Development 7 min read

Vibecoding and the illusion of competence

James Barnes Design Engineer & Front-End Developer

Let me say this up front: AI has been one of the biggest accelerants of my working life. It drafts, it debugs, it explains unfamiliar code, and it turns an afternoon of boilerplate into ten minutes. I use it every day.

I also don’t blame anyone for diving in. If you’re learning to build software with AI right now, good. Knowing how to work with these tools is quickly becoming a core skill, and the people who learn it will have options others won’t.

But alongside the tools, a story has grown that I don’t buy: “You don’t need to learn to code anymore. AI can do everything.” That story is where things start to go wrong, for individuals, for the craft, and for the world more broadly.

The confidence gap

In 1999, psychologists David Dunning and Justin Kruger described a familiar pattern: people with the least skill in an area tend to overestimate their ability the most. The knowledge you need to do something well is the same knowledge you need to notice you’re doing it badly.

AI widens that gap in a new way. A beginner used to hit a wall quickly. The code wouldn’t compile, the page wouldn’t load, and reality gave honest feedback. Now AI clears that first wall for you. The app runs, the demo works, and because it works, it feels like mastery.

What’s easy to miss is everything underneath: the SQL injection in the login form, the API key committed to a public repo, the bug that only shows up with two users at once. Experienced developers spot these because they’ve been burned by them. Someone who has never written code by hand often doesn’t know to ask.

That isn’t a reason to gatekeep. It’s a reason to keep learning. Shipping something is a brilliant start. It just isn’t the same as understanding it.

How you use it matters

In January 2026, Anthropic published a randomised trial with 52 mostly junior developers learning a new Python library. The group using AI scored 17% lower on a quiz about concepts they had used minutes earlier, with the biggest gap in debugging.

The interesting part was the split inside that group. People who used AI to ask questions and get explanations kept their understanding. People who simply handed over the writing didn’t. Same tool, very different outcomes. AI can be the best tutor you’ve ever had, or a way to skip the learning altogether. That choice is ours.

Why many of us code

For a lot of developers, writing code was never just a means to an end. We got into this because we loved it: the puzzle of a tricky bug, the satisfaction of an elegant function, that 1 a.m. moment when something finally clicks and you understand not just that it works but why.

And that’s the part AI is quietly taking from a lot of us. Many developers I know are genuinely miserable right now, and it isn’t because the tools don’t work. It’s because the job has changed underneath us. We aren’t really programming anymore. We’re orchestrating: writing prompts, reviewing output, nudging an agent back on track, approving a diff. The work gets done, often faster than before. But when it ships, it doesn’t feel like we did it.

That sense of completion, of wrestling a problem to the ground and knowing you solved it, is a huge part of why this job is satisfying. Take it away and you’re left with output you supervised but didn’t make. For people who code because they love it, that’s a real loss, and it deserves to be said out loud rather than brushed off as resistance to change.

The pressure to keep up

The pace right now is relentless. A new model every few weeks, a new “10x” workflow, a new post implying anyone not using it will be unemployable by Christmas. That pressure pushes people to adopt tools out of fear rather than fit, and to hand over more than they meant to.

The costs are quiet but real:

  • Skills fade when you don’t use them. This is the one I worry about most. Programming is a skill like any other, and skills atrophy when you stop practising them. Stop writing tricky logic yourself and, a year later, you may reach for the prompt box before you even try, and struggle when the AI gets it wrong and you have to fix it yourself. The better the tools get, the easier it is to stop practising without noticing.
  • Debugging gets harder. Reading code you didn’t write is harder than reading your own, and AI-generated code is always code you didn’t write.
  • Focus fragments. Deep work can turn into a loop of prompt, wait, review, correct.

None of this means don’t use AI. It means use it on purpose, not out of panic.

Everyone has the same shovel

There’s also the dream that AI is your ticket to building the next big app and becoming a millionaire overnight.

But AI didn’t give you a secret advantage. It gave everyone the same one. If you can build a habit tracker in a weekend, so can the next person, and so can the thousands who already did. When anyone can build something, the thing itself stops being scarce, and scarcity is what people pay for.

Honestly, that’s a relief. There’s no race to win, so you can relax. Instead of trying to sell people something they could make themselves, point AI at your own life. Use it as an assistant and a coworker: automate the tedious parts of your business, draft the emails you dread, learn something new, build the small tools only you needed. The real payoff is a better life and a better business of your own.

The bigger picture

Zoom out from our desks and the stakes get much larger. The risks AI poses to society and the environment are massive, and they deserve more attention than they get.

  • Energy and the environment. The International Energy Agency estimates data centres used about 415 TWh of electricity in 2024, roughly 1.5% of the world’s total, and expects that to more than double to around 945 TWh by 2030, with AI as the main driver. A typical AI-focused data centre uses as much electricity as 100,000 households, before you count the water used for cooling and the mining behind the chips.
  • Work and learning. Entry-level roles are where people learn their trade. If AI absorbs that work, it’s not clear where the next generation of experienced developers comes from.
  • Truth. Cheap, convincing text, images and voices make misinformation and scams easier than ever.
  • Power. Building frontier models costs enormous sums, which concentrates a great deal of influence in a handful of companies.

But like it or not, the genie is out of the bottle. These tools aren’t going back in the box, and refusing to touch them won’t make the risks go away. It still pays to adapt. The goal is to use AI thoughtfully, with your eyes open, and to be wary of anyone telling you it’s all upside.

Use it, and stay in charge

So here’s where I’ve landed:

  • Learn to code, and learn AI. You can’t judge output you don’t understand, and the better AI gets, the more that understanding is worth.
  • Use it as an accelerant and a tutor. Ask why. Ask for the trade-offs. Make it explain itself.
  • Keep your hands on the hard parts. Let AI take the boilerplate; keep the logic that matters.
  • Protect the joy. If you love coding, make time to just code.
  • Remember the cost. Every prompt runs on real electricity in a real building somewhere.

AI can build you an app, and it can make a good developer dramatically faster. It just can’t make you a developer on its own. That part is still up to you.

Sources

Continue Reading
enfr