Worried AI will make you irrelevant? Another tutorial won't fix it

Every week brings a new demo of an app made in a few prompts, and a new claim that developers are finished. We use coding agents at Hexerize every day, so we've seen how good they've become. We've also seen how much they still depend on someone knowing what to ask for.

If all this has you worried about your career, you're not alone. Here's how we'd handle it.

How worried are people?

It depends who you ask. In Devographics' 2026 survey of 7,258 web developers, nearly half worry AI will displace their jobs, though its researchers note that the topic may have drawn people already interested in AI. In Stack Overflow's 2025 survey, 15% of developers called AI a threat to their job, up from 12.1% the year before.

The job numbers are mixed too. Stanford's latest payroll research found no widespread displacement, but employment among 22 to 25 year olds in highly AI-exposed jobs is about 19% below where it would be if it had kept pace with people the same age in less exposed jobs. The researchers say that's a pattern, not proof AI caused it.

So we won't tell you your job is safe. Nobody can promise that. What we can say is that alarm isn't a plan, and these three things help more.

1. Stop comparing your job to a demo

A demo only has to work once, on the path its maker chose. A product has to work on the days nothing goes to plan. A few cases from our own work:

• School fees. In HexerSMS, a parent can pay part of a term's fees by Mobile Money. The school office can put it towards the oldest debt first, and the parent's text receipt still has to show the right balance.
• Blocked words. On DropAsecret, you can keep certain words out of your inbox. Blocking "ex" mustn't catch "next".
• Slow payments. A Mobile Money approval can take a minute or two. The page has to keep checking and say so, so that people don't pay twice.

None of that makes a good demo, and all of it is the job. An agent can write the code for any of these once someone has thought to ask. Knowing what to ask comes from working with real users.

2. Learn from a real problem

A saved tutorial feels like progress, but nothing is asking you to use it. So start with a problem in the work you already do.

The people we build for often pay for every page in mobile data, so knowing what a page weighs is worth learning, whichever tools win. Pick something like that, try an agent on it, and note where it helps and where it goes wrong. You'll learn more about the agent from that than from any thread about it.

You don't need every tool either. In a BCG study of 1,488 US workers, people reported being more productive with up to three AI tools, and less productive from the fourth.

3. Close the laptop

As agents write more of the code, more of the job is judgment: deciding what to build and whether the result is right. That's hard to do tired.

In the same study, 14% of workers using AI reported mental fatigue from using or overseeing it beyond what they could handle, and software development was among the fields where it was most common.

An evening off won't answer your questions about the future. But you'll check an agent's work better after a night's sleep than after another hour of reading about the future of your job.

One thing to do this month

Nobody has the next five years worked out, including the people posting about it. Pick something you want to be better at this month, and a problem from your own work to use it on.

And if the worry is costing you sleep, tell someone: a colleague, a friend or a doctor. It's a heavy thing to carry alone, and you don't have to.