Intro
About two years ago I wrote about how AI will replace programming. Back then I predicted that most programming jobs would look fundamentally different in ten years. I still believe that. But now I want to get more specific.
I’ve learned that predicting the future is extremely hard. And I am especially bad at it. That hasn’t changed. But here are my bets anyway. Today is 2026-03-18. Let’s revisit in a few years.
Teams Will Shrink
We’ll see a lot more PM + 1 Engineer teams. When AI handles most of the coding, you don’t need five engineers to build a feature. One strong engineer paired with a product manager who knows the domain can move remarkably fast.
AI will also disrupt DevOps. Infrastructure-as-code, CI/CD pipelines, monitoring setups — all of this is highly repetitive and pattern-based. Exactly the kind of work AI excels at. DevOps as a separate discipline will shrink significantly.
Writing Code Will Become the Exception
Most code will be generated. The engineer’s job shifts from writing code to reviewing, curating, and maintaining it. Making sure that generated code fits the quality bar and stays maintainable — that’s the core engineering task going forward.
This is a bigger shift than it sounds. Writing code and reading code are very different skills. Engineers will need to become excellent reviewers, not just excellent writers.
The Human Tasks That Stay
Doing the right thing will stay as relevant as before. AI can produce code at incredible speed. But producing the wrong thing fast is worse than producing the right thing slowly. Understanding what to build and why — that remains a deeply human task.
Breaking down larger chunks into workable pieces of software and increments will stay just as critical. And let’s be honest — we are often very bad at it. AI won’t fix that for us. Decomposition requires understanding the business, the users, and the constraints. No AI assistant can do that for you.
Having the big picture in mind is a human task. Knowing what the AI produces, how it fits together, and where it’s heading — that requires human judgement and intervention. The moment you lose oversight of what the AI generates, you lose control of your system.
Simplicity Wins
Simplicity will reign big. One codebase, easy to build, easy to understand, well tested. Companies that operate this way will thrive with AI assistance.
Complex setups will lose. Multiple repos, hard-to-replicate environments, byzantine build systems — these will become liabilities. AI tools work best when the context is clear and contained. If your setup is already hard for humans to understand, AI will make it worse, not better.
The Market: 80/20 Rules
The 80/20 rule is real and it will intensify. We’ll see big winners and a lot of small, unnoticeable companies. When the cost of building software drops dramatically, everyone can build. But not everyone can distribute.
We’ll see a lot more choice in products. The barrier to creating software drops to near zero. That means more niche products, more experiments, more competition.
But Microsoft, Google, and the other incumbents will still dominate. They have distribution, data, and infrastructure. 80/20 will stay true — even in the times of AI. Maybe especially so. The big players will adopt AI faster and more effectively than anyone else.
Demand for Software Engineers Will Rise Again
In 2026, the job market for software engineers is brutal. Finding a new role is difficult, employers have more leverage, and many companies are laying off portions of their engineering staff.
My expectation is that demand will rise again after this initial slump. As software becomes cheaper and faster to produce—especially with new tooling—organizations tend to build more of it, not less. More software in production still creates substantial needs: integration, reliability, security, compliance, and ongoing maintenance.
This dynamic resembles Jevons paradox: when efficiency improves, overall consumption can increase rather than decrease. If software creation becomes more efficient, we will end up deploying software in more places—raising total demand for software engineering work over time.
Conclusion
These are my predictions as of 2026-03-18. Some of them will be wrong. But I’m fairly confident about the direction: smaller teams, less hand-written code, more emphasis on judgement, simplicity, and doing the right thing.
The engineers who thrive will be the ones who can think clearly about what to build, break it down well, and keep the big picture in mind while AI does the heavy lifting.
More
- AI Replaces Programming — my earlier take from 2024
