The Real Threat to Programmers Is Not AI — It Is Standing Still

This article is part of Beyond the Code, Computhink Academy’s quarterly series featuring perspectives from the frontier of technology, AI, and the future of learning.

In response to Anthropic’s article, Labor market impacts of AI: A new measure and early evidence

https://www.anthropic.com/research/labor-market-impacts

Key Takeaways

• Programming jobs are not disappearing. They are being restructured, and the restructuring is accelerating.
• The programmer’s role is shifting from writing code to designing systems, guiding AI, and owning the quality of the result. That makes the job harder, not easier.
• Keeping pace is now a core professional skill. Agents, MCP, and new AI frameworks, the landscape changes almost weekly. Programmers who stop learning will fall behind fast.
• Learning to code still matters deeply, not just to build things, but to evaluate what AI builds. Without that foundation, you cannot judge whether the output in front of you is correct, secure, or fit for purpose.
• More people will need programming literacy, not fewer, as AI enables non-technical teams to build their own tools and automations.
• For Computer Science students, the real advantage lies in strong fundamentals combined with staying current. Generating code quickly is the easy part. Thinking clearly about systems and never standing still is what sets you apart.

Anthropic’s recent labour-market analysis found that Computer Programmers show one of the highest levels of AI exposure, with a large share of their work linked to tasks such as writing, updating, and maintaining software programs. Many will read this and conclude that programming jobs will disappear. I think that conclusion is too simplistic.

My view is that programming work is not disappearing. It is instead being restructured.

In the past, if a business unit needed a system, a dashboard, an automation flow, or an internal tool, they often had to go through an IT department or an external vendor. Today, AI is changing that work model. Some types of software can now be created much more directly by the business units, business analysts, operations teams, or domain experts who understand the workflow well and know how to work with AI tools.

That does not make programming irrelevant. In fact, it raises the bar for computer programmers.

The value of the programmer is shifting away from simply writing code line by line. The modern programmer must increasingly define the problem clearly, design the solution architecture, guide the AI to generate usable code, validate outputs, test edge cases, ensure maintainability, and produce reliable documentation. In other words, coding knowledge still matters, but it is no longer enough on its own.

This is only the beginning of what the programmer must now master. Building systems today increasingly means building systems with AI Agents. Understanding how to design agentic workflows, leverage new knowledge like Skills and MCP (Model Context Protocol), and integrate tools that did not exist six months ago. This is a continuous re-learning of how systems are built.

Programmers in IT have long accepted (or complained!) that the job requires keeping up with new frameworks, new languages, and new system architectures. That was always a common complaint, and a fair one. But the pace has fundamentally changed. It is no longer a matter of learning a new framework every few years. Today, something new and worth trying emerges almost every other week. Agents, orchestration patterns, memory systems, context protocols, the possibilities are compounding faster than any one person can fully track. For programmers who lean into this, it is an extraordinary time to build things that were simply not possible before. For those who do not, the risk of irrelevance is real.

Beyond keeping pace with new tools, the programmer’s responsibilities have also grown more demanding in a less obvious way. Getting AI to generate code is not the hard part. The hard part is everything that comes after: verifying that the output is correct, checking that it is secure, confirming it meets specifications, and taking full accountability for the final system. The programmer who understands the underlying logic can do this well. The programmer who cannot is flying blind. In that sense, the job has not become easier; it has become harder in the ways that matter most.

This is why I believe the real risk is not that all programmers will be replaced. The real risk is that programmers who remain only code writers will become vulnerable. As AI takes over more routine implementation work, companies will place a higher premium on those who can combine programming knowledge, systems thinking, business understanding, judgment, and accountability.

AI may simplify parts of software development, even as it makes the role of the programmer more demanding. Less effort may be spent on actually writing code, but more will be expected in solution design, decision-making, integration, quality control, and overall speed of execution.

The future programmer is not just a coder. The future programmer is an AI-enabled solution builder. And the one thing that future cannot accommodate is standing still.

Then what does this mean for young students learning programming today, or university students considering a Computer Science degree? Does the rise of AI mean there is no longer value in learning to program?

I would argue the opposite. Learning to program was never only about producing code. At its core, it is about learning logic, critical thinking, and computational thinking. It teaches students how to break problems down, think in steps, test assumptions, and build solutions systematically. These are valuable skills regardless of the field a student eventually enters.

Learning to code is first and foremost about learning the language of logic and structured problem-solving. This foundation matters now more than ever, because it is precisely what allows a person to evaluate whether AI-generated code is actually correct. AI can produce code quickly. But without the underlying ability to reason about what the code should do, how it should behave at edge cases, and whether it is secure, a developer cannot judge the output in front of them. The ability to check the AI’s work is not a bonus skill. It is the skill.

Even if AI becomes increasingly capable of generating code, students still need to know how to think clearly, evaluate outputs, and judge whether a solution actually makes sense. Without that foundation, they may be able to generate something quickly, but they will struggle to know whether it is correct, reliable, or suitable for the real problem at hand.

I would go further and say that more people should learn some form of programming literacy, not fewer. If AI allows business units and non-IT workers to create smaller systems, automations, and internal tools without relying fully on software engineers, then programming knowledge becomes useful to a much wider group of people. It becomes less of a narrow specialist skill, and more of a practical capability that helps people work more effectively in their own domains.

For students who still want to pursue a Computer Science degree, the importance of learning to code properly has only increased. It is not enough to rely on AI to generate code. Students must still be able to write good code, understand why it works, and recognise when code is inefficient, insecure, or badly structured. More importantly, they must develop the ability to recognise and curate good code and take responsibility for the final result.

But learning to code well is only half the equation. The other half is staying aware of how the industry is changing, and being willing to keep pace with it. The same imperative that applies to working programmers applies to students: do not stand still. A Computer Science degree built entirely on yesterday’s tools and yesterday’s workflows will not be enough. Students who combine strong programming fundamentals with genuine curiosity about what is emerging — Agents, MCP, new AI development patterns — will enter the industry with a real advantage.

The temptation to skip the hard work of learning to code properly, and simply use AI to generate everything, is understandable. But it is a false shortcut. Students who take it will lose the very thing that makes a Computer Science background valuable: the ability to reason about systems, judge the quality of code, and take ownership of solutions. That is not something AI gives you. It is something you build through learning.

So no, the rise of AI does not mean there is no value in learning to program. If anything, it makes that value clearer than ever.

For younger students, coding remains one of the best ways to build logic and problem-solving skills. For non-technical learners, it becomes a powerful modern literacy. For future Computer Science students, it is the foundation for higher-level work in design, judgment, and technical decision-making.

The future will belong to those who can think clearly, use AI technology wisely, and who never stop learning.