Ignore the Grifters - AI Isn't Going to Kill the Software Industry

I feel like half of my social media feed is composed of AI grifters saying software developers are not going to make it. Combine that sentiment with some economic headwinds and it’s easy to feel like we’re all screwed. I think that’s bullshit. The best days of our industry lie ahead.

It’s highly unlikely that software developers are going away any time soon. The job is definitely going to change, but I think there are going to be even more opportunities for software developers to make a comfortable living making cool stuff.

Here’s your white pill. It’s all going to be okay.

Economics

The field of economics has a lot to say about automation, productivity gains, and how they effect the economy. There’s no shortage of people saying “this time it’s different”, but those people have been around for every other major technological advance and they have yet to be correct. I wouldn’t bet on the doomers.

Jevons Paradox

AI tools create a significant productivity boost for developers. Different folks report different gains, but most people who try AI code generation recognize its ability to increase velocity. Many people think that means we’re going to need fewer developers, and our industry is going to slowly circle the drain.

This view is based on a misunderstanding of why people pay for software. A business creates software because they think that it will give them some sort of economic advantage. The investment needs to pay for itself with interest. There are many software projects that would help a business, but businesses aren’t going to do them because the return on investment doesn’t make sense.

When software development becomes more efficient, the ROI of any given software project increases, which unlocks more projects. That legacy modernization project that no one wants to tackle because it’s super costly. Now you can make AI do most of the work. That project now makes sense. That cool new software product idea that might be awesome but might also crash and burn. AI can make it cheaper for a business to roll the dice. Cheaper software means people are going to want more of it. More software means more jobs for increasingly efficient software developers.

Economists call this Jevons Paradox.

Comparative Advantage

When we think of future AI, we imagine a system that can do everything better than a human being. The all-knowing AI is better at programming, art, diagnosing diseases, and coming up with new ways to make toast. How can any human compete with that?

Turns out, it doesn’t really matter. There are already lots of situations where people who aren’t the best at something still find work.

Imagine a small business owner who is absolutely the best at everything in their business. They know every aspect of their business and can do any job in it better than anyone they can hire. Regardless of their genius, they only have 24 hours in a day. If they want to scale, they will need to hire people so they can focus on their most important work. For example, the business owner may be a better bookkeeper than their accountant, but the opportunity cost of not working on other things justifies hiring someone to keep the books.

This is an example of comparative advantage. Comparative advantage is when a country, company, or person can produce something at a lower opportunity cost than others. Even if that entity is not the most efficient producer, they’ll still be able to sell their product as long as someone else has more important things to do.

While AI is powerful, it’s also computationally expensive. Unless someone decides to rewrite the laws of physics, there will always be a limit on how much artificial intelligence humanity can bring to bear. This means that we’ll eventually allocate our scarce AI resources towards the things they are best at, which leaves plenty of things for humans to do.

Solow Model of Growth

The Solow model shows that economic growth is a product of capital (factories, data centers, corporate relationships, land, etc…), labor, and technological progress. In the long run, the only reliable driver of economic growth is technological progress. Our society gets richer by learning new ways to deploy scarce capital.

Another way to think about this on a micro level is that you can’t sustainably make more money than the economic value you create. A person who digs a hole with a shovel is not going to be able to make as much money as someone driving a bulldozer.

Widespread adoption of artificial intelligence will greatly accelerate technological progress. This acceleration will create a massive increase in economic growth. This rising tide will create more resources for everyone. If you’ve spent any time following the e/acc community on Twitter, this is what they’re banking on.

The Solow Model is more of a macro level concept and doesn’t specifically apply to software development as an industry. While I don’t think software development as a career is going away any time soon, if it does, we’ll still probably be better off. It’s better to be a barista in Star Trek than a noble in Game of Thrones.

Widespread AI

While increased societal wealth is nice, what happens if all of that wealth is controlled by a handful of large companies run by a small cadre of tech elites that hoard all of the AI compute?

This is unlikely for a lot of reasons, but one of the big reasons is that AI innovators are committed to making AI accessible. While there are plenty of folks that offer access to AI-based tools, you can also DIY your own AI.

On the model side, many producers of models release them for free. While the top tier Open AI models are locked behind their walled garden, you can fire up Ollama or CivitaI and download lots of very capable models.

On the hardware side, Nvidia is also pushing local model execution via things like their Jetson Nano products and Digits mini computer. This doesn’t even include things like consumer video cards, or the M4 Mac Mini, which people have been using to make their own AI clusters. AMD is also working hard to make it easy to run local models.

There’s a huge incentive to make AI resources accessible to most people. If we amplify everyone by making it easy to access their own small army of personal assistants, innovation skyrockets. AI has the potential to enable millions of small creators to build sustainable businesses.

The Job of Software Developers

Now that we’ve covered a few macro reasons why we aren’t screwed, let’s look at some more down to earth reasons why we’re still going to need software developers in the age of AI.

The 70% Problem

AI models are fantastic at producing a wide variety of things. Sometimes the results are good and sometimes they’re weird. AI generated code is no different. I have yet to prompt an AI generated function that’s 100% correct. Gen AI is great for getting started, but it won’t write your whole app for you. This is occasionally referred as the “70% problem”, where AI can get you most of the way there, but it falls down at the end.

A lot of folks who don’t know how to write software and think they can prompt their way to success will get stuck in a loop where each prompt fixes one issue and causes two more. This is the 70% problem in action. You can generate code all day, but there’s no guarantee it’s going to be the right code.

Even if you get the AI to generate all of your code for you, it still needs to be tested, monitored, deployed, and maintained. Even if you get AI to write most of your code, there’s still plenty to do.

AI code gen also tends to fall down in complex enterprise systems. You can crank out cute demo apps all day long, but most systems don’t resemble cute demo apps. This isn’t much different than the Ruby on Rails 15 minute blog app scaffolding demos from back in the day. They looked cool, but it was only the first step.

Much of our job isn’t writing code anyway

The folks who think AI is going to wipe out the software industry don’t fully understand what software engineers do day to day. After 20 years, coding is honestly one of the easiest parts of the job for me. There’s lots of time spent talking with other engineers, understanding the business domain, figuring out the best approach, weighing options, and designing systems. Even if AI generates most of our code, we’ll still need to tell the thing what to write and that’s going to require software developers.

We have so much software left to build

The biggest assumption of the doomer crowd we’re close to having “enough” software and increased efficiency is going to quickly exhaust the project queue. I have yet to work for a company that doesn’t have a near infinite backlog of things they want to do. If software development moves faster, we’re just going to build more software.

What do I need to do?

Some developers think AI isn’t going to change much of anything and we should just sit tight and wait for it all to blow over. That view is just as short sided as the doomer side of the equation. Software development has always been a career where you are either learning new things or stagnating. AI doesn’t change the need to keep learning and evolving.

The best thing you can do is learn how the current stack of AI tools work. If you can use Copilot at work, go for it. Otherwise, spin up your free Copilot trial and build something. Beyond that, take some time to learn how AI works and think about other ways it can improve your workflow. Maybe get it to crank out some of that documentation you don’t want to write anyway.

Beyond that, keep learning about what it takes to build good software. Work on your architecture & design skills. Consider branching out into other areas in the stack like product development. While the value of raw coding skills might go down, the value of everything else around them is going to continue to rise.

The software developer job is going to change a lot in the next five years. The AI revolution is similar to the introduction of compilers. If you keep up and learn the new paradigms, then you’re going to be okay. Otherwise, you might find yourself in a bad place.

Things are probably going to get a little weird, but it’s not like you signed up to be a developer because it was the same thing every day. Crack the books, fire up your copilot enhanced editor, and make cool stuff.

Links

If you want a longer description of this topic, check out this article from an actual economist:
Plentiful, high-paying jobs in the age of AI

A good video series on the Solow Model:
Intro to the Solow Model of Economic Growth