AI and Software Craftsmanship

Or, an old guy whines about stuff

I’ve been working as a computer programmer for 50 years, and was paid pretty well for 39 of those years. It was good run, but I’m glad I retired in 2015, because it allowed me to work for peanuts running the catalog/circulation web sites for three Vermont libraries. (Pro Tip: you’ll never get rich working for libraries; they usually don’t have a lot of money.)

Nowadays I have another reason to be glad I’m retired: the use of AI in software production. According to AI proponents who inhabit Hacker News, you’re an idiot if you don’t use AI to churn out code, because the world is leaving you behind. Many companies now require programmers to use AI tools, and I’m glad I don’t have to work under those conditions.

During my decades in Silicon Valley, I learned that there were two ways for a programmer to get ahead in the corporate world. The first was to be a genius, which would allow you to become a Senior Staff Engineer, because then you could work on whatever you felt like. The second way, for mediocre brains like me, was to go into Project Management. If you took this path, you’d make a lot more money, but you’d never get to write or even look at code again, and you’d spend all your time in meetings and trying to get other people to do stuff for you.

Project Management sounded like a terrible career choice, so I just kept plugging away at writing code. At my last job, at a Very Large Software Company in Palo Alto, not being a Project Manager meant that I could do everything from assembly language (a step up from raw machine code) to web apps in Ruby On Rails. This was pleasing to me, because it allowed me to learn new things, which, more than anything else, is what keeps me interested in programming. It’s why I still write code, creating things for myself, not for money.

But AI proponents say this kind of work is all in the past, and we programmers should realize that resistance is futile, and that we should join the Borg. One of the most vocal of these evangelists is Steve Yegge, a famous programmer who achieved some of his fame by very vocally quitting highly-paid jobs at big companies like Google and Amazon. Yegge says we should stop looking at our code (something I do every day) because AI-generated code is like a fire-hose spraying thousands of lines of code every day. Yegge says that in a six month period last year, his AIs were able to spew out a hundred thousand lines of code. No mere human could possibly keep up with that!

So Yegge puts together what he calls “AI orchestrators”. He has disdain for those low-life programmers who just let the AI write the code without any guardrails. Those idiots are doing it all wrong! You need multiple AIs! You have one AI looking at (and probably stealing) other people’s code and trying to understand it. You have another AI creating specs based on the first AI’s results, plus a few prompts from you, the human. Then you have another AI to check the second AI’s specs for errors. Then you have another AI try to write code based on the now-checked spec. Then you have another AI looking for bugs in the newly generated code slop. I’m sure I’m leaving out steps here, because I’m not a genius like Steve Yegge.

Anyway, it all comes together like magic, and all you have to do is manage all of these AIs somehow. So of course, Yegge (or rather, the AI he uses), has created some kind of orchestrator to do this for you. This sure sounds like those Project Management jobs I managed to avoid for decades, with the additional benefit of using colossal amounts of energy to power the AIs doing this job.

In order to not sound like an unbalanced crazy person, Yegge talks about how you can get burned out being so darned productive this way, but it’s all good because Productivity!

So now we come to the real reason why we’re supposed to abandon our old ways of writing software, and join the Glorious Revolution: it allows you to be more productive! Instead of writing a few lines of carefully considered and crafted lines of code per day, you can churn out many thousands of lines of code. Yegge is not the only AI proponent making this point. I’ve seen plenty of other Hacker News commenters talking about how they can write tens of thousands of lines of code in a few hours.

But why do we need to be so terrifyingly productive? This question doesn’t seem to be very important. The AI Psyop says that you need to jump on the bandwagon or be replaced. This was was stated plainly on a billboard I saw in San Francisco a few weeks ago: “AI will replace you unless you use it.” So apparently you have to use it, because everybody else is using it.

What this means for programmers who follow this path is that not only they will stop writing code, but as Steve Yegge dictates, they will no longer even look at code. Trust the AI! That’s the message.

So human programmers will lose the skills they honed for decades. This is not the same situation as using compiler to write code in a more expressive language. It’s a complete abandonment of the craft. It’s as if I were a music composer, and I gave up writing original music, and instead asked an AI to “write music in the style of Chopin.” Maybe the resulting music would sound vaguely like a Nocturne or a Ballade, but it would be missing the brilliance and originality of late works like the Barcarolle, that could not have been predicted by Chopin’s earlier works. This is not to say that any code I’ve written was brilliant, but at least a lot of it was original.

We’re supposed to applaud this change. But old farts like me who lament this state of affairs are, as Steve Yegge says, just in the first stage of grief, and we are stupid to be stuck there.

It will be interesting to see how this turns out, but I’m pretty sure it will not end well. I see a future where software built by AIs becomes so bloated and buggy, due to the fact that it’s being trained on the slop that it’s already produced, that we’ll pine for the good old days of the dot-com bubble. For me, “Trust the AI” is about as sensible as “Trust the Science™” was during the Great Covid Psyop.

But the Crappification of Software is only one aspect of the Glorious Revolution that alarms me. Please read Gabe of Libre Solutions Network and V. N. Alexander (no relation) of The Posthumous Style for insight into the bigger picture of the changes that AI is inflicting on our world. They also offer some hope for an alternative to the AI dystopia, and goodness knows we need some of that hope right now.