A few days ago Elon Musk suggested that within a year or two people may stop writing code altogether and simply describe what they want AI to build. Meanwhile, I’ve been increasing the amount of time I spend learning JavaScript.

Given AI is getting better at writing software every month, why spend hundreds of hours learning a skill that some people believe is becoming obsolete? Since committing to learn how to code, I’ve periodically found myself asking myself this very question.

A few weeks ago when I reviewed my progress against my goals for the year, I realised I had barely touched my coding learning for a month. This wasn’t because I had lost interest, but whenever it came down to it, using Codex to ship something felt more productive than working through programming fundamentals. I was feeling behind on shipping, so deprioritizing coding felt like the right decision to make.

The problem was that whilst I was learning how to build with AI, I wasn't getting much closer to understanding what was happening beneath the surface. I could deliver features, but my ability to reason about the code, troubleshoot problems, or make informed technical decisions wasn't improving at the same rate.

Once I noticed that pattern, I changed course. Over the last few weeks I've carved out time each morning to study JavaScript before switching over to building Pando. The question is whether that investment makes sense. Elon might disagree, but right now I think it does.

Vibe coding is here, and it’s here to stay

The reality is that AI-assisted coding has already won. In less than four years we've gone from GitHub Copilot autocomplete to coding agents capable of building and deploying working applications. Today it's entirely possible for a single person to create software that would previously have required a small team. I can understand how software engineers who have experienced this shift feel anxious about their place in the industry.

Andrej Karpathy famously coined the term "vibe coding" to describe this shift:

"I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works."

That might sound flippant, but the shift is real. Stack Overflow's 2025 Developer Survey found that over half of professional developers use AI tools daily, while Gary Tan reported that some YC founders were generating as much as 95% of their code with AI. To quote him:

“This isn't a fad. This isn't going away. This is actually the dominant way to code.”

The direction of travel is obvious. AI is making individual builders dramatically more productive, and the amount of software one person can create today is far greater than it was even a couple of years ago.

Given that reality, I can understand the temptation to stop studying and simply focus on shipping. In many cases, that's probably the rational decision. Many non-technical users are going to build useful solutions without ever going deeper than prompts.

However, the question is whether learning to code still creates an advantage regardless of this shift.

The best managers know how the work gets done

The reason I'm continuing to learn isn't because I expect to spend my future manually writing thousands of lines of code but rather because I think understanding what's happening beneath the surface compounds as an advantage.

When you're relying entirely on AI, it's easy to build things without fully understanding how they work. That can be enough to get a prototype live, but as systems become larger the gaps start to matter. You need to make decisions about architecture, data models, security, monitoring, performance, costs, and trade-offs. AI can help with all of those things, but ultimately somebody still needs to exercise judgement.

Crucially, when something goes wrong, understanding the underlying system gives you more options and improves the likelihood of a timely resolution and future mitigation. This matters when you have production issues. You can inspect what's happening, form hypotheses, challenge the AI's suggestions, and make informed decisions rather than relying on a solution being generated for you.

Gary Tan captured this well when he said:

"The people who are most likely to get wings are the people who have taste and understand technology."

I agree that AI doesn’t remove the value of technical understanding. If anything, it increases the value of good judgement because the cost of implementation is lower. This means that the builders who can combine product thinking, technical understanding, and AI effectively are likely to outperform people who rely entirely on any one of those things in isolation.

And so?

Ultimately, as somebody learning to code in 2026, my goal isn't to become the world's best programmer. It’s to become somebody who can take customer problems and solve them with software independently.

To do that, I want enough technical understanding to read the code I'm generating, troubleshoot issues when they arise, evaluate trade-offs, and make informed decisions about what I'm building. I don't need to focus on every implementation detail, but I do want to understand enough that I'm not entirely dependent on AI whenever issues arise, or something becomes more complex than expected.

I see AI as a force multiplier rather than a replacement for expertise. The better I understand software, product development, and customer problems, the more effectively I can use AI to extend my capabilities. This applies not only to building products, but across operations, customer support, distribution and everything else required to run a business.

With that in mind, my learning path is relatively straightforward. After spending a few months learning Python, I'm currently working through JavaScript fundamentals. From there I'll move on to React and TypeScript so that I can better understand, modify and maintain the applications I'm building. Alongside that, I'm continuing to build Pando, which is giving me exposure to databases, authentication, APIs, deployments, Git and AI-assisted development in a real environment.

At present, it’s impossible to say whether this will prove to be an optimal investment. The software industry is changing too quickly to know. It's certainly possible that AI capabilities advance far faster than I expect and make much of this unnecessary.

However, my current belief is that people who combine product thinking, AI fluency and software literacy will have an advantage over people who rely entirely on tools they don't understand. At the very least, learning these skills will give me the freedom to build for customers without waiting for somebody else to do it for me.

If I can achieve that, the next challenge is finding the right problems and reaching the right audience. Right now it seems like the best way forward is to keep learning and keep building.