Devs are Fine

I don’t think AI is going to have an especially large impact on the total number of people engaged in software development.

Hot take, I know, but let me explain.

For the past couple weeks, there has been an explosion of various predictions that amount to “most software engineers will be out of work in [x] months/years.” Software development may change dramatically, but software is an unusual good in that there is infinite demand for it.

For most things in an affluent economy, making them cheaper does not linearly create more demand. If cars were, say, an order of magnitude cheaper, there probably wouldn’t be many more of them on the road. There likely would be a small increase, but most people who want a car have a car, even if they’d perhaps like a fancier one. If someone gave me a car tomorrow, I’d sell it; I don’t need another car and don’t have a convenient place to put it. That’s true with clothing (I only have one body), food (I don’t need more than a few thousand calories every day), houses (I can only be in one place at a time)—most material goods, really.

Following from that, if we automate 50% of a given job, then ~50% of workers are likely permanently out of that specific job. Sure, the economic theory will tell you that the economy will expand as productivity increases, and over time, that labor will get reabsorbed elsewhere, but those specific jobs are lost forever.

I’m not sure software works like this, because I think there’s effectively infinite demand for it. It is more like energy—an input that affects and improves almost every human activity.

The software industry as a whole is not going to change nearly as much in terms of total employment as common wisdom has it today.

Talk to anyone who knows how to code, even at a basic level, and we all have little random programs and scripts doing things for us. I would love more software! I can come up with a hundred things that I’d like to automate, but even with coding agents, I’m never going to get around to most of it. I probably wouldn’t even think of a lot of potential applications, but I’d happily purchase them.

There are entire businesses out there that run on Excel-as-a-database or use software from companies like Oracle that are stuck in the 1990s. Many small businesses’ entire web footprint is still via Facebook. Entire industries—probably most—are woefully under-digitized. The amount of improvement possible with better, faster software development is simply enormous.

Unlike information, software doesn’t necessarily demand attention, either. I can only read so many books in a lifetime, but software can actually reduce attentional needs by automating things. It takes up no physical space.

In most industries, making workers more productive means slimming down the labor force. In software, I think it just means we’re going to get more software. The specific type of job might change—what we think of today as a software engineer is probably a lot different from what it will be in a few years—and certainly some people may see their specific skills obsoleted, but I don’t think the software industry as a whole is going to change nearly as much in terms of total employment as common wisdom has it today.

Finally, we are at the very early stages of building great software. Most of it is really not very good (unlike, say, cars and airplanes, where we have probably reached engineering asymptotes). So, there’s also enormous room to improve the quality, not just the quantity.

Now is a fantastic time to begin building new companies leveraging tools to build more goods and services. The benefits of these coding agents inure to smaller, more nimble companies that can hire AI-first employees willing to exploit those tools to the utmost.

SWEs, get to it! Lots of work to do. Even better, start a company then call us.