How to Not Lose Your Engineering Job to AI

This is the original article from 2024. In 2026, I revisited what held up and what changed in Part Two.

A couple of years ago, AI could barely write a simple HTML and CSS webpage without making mistakes. Today, it can help write a functioning web app, solve problems in your enterprise codebase, and answer nuanced questions about algorithms. In just a short time, it’s moved from barely useful to becoming a super-powered junior developer. And it’s improving fast—so rapidly that it raises questions about the future of traditional development roles.

If we had to put a skill level on it, I’d say today’s AI is better than a junior developer at writing code and understanding tasks. It's almost at the level of a solid mid-level developer with a few years of experience under their belt—sometimes even better when it comes to sheer speed or breadth of knowledge.

But the trajectory is what’s startling. The same AI that’s now almost as good as a mid-level developer was barely competent two years ago. Context sizes—the amount of information these models can hold in mind at once—are scaling up. Two years ago, 4K tokens was considered a lot of context. Today, models are handling millions. Soon, it won't be crazy to think that you could fit an entire codebase in one prompt. Right now, tools like Cursor or VSCode Copilot bridge that gap with clever retrieval techniques, but soon that bridge won’t be necessary. The AI will simply know your whole codebase. This will allow the AI to develop a deeper understanding of code dependencies, reducing second-order errors when making changes. It will also be able to quickly identify redundant or unused code throughout the entire codebase.

This presents a stark reality: The days of being valuable purely for your ability to crank out code are numbered. For example, Klarna's recent AI assistant, powered by OpenAI, has been operational globally for just one month and has already handled 2.3 million customer conversations—equivalent to the work of 700 full-time agents. The increased productivity is undeniable. Similarly, I find myself doing in one day what used to take me two or three. It’s going to have an impact, either by allowing companies to achieve more with fewer developers or by raising the bar for what’s expected of each developer. The efficiency gains are real, and they’re only going to increase.

So what does this mean for software engineers? How do you stay relevant in a world where AI can do much of the coding work?

The answer lies in shifting your value away from just writing code. Here are some of the places where humans are still needed, and likely will be for a while:

Keeping the Lights On

AI can generate code, but it’s not good at managing infrastructure—at least not yet. DevOps, ensuring correct system deployment, and maintaining smooth operations still require hands-on oversight, quick problem-solving, and a deep understanding of context—tasks that AI struggles with. This is mainly because these responsibilities span numerous services across different domains, where AI lacks overarching access and context.

Reviewing AI-Generated Code

AI might write a lot of code, but someone needs to review it. Bad code, security vulnerabilities, or even malicious bugs can slip through if there’s no human in the loop. Your role as a code reviewer is crucial for ensuring the quality and safety of what’s being produced. Think of yourself not just as a coder, but as a quality gatekeeper.

Directing the AI

The ability to clearly define a problem, guide the AI towards an acceptable solution, and iterate on that solution requires skill. You need to understand the architecture of the system you're building and make sure the AI is implementing things in a way that aligns with that vision. It's your job to make sure the AI is solving the right problem, in the right way.

Translating Business Requirements

Perhaps the most important skill to develop is the ability to act as a bridge between the business and development. AI might be able to write code, but it can't understand the needs of the business the way a human can. If you can translate vague business requirements into clear specifications that an AI (or a human team) can execute on, you’ll always be in demand. Companies need engineers who understand not just how to build, but what to build and why.

The quality of AI output is often dependent on the quality of input. If you can create great specifications for what the AI should do, you’ll get better results. This means understanding the problem deeply enough to give precise directions, breaking down complex tasks into manageable components, and anticipating the edge cases and pitfalls the AI might encounter.

Moving Forward

The future of software development is one where humans are replaced, just partly by a role change. Coding is becoming a commodity skill. The real value lies in what surrounds the code: the ability to understand the problem, to direct AI effectively, to ensure quality and security, and to bridge the gap between business needs and technical solutions.

It’s worth noting that this is not a new phenomenon. Programming languages themselves were an abstraction that made coding more accessible and efficient. We went from assembly to higher-level languages like Python or JavaScript, and the job of the programmer shifted each time. The rise of AI is just another step in that evolution—just a really big one.

If you're a software engineer wondering how to stay employed in the coming years, the answer is this: Don’t be just a coder. Be an architect. Be a communicator. Be a problem-solver. The AI revolution in software is coming, and it’s bringing incredible tools that will change what it means to be a developer. Embrace it, adapt to it, and find your new role within it. The engineers who do will find themselves more valuable, not less.