“Software is eating the world,” as Jeff Bezos famously put it, capturing how software has infiltrated every industry. Now, artificial intelligence is reshaping how that software is built — and changing the narrative around the role of programmers and software engineers. With AI writing code and “vibe coding” on the rise, many wonder: Is learning to code still relevant? What does the future hold for developers in an age of GitHub Copilot, Amazon CodeWhisperer, Replit’s Ghostwriter, Cursor, and other AI pair programmers? This post explores these questions and why, despite AI’s rapid advances, skilled software engineers are more important than ever.
Is Learning to Code Still Relevant in the Age of AI?
In a word, yes. Learning to code remains highly relevant, arguably even more so now. Some voices have argued that AI will soon automate programming entirely, suggesting that aspiring developers might skip learning to code. AI visionary Andrew Ng strongly disagrees, calling such advice “some of the worst career advice ever given.” He notes that as tools make coding easier, they expand opportunity rather than eliminate it. From the move to high-level languages to today’s AI-assisted coding, each step that lowered the barrier has increased the number of people who can program.
“This is the best time yet to learn to code,” Ng writes, emphasizing that understanding the “language of software” is key to telling an AI exactly what you want it to do. In other words, coding is not obsolete — it’s evolving. Knowing how to program allows you to harness AI coding assistants effectively instead of being at their mercy. It builds the computational thinking needed to break down problems and guide AI tools with precision. Rather than replace human coders, AI is a multiplier for those who do understand code. Developers who learn to use AI assistants become what Ng calls “10x professionals” by achieving far greater impact with the help of these tools.
The Rise of AI Coding Assistants — “Code Vibe” is the Order of the Day
Over the last few years, AI coding companions have moved from novelty to normal. Integrated into IDEs and editors, systems like GitHub Copilot, CodeWhisperer, and Cursor can autocomplete code, generate functions from comments, and even detect bugs. As AI researcher Andrej Karpathy quipped, “The hottest new programming language is English.” He’s pointing to “vibe coding,” a trend where you describe what you want in natural language and let the AI handle the syntax. In fact, some developers barely glance at the generated code at first, focusing instead on whether the feature works. Andrew Ng describes vibe coding as a practice where you “barely look at the generated code, and instead focus on the architecture and features of your application”. In other words, coding becomes more about high-level design — the vibe of what you’re building — rather than typing out every line.
This AI-assisted approach is accelerating development. Even large tech companies are embracing it; at Google, more than 25% of new code is now automatically generated by AI. And according to Anthropic (the team behind Claude), about 37% of all queries their AI model handles are for coding help — tasks like modifying software or debugging. Clearly, “code vibe” isn’t just a buzzword; it reflects a real shift in how software is created. By offloading boilerplate and routine code to AI, engineers can work faster and focus on bigger-picture problems.
However, using these tools effectively requires skill. It’s not as simple as “write my app for me” — prompt an AI agent incorrectly, and you might get poor results or nonsense. Ng cautions that “asking an LLM to do everything in one shot usually does not work.” Successful AI-augmented coding needs an iterative, guided approach. Developers break problems into parts, craft specific prompts for each piece, review the AI’s output, and refine it. In practice, the human developer is still in the driver’s seat — orchestrating the AI and making judgment calls at every step. This nuance is why coding knowledge and logical thinking remain essential; without them, one might accept flawed code that looks correct at first glance.
Everyone Can Build an App Now — But Expertise Still Matters
One of the exciting promises of AI coding assistants is democratization: “everyone can code” is closer to reality. A non-programmer with a great idea can describe it to an AI and get a working prototype. We’re already seeing people with little coding background build the app of their dreams by leaning on AI. This is a powerful development — lowering the barrier to entry unleashes creativity and lets anyone turn ideas into software.
Yet this comes with a big asterisk: building a maintainable, production-quality application still isn’t easy. Just because anyone can build an app with AI doesn’t mean it will be a good app without experienced oversight. There’s a difference between a demo that works and a polished product that scales. AI-generated code may run initially, but is it secure? Efficient? Structured correctly for long-term growth? If you don’t understand the code, debugging or improving it can become a nightmare. As one TechRadar analysis put it, today’s AI coding tools are like junior developers — they can draft code to get you started, but “complex projects still require more time, deeper expertise, and often the involvement of a professional developer.”
Let’s break down a few areas where human developers provide irreplaceable value in ensuring software is optimal, scalable, secure, and industry-standard:
- System Architecture & Design: Seasoned engineers act as architects, deciding how all the pieces fit together. They design data flows, choose appropriate frameworks, and set the technical direction. AI can generate code snippets, but it cannot yet make the nuanced design trade-offs that determine a system’s robustness. Software developers aren’t just writing lines of code; they’re architecting systems — deciding how data flows, ensuring security, and designing infrastructure that can scale. This high-level vision comes from experience and understanding of the domain — something an AI lacks.
- Quality Control & Ethical Oversight: AI might crank out code quickly, but it has no intuition for good practices or ethics. Developers must review and test AI-written code. Bugs, security vulnerabilities, or biased outcomes can slip in just as easily with AI as with human coders. It falls on engineers to rigorously test, debug, and validate the outputs. For example, an AI could inadvertently introduce a security flaw or produce a result that violates privacy norms. Human oversight is needed to catch these issues and ensure the software meets quality standards and ethical guidelines. The engineering discipline of code review, testing, and considering edge cases becomes even more critical when an AI is involved because the AI won’t warn you if it misunderstands the requirement.
- Domain Expertise & Compliance: In regulated industries like healthcare, finance, or aerospace, there are strict standards and laws that software must comply with. A generative AI won’t inherently know the regulatory constraints or industry best practices that apply, especially since it's trained not on date data. A skilled software engineer with domain knowledge has to guide the AI, embedding those requirements into the code. As Dr. Thilina Halloluwa notes, an AI can generate snippets, “but domain knowledge is paramount. The software engineer knows the specific laws, standards, and constraints that must shape any AI-driven solution”. This expertise ensures the final product is not only functional but also legal, safe, and appropriate for its context.
In short, AI can help build faster, but human expertise ensures you build the right thing and build it right. Without understanding software architecture, a generated app might work for 10 users but crash under 10,000 users. Without secure coding practices, an AI-written feature might expose user data. And when something goes wrong (as it inevitably does), who will debug it? Poorly understood AI-generated code can lead to painful debugging sessions and “mystery” bugs that only someone with a coding background can unravel. That’s why companies still need experienced engineers to vet AI contributions and apply rigor. As one engineer quipped, accepting AI-written code without understanding it is a growing trend, but it’s “a trap in the long run” for any serious software project (nobody wants to be on call at 2 AM trying to decipher an AI’s code logic).
Notably, AI pioneers themselves advocate this balanced view. Andrew Ng readily admits that “large-scale, critical systems will rely on highly skilled software engineers, data engineers, DevOps specialists, and more”, even as smaller applications become easier to spin up with AI. In other words, vibe coding can accelerate prototypes and help non-developers get going, but it won’t replace the need for professional software engineering when it comes to building robust, reliable systems.
“Software Engineers are the New Architects”
Far from making engineers obsolete, AI is elevating the role of software engineers to a higher level of abstraction. Andrew Ng mused that software engineers are becoming “the new architects” in this AI-driven development paradigm. What does this mean? Consider a traditional architect: they don’t personally lay every brick; instead, they design the blueprint and oversee construction specialists. Similarly, developers are shifting toward an architect’s mindset — designing and guiding the construction of software rather than hand-coding every function.
With code generation on tap, the developer’s focus moves to what the software should do and how it should be structured rather than the low-level details of writing the code. We already see this in the vibe coding approach: you concentrate on the features and overall architecture of your application and let AI handle a lot of the boilerplate implementation. The engineer’s job becomes defining the problem, breaking it into components, and integrating the AI-produced pieces into a coherent whole.
In practice, acting as the “architect” means developers must think deeply about system design, user experience, and long-term maintainability. They might draft a high-level design (like how data flows through the app, which services communicate, and where to enforce security), then use AI tools to flesh out each part. Creative problem-solving and decision-making remain human-led. AI is great at producing code that follows a pattern, but deciding which pattern to follow and why is an architectural question left to the human expert. This shift is making the work of coding more strategic. Good software engineers will increasingly be those who can leverage AI as a tool while still providing strong technical leadership and architectural direction. They become, in effect, AI-era software architects — a hybrid of coder, designer, and planner.
Ng’s observation that “software engineers are the new architects” highlights a positive future for developers: rather than being replaced, they are evolving into higher-impact roles. By mastering AI coding assistants, tomorrow’s developers can handle bigger projects and multidisciplinary concerns, much like an architect coordinates various aspects of building design. This evolution might even attract more talent into software engineering, as the job becomes less about rote coding and more about creative design and problem-solving at scale.
Keeping Humans in the Loop — Augmented, Not Replaced
The key to harnessing AI in software development is collaboration. AI excels at speed and scale; humans excel at judgment, creativity, and understanding context. When we combine the two, we get the best results. As Dainius Kavoliunas wrote, “The future of coding isn’t about choosing between AI and human developers — it’s about collaboration. The best developers will be those who know how to harness AI effectively.”.
In practical terms, this means keeping a human in the loop at all critical stages of development. AI might generate the code, but a human should review it, test it, and decide if it aligns with the product vision. AI might suggest ten different ways to implement a feature; an experienced engineer evaluates which approach is best for the users and the business. Human insight guides the AI, ensuring the end product truly solves the right problem in an efficient, cost-effective way. It’s all too easy for an AI to write code that technically meets a prompt but isn’t optimal for the situation (for instance, it might use an inefficient algorithm that only an expert would recognize as problematic, which i can attest to). By keeping humans in charge of directing and approving what the AI does, companies make sure they aren’t just building software quickly but building the right software.
Another reason the human loop remains vital is accountability. If an AI writes a flawed piece of code, who is responsible when it fails? Ultimately, the engineering team is. Skilled engineers overseeing AI contributions can catch issues early, whereas a fully automated pipeline could deploy faulty code without anyone noticing until it’s too late. This is particularly important for mission-critical systems. AI doesn’t feel the weight of responsibility — but human engineers do, and that drives them to double-check and test rigorously.
Finally, involving humans keeps development aligned with real user needs. AI isn’t going to question the spec you give it; it will do exactly (or approximately) what you ask, for better or worse. Developers, especially those with product sense, can think from the user’s perspective and adjust course as needed. They act as the conscience and compass of the project, steering the AI to ensure the final product is user-friendly and provides real value, not just technically functioning. This human-centric approach is what differentiates a successful app from a merely functional one.
A New Narrative for Programmers in the Age of AI
AI is undoubtedly changing the game for software development. Coding is faster and more accessible than ever. A solo developer today, equipped with AI pair programmers, can accomplish what used to take a whole team weeks or months. The narrative is shifting from “How do I code this?” — Stack overflow era to “What do I want to build?” — LLMs efdcra. This is an empowering change, opening the door for more innovation and allowing developers to tackle higher-level challenges.
But this doesn’t spell the end of programmers — it marks a new beginning. The role of the software engineer is evolving, not disappearing. Routine coding might be handled by AI, but the architecting, innovating, fine-tuning, and decision-making are squarely in human hands. Those aspects become more important as AI takes on the grunt work. It’s a bit like autopilot in planes: it handles the steady cruising, but you still need a skilled pilot for takeoff, landing, and when there’s turbulence. Likewise, AI can autopilot a lot of coding, but human engineers are the pilots charting the course.
Going forward, learning to code remains a smart move — not to out-compete AI at rote coding, but to better collaborate with AI and supervise its work. The developers who thrive will be those who blend technical knowledge with strategic thinking: part programmer, part architect, part product thinker. They will embrace AI tools without becoming over-reliant on them, maintaining the discipline to verify and improve AI outputs. This balanced mindset ensures that we use AI to accelerate development while still upholding quality, security, and user experience.
In summary, AI is changing how we write software, but it isn’t replacing the need for human software engineers. Instead, it’s elevating their role. Software engineers are becoming the architects of the digital world, using AI as a powerful tool in their toolbox. With AI’s help, they can build more, build faster, and build smarter — but their guidance remains the blueprint behind it all. The new narrative for programmers is not one of obsolescence but one of augmentation and growth. By staying curious, continually learning (about both coding and AI), and keeping a user-focused, ethical approach, today’s engineers will lead the charge in this exciting era of AI-assisted development. And that is good news for tech professionals, aspiring developers, and non-tech founders alike: The future of software will be built by human creativity amplified by AI, not the other way around.
(Human + AI) > AI alone — the best software of tomorrow will come from this synergy.
AI is undoubtedly reshaping software development, but far from replacing human engineers, it elevates their roles, transforming them into strategic architects. Engineers skilled in guiding AI tools, ensuring software integrity, security, and usability, will drive the future of technology.
References to read further
[1] Marc Andreessen (2011). “Why Software Is Eating the World”.
[2] Andrew Ng (2025). “Not Learning To Code Because Of AI Is The Worst Career Advice Ever”.
[3] Andrew Ng (2025). “A 10x Engineer: A Widely Accepted Concept”.
[4] Simon Willison (2025). “Not all AI-assisted programming is vibe coding (but vibe coding rocks)”.
[5] Hugh Langley (2024). “Google CEO says more than a quarter of the company’s new code is created by AI”.
[6] Andrew Ng (2023). “ChatGPT Prompt Engineering for Developers”.
[7] TechRadar (2023). “Securely working with AI-generated code”.
[8] Thilina Halloluwa (2022). “AI as a Replacement for Software Engineers?”.
[9] Bhaskar Chakravorti (2024). “AI’s Trust Problem”.
[10] McKinsey & Company (2023). “AI in the workplace: A report for 2023”.
[11] Dainius Kavoliūnas (2023). “Vibe coding isn’t here to take developer jobs. It’s here to transform them into AI architects”.