Unfortunately, Plastic Software

Plastic Land

In my last post I talked about the hopeful idea that the software industry might use more verification techniques; hardening our systems while using LLMs to our intellectual advantage. Unfortunately, I am now of the opinion that we have actually invented plastic software, and whether we use that for hardening our systems might not change at all, or perhaps more likely, be even less than we do now.

What do I mean by “plastic software”? With LLMs, we could use them to explore new ideas and enhance our own capabilities; e.g. if you have been solely a backend engineer throughout your career, like I have been, you could now dip your toes into some frontend work. Or like Kleppmann proposed in the article from my last post, use it to help formally verify your projects. But instead we have invented the means to mass produce software that is very brittle, at a glance looks correct, and is absolutely terrible for the environment.

We have invented plastic software and the polymer is Python and JavaScript.

Slime

Now, I don’t think this is a completely bad thing. I grew up in the 90s which, if I were to draw an analogy between LLM models and plastic production, could have been the ChatGPT 3.5 era of our times. Did you have a Creepy Crawlers kit growing up? They literally made molding plastic into a child’s playtime activity and I loved it.

On the subject of nostalgic toys from my childhood, I also remember having pressed-steel, metal, construction vehicles. And model trains.¹ I would hold one of these things and they would have weight. Set it down on a wood table and there was a thunk. I could actually use the metal excavator to excavate dirt in my backyard.

No matter what though the plastic is here to stay. Nowadays we have some pretty cool kinds of plastic. Almost every manufactured material has some kind of plastic in it. So, it’s here to stay.

Copper Versus PVC

When I explain to someone what I do as a software engineer I like to make an analogy with the trades and say I’m a “data plumber”. I do mostly data-related, backend work, so I think the comparison is fitting; and it certainly helps people conceptualize what I do, if at least superficially.

To stretch this analogy even further, as a data plumber, the materials that I do my plumbing with should matter. And if you were to hire this data plumber, would you prefer that I used copper or PVC to do my plumbing?

Now, I’m fully aware the correct answer is context dependent. And maybe that makes this analogy all the better. But, generally speaking, I think people would say copper–at least the folks who know little about plumbing (such as myself)–and I think the reason for that is because it is commonly thought of as the more durable option.

Really, that’s the main point. So I’ll happily stop this poor analogy there. We do not want to lose durability in our software systems. We can allow them to be brittle and fragile in certain areas, sure, but let’s not do that to everything. Like we did in the 90s.

Garbage In, Garbage Out

I’m going to assume the government is reading my drafts because while writing this article the US Department of Education posted this tweet.

How wonderfully appropriate and terribly disgusting is this? Here we have an AI generated image–the pinnacle of plastic these days–showing what appears to be two plumbers doing what I guess are plumber things. The image is full of inaccuracies and errors, but it wants you to think these two are doing plumber things with copper pipes; the good, durable kind of plumbing.

How on earth am I supposed to think this message is authentic? Have you ever gone to a costume store looking for your favorite Power Rangers outfit, only to find they just have cheap plastic masks? That’s how this post makes me feel.

Perhaps I wanted some actual guidance from the US Department of Education, but instead I got the cheap plastic knockoff. They want my attention, not to inform me of some great opportunities in the trades.

But that would take time, patience, and integrity.