Does the Future Need Programmers? Part 1

All of them, or none

There’s a common complaint that goes something like this:

“Software companies are no longer hiring junior programmers because of AI. But they’re shooting themselves in the foot, because they still need senior programmers. And where do they think the next generation of senior programmers will come from if there are no more juniors?”

There are plenty of good reasons to worry about AI and jobs, but this isn’t one of them. If AI’s impact on the job market is relatively benign, junior programmers will be more in demand than ever. And if AI eats the market for junior programmers, I fear it’s only a matter of time before it comes for the senior programmers also.

Both scenarios are plausible, although over a five to ten year horizon my money says AI will eat its way to the very top of the programming profession. Part One of this piece explores how software development is already changing, and Part Two maps out the dynamics that will determine which way the industry goes.

I’m focusing on the cutting edge of software development because that’s where we can most clearly see how AI will impact employment. But that’s just the beginning: the dynamics we see there will quickly spread within a few years, first to the whole of the software industry and then to most white collar professions.

Are junior programmers losing their jobs?

There’s some controversy about whether AI is actually destroying entry level programming jobs. The data are confusing and I don’t think it’s possible to say definitively whether or not we’re seeing the early stages of a major disruption. Adding to the confusion, AI is frequently used as an excuse for layoffs that are happening for mundane business reasons.

There are early signs that junior developers are becoming less useful at the most forward-looking companies, even if that hasn’t yet resulted in significant cuts. Anthropic’s Jack Clark puts it very diplomatically:

Something that we found is that the value of more senior people with really well-calibrated intuitions and taste is going up, and the value of more junior people is a bit more dubious.

Since there are no answers in the employment data, let’s look at how programming itself is changing.

What do programmers do all day?

Perhaps the strangest thing AI has done to programmers is to upend our understanding of what we do for a living. A year or two ago, if you’d asked what we do, most of us would probably have told you “I write code”. And yet today, many of us no longer write any code at all. Nobody ever got paid to write code: our job is to create useful software, and that’s as true now as it ever was. But what that means has changed profoundly.

We often say that every programmer is now a manager, supervising multiple coding agents. That’s a great description of how it feels, but it doesn’t help us think about what’s coming next. For that, it’s helpful to think of programmers as having two jobs:

• Coordinating with other people. We spend much (often too much) of our time writing status reports, working with designers, and reading & writing specs. AI hasn’t really changed this part of the job, although that’s starting to change.
• Building software. “Writing code” is just one part of building software. Creating useful software involves designing high-level architecture, choosing libraries, crafting interfaces between components, and ensuring that the entire product is reliable and maintainable. This half of the job has been completely transformed over the last year.

Programming has changed over the years: we’ve gone from handcrafting assembly code to working in high level languages, and the scope of our ambitions has expanded as our tools have improved. But until 2025, we had always spent most of our time writing code. AI completely changed that equation: most cutting edge programmers now write little to no code, focusing instead on architecting and reviewing code that is written by AI.

That’s great news if you’re an experienced, ambitious developer: instead of writing code line by line, you can focus on high-level architecture, telling teams of agents how to build your product for you. You can produce far more (and better) software than you ever could before, and you are therefore more valuable than ever before. But what if you don’t have a decade or two of experience building software? As a junior developer, can you find a way to be useful, or are you just getting in the way of the senior developers and their robot armies?

I see two possible futures: in one, junior developers also experience large productivity gains, and their job prospects are better than ever. But in the other, it becomes clear that junior programmers are just getting in the way—they quickly become unemployable, followed soon after by their more experienced colleagues. Three dynamics will determine which way the industry goes:

• Will the coordination part of the job become a bottleneck that only humans can do?
• Can we teach high-level software-building skills the same way we teach programming?
• Will coding agents reach a capability limit where they can augment but not automate the work of senior developers?

We’ll tackle those in Part Two.