has started to compress into conversations. The hours I used to spend wrestling with syntax, debugging, formatting and searching now often happen through back and forth with GenAI tools.

And what remains, what I’m left doing feels closer to art.

We used to valorize the craft: the muscle memory of language semantics, the encyclopedic knowledge of bash commands and language syntax, the ability to both architect a system and bang out the code. That craft mattered, and it still does. But it was always a means to an end.

Almost nobody hires a software engineer because they want lines of code. They hire them because they want the ability to solve problems, build systems and create value. The code is the artifact, not the point.

When I work with GenAI tools now, the dynamic feels less like “using a tool” and more like directing a skilled apprentice. I describe what I want. I critique what comes back. I refine, redirect, reject, reflect, reset and redesign.

The actual keystrokes that produce the code aren’t always mine anymore, but the outcome can still be exactly what I want. I’m not a believer in vibe-coding, but guide-coding is what I do. I prefer operating at a higher level of abstraction and I still want the opportunity to learn as I go.

This isn’t unprecedented in other creative domains:

Many of Andy Warhol’s paintings were done by other artists and by machines,
while he supplied the ideas and retained authorship.
Some famous California rock bands of the ’60s didn’t play on their own albums.
And some prolific authors just invent characters and story lines,
and leave it to other writers to fill out the prose.

That resonates with how building software feels for me now: authorship is increasingly defined by intent, direction, and judgment, not by who physically produced every brushstroke, guitar riff or line of code.

A recent chain of events went like this. I love learning about LLM technology and I’ve been at it for years. I recently came across Stanford CS336 and it does a great job lining up the research for anyone who wants to understand LLM evolution.

$ feh Calculus-3-2015.jpg

[brighten]

< >

I’m also a cheat-sheet fan. I like creating them because it’s my way of reviewing material I’ve learned. Going through the course, I could tell there was so much to understand deeply that I attempted to create a cheat sheet and very quickly realized I might as well be writing a book.

That seeded another idea: what if this became a website? CSS alone is enough to turn my curiosity into resistance, when it comes to making websites. Still, I started experimenting with what the site might look like. GenAI tools helped me craft the site, and I learned Svelte along the way. It became a living cheat sheet, something I could refer to and update whenever I wanted. I delegated much of the craft of creating the website.

Was the resulting website art? To me, yes. It serves my purpose. It reflects my taste. It embodies choices about what matters, what to emphasize, what to connect, and what to leave out.

I’ve always identified as an engineer: objective, systematic, focused on how things work. Never as an artist. But Rubin captures something important here:

Whereas the Experimentation phase is about what the seed has to offer,
now we are applying our filter. Reviewing the totality of our experience
in the world and searching for connections: What does this remind us of,
what can we measure it against, what does it relate to
that we've noticed over the course of our lives?"

That filter, that judgment, still feels irreducibly human. The AI can craft, but it can’t taste.

This might be a renaissance for software. I wanted a dictation tool; there are fifty of them. I tried ten, picked one, and it works beautifully. I wanted a feature for a word processor; I created (forked & edited) a word processor. That’s the world we’re entering: a world where more people can manifest their visions without being stopped by the long staircase of craft.