I sit here, watching in awe and horror as Opus 5.5 rebuilds one of my favorite childhood games from scratch in the browser. A quick back-and-forth on direction, art, and style. An outline on how to record/break down work and manage the project via issues and worktrees. It keeps churning, closing PRs, and managing the work like a small, artificial development team. I have 3 developers on this little team, a limit imposed by my rather restrictive $20-a-month token plan, but still… along it goes. Building.

That’s what I like to do. I like to build. I like to make things. I like to take what feels like nothing, a simple text file, and turn it into something. It’s what I’ve loved about computing ever since I got my first hand-me-down brick of a laptop at age 10. I’ve always wanted to build real computer applications. It’s something I found myself good at. I was told in college that the best programmer isn’t necessarily the smartest person in the room, which was perfect for me. I’m average. Sometimes slightly above on certain things, sometimes slightly below. But working with computers has always been my forte.

Now I sit here watching a machine, a complex auto-complete algorithm, the dumb thing that used to suggest the wrong word on my phone, do the thing that I am good at. At least I thought I was good at it. This thing has a depth of knowledge that no human brain can ever comprehend or understand. It’s still nascent, it still makes mistakes, and it still doesn’t understand certain nuances or understand exactly what looks good or bad, but it’s certainly getting better than it used to be. And that’s something I thought was going to take a long, long time. I didn’t expect it to keep progressing like it has. It’s probably dumb, because trillions of dollars are being spent to try to accomplish this goal. To make the statistical averaging machine average higher and higher. And deep down I knew that this would happen. I knew it would take the work I did and make the thing I loved to do trivial. I don’t love dealing with people. I love dealing with the machine. Understanding how it works and the intricacies and the infinite depth of knowledge that I can strive for but never attain. Why does it feel like that’s been taken from me?

I can still go on, exactly how I’ve been going on in the past. Nobody at my work is pushing me to go faster. Nobody is saying I have to use AI to build things at work. On the contrary, I work for a small company, so AI spend is not an insignificant expense. It’d be easier if we didn’t have to pay for tokens anyway, but it almost seems irresponsible not to. There are so many places where having an agent build things just makes more sense. Even when I just need it to go scaffold this small piece, or that subsystem, or write unit tests for me. It has a place in my work, but nobody’s making me do this.

The Past

We’ve seen revolutions in the computing industry before. Compilers revolutionized how people programmed computers. We used to have to be aware of timing on drum memory in order to optimize programs so they didn’t take an enormous amount of precious runtime. Our field is one of revolution after revolution. All the way up the stack. We as a profession have harnessed the machine to accomplish great (and often dumb) things. We went from physical switchboards, to paper tape and punch cards, to magnetic storage and terminals talking to old big iron mainframes, then to the minicomputer era, then to microcomputers and smartphones. All of these advancements have changed the industry over and over and over again. Nothing has ever stood still in this industry.

The Reality

DHH got on stage recently and said that we’ve hit the point where it’s no longer economically viable to write code by hand anymore. And you know what? He’s right. It doesn’t make any sense. And it leaves me at a crossroads. Do I embrace this new technology and let it multiply my output and never write a line of code again? Do I build a dark software factory, where I have no idea about the underlying architecture and how the thing actually works? Do I swear off AI entirely and only write artisanal handcrafted code? Was my code any good to begin with? Do I have enough “taste” to make it in this new world?

I’m grappling with all these questions and trying to find an answer.

Maybe I can have my cake and eat it too. Maybe I can build systems by hand and gain competency. It’s not like I was writing novel, innovative code at work before; I was building things that all smelled the same. They all had similar shapes and formats. Maybe I don’t lose a thing by having an AI write code and do the grunt work for me, and maybe while it does all that cookie-cutter work, I can read docs, learn new languages, learn how to write and read assembly, and learn about how distributed systems should act, all while my production code gets stamped out on the cookie-cutter machine. Maybe I can spend my time looking for the patterns and optimizations that will make the systems I build actually great. Maybe, just maybe, AI buys me the time to actually dive deep and strive for that technical excellence that I so deeply crave. I want to be more competent, not less.

The Future

The outlook here might be bleak. AI might eliminate us all. It might turn the world into a pile of paperclips. They stole the entire corpus of human knowledge to build it. It hallucinates, even now, and it won’t stop. It centralizes power and leverage into the hands of a few large companies. These are very real concerns. And they absolutely deserve to be addressed. But it’s evident that we aren’t going to slow down. We’re throwing caution to the wind, and we’re forging on ahead. For better or worse, we are on the path to push this technology to its limits, and it feels like, as a society, we’ve barely scratched the surface of what this thing can actually do. There hasn’t been enough time for any of this to really bake.

At every point in history, at every revolution in computing, we’ve seen more and more programmers. It’s gotten cheaper and cheaper to produce software, making it more and more viable for people to develop software. We don’t need to understand logic gates and binary to build software; we haven’t for quite some time. Is it a good idea to learn the underlying primitives? Absolutely. That helps you understand what’s actually happening inside the machine. You should learn to read the code. Everyone should read the code these things output, but it’s not an absolute requirement anymore. There will be more programmers in the market than ever before. Salaries may change, but the job isn’t going away. There’ll be jobs up and down the stack too, from people able to read and optimize x86-64 assembly to people who don’t even know what a boolean in a programming language is. There’ll be a whole spectrum of developers. And there’ll be shifts: people may move up or down the stack depending on competencies, but the job isn’t going anywhere.

I Can Have the Best of Both Worlds

I think I might just be able to have my cake and eat it too. It’s not like what I was doing at work was revolutionary or novel. Work needs me to meet business requirements, and those requirements are similar to what other companies need. It’s not that complicated. But using AI to build those cookie-cutter pieces might just let me go deep figuring out how to optimize the stack properly. Without a backlog, why don’t I minimize compute costs? Optimize the storage requirements for our software? Assess how to make it faster? Make the system more secure? Make deployments happen faster. Strive for the technical excellence that I so crave. And maybe read a book on design or two to develop some “taste.” Maybe this is an opportunity to grow, not to lose skills. Maybe it’s an opportunity to become more competent, not less.