Let’s address the question once and for all. Is coding really dead?
Hype is everywhere. Influencers are making bold claims. CEOs are dropping sensational headlines. It’s enough to discourage anyone, especially beginners just starting their journey. You’ve probably seen the headlines: Anthropic’s CEO suggesting AI will replace software engineers within 12 months, or Nvidia’s CEO predicting the death of coding. Blog posts declare that “code is cheap,” and even the creator of Node.js has stated that the era of humans writing code is over.
If you’re feeling discouraged, it’s completely understandable. But before you abandon your path, I want you to step back and think with me for a moment. This is a perspective from a developer with 13 years of hands-on coding experience, who is also a top-tier user of these new AI tools.
The Two Halves of Software Engineering
To understand the future, we must first understand the present. Software engineering, in almost any company or context, can be broken down into two fundamental parts.
First, there is the thinking part. This is the “what” and the “why.” From a business perspective, this involves identifying customers, defining the service, and outlining the entire process. What is the company about? What problem are we solving? This isn’t about code; it’s about strategy, vision, and context.
Second, there is the implementing part. This is the “how.” In our world, we call this coding. It’s spinning up your IDE and translating the solution into lines of code.
This division isn’t unique to software. Imagine you want to transform your life—earn more money, live in a better home, or get fit. You can’t just start doing something randomly and hope for the best. You must first think. You have to create a plan of action. What will you do over the next few months? Only then can you begin the work of implementation.
This thinking part is deeply contextual. A personal goal has personal context. A business problem is steeped in business context: How much capital do you have? What is your vision for the world? What kind of company are you building?
Where AI Shines (and Fails Miserably)
Large Language Models (LLMs) are becoming incredibly proficient at the “how.” They can reason through implementation steps and generate code. This is where the panic comes from. But most people are missing a critical detail.
LLMs are extremely bad at reasoning about the what or the why.
They fail here because they lack the infinite context of your life, your business, and your specific goals. An LLM doesn’t know why you, as a software engineer, are building a particular feature or why the company needs it.
A Test You Can Run Yourself
Open your AI coding assistant of choice. Now, give it a truly absurd task.
Let’s say your company has an internal analytics dashboard. Give your AI this prompt:
“I think it would be a great idea for people to also be able to listen to music while they are browsing our internal dashboard. I want to implement a music player inside the dashboard where people can think of any song and AI would generate it on the fly.”
I can bet you that every single model will respond with something like, “This is a great idea! Listening to music can increase productivity. Let me go ahead and do that.”
The AI will then do a fine job of figuring out how to implement this feature. It will explore the codebase and start building. But it will never stop to ask, “Why are you even doing this? This is a stupid idea for an internal tool that nobody will use.”
It simply doesn’t have the context to make that judgment. This extreme example is often indistinguishable from a real-world task that might seem reasonable to a junior developer but is obviously a waste of time to a senior.
Your New Job Description: The Thinker
Your role as a software engineer has not been eliminated; it has shifted.
Previously, your job might have been 50% thinking and 50% implementing. The scales have now tipped dramatically.
The thinking part is now 70% of your job, and implementing is 30%.
This doesn’t mean you do less work. It means the nature of your work has changed. The time you spend on manual implementation has been compressed by AI. Your value is no longer in your typing speed but in your thinking speed and depth. You must get better at critical thinking.
The “what” and “why” are entirely on you. The “how” is something you now supervise. And you must supervise it closely. In my experience, LLMs take a suboptimal approach about 95% of the time. They don’t choose the best possible solution for your codebase. You can suggest a better approach, and it will implement it, but it won’t get there on its own.
Answering Your Biggest Fears
Let’s tackle some common questions head-on.
Should I stop learning to code?
Not at all. But you must change your focus. It’s no longer about memorizing syntax. Your job is now to:
- Read and understand AI-generated code.
- Read your peers’ code.
- Catch the AI when it generates bad, inefficient, or insecure code.
- Develop an inside-out, expert-level understanding of your product and systems.
You have to be skilled enough to know when the AI is confidently wrong.
Can AI really build production systems?
This is partially true, but also a bit of a cope. AI alone cannot build a robust production system from scratch without guidance. However, a skilled developer who properly steers the AI can generate code that is ready to ship. It’s not true that every line of AI code needs to be rewritten. But it is true that it needs to be expertly guided.
What about my job security?
You should not be worried about your job security… IF you are learning to use AI.
You should be worried if you are lazy, not closing your knowledge gaps, and not striving to be a good developer. You know in your heart if you’re slacking. If you lose your job for any reason but are a good developer who understands the “what, why, and how,” you will find new employment in no time.
This is one of the best times to be a developer. The demand for good people—people who can think—is astronomical. The implementation part has been reduced from 50% of the job to maybe 5%. That’s why the emphasis is on engineering, not just coding.
What skills should I focus on?
First, get exceptionally good at whatever you are doing right now. You can’t supervise the AI if you’re bad at coding yourself. Fix your knowledge gaps. Beyond that, focus on these areas:
- System Design: AI is terrible at this because it lacks the holistic business context.
- Debugging: This will be one of the last frontiers for AI. It requires stitching together information from logs, telemetry, and multiple services—a task that demands deep critical thinking.
- Communication: Get good at communicating your ideas, not just to leadership, but to the AI itself. Learn to write clear, precise product specs. If you’re an introvert who dislikes communication, it’s time to change.
- Critical Thinking: This is the foundation of everything. Solve math problems, study logic, do whatever it takes to become a better thinker. If you can’t do the “what, why, and how,” someone else will, and the AI will handle the implementation.
Is a computer science degree worthless?
A big, fat NO. A good CS degree is incredibly useful. It provides the prerequisites—the fundamentals of networking, operating systems, data structures, and algorithms—that you need to build higher-level knowledge in areas like system design.
The world isn’t changing every 10 seconds. Your degree is still valuable today. It might be less valuable in 5 or 10 years, but right now, it provides a critical foundation. Make the best use of it. Don’t just get a piece of paper; absorb the knowledge.
The Road Ahead
Software engineering is not dead. It’s evolving. The mindless part of the job is being automated, leaving behind the challenging, creative, and deeply valuable work of thinking.
The future belongs to developers who can architect solutions, debug complex systems, and provide the critical context that AI lacks. The future belongs to engineers, not just coders.