Every few years, developers are told their jobs are about to become obsolete.
Early programmers, often working in Assembly or machine code, were told that the rise of high-level languages like C would mean anybody could write code in a language close (or at least closer) to natural language, and without needing to understand the hardware.
Similarly, the rise of no-code tools throughout the 2010s, and the continued popularity of CMS’ like Wordpress, meant many onlookers felt like it was a sign of a shift away from complex coding, towards drag-and-drop and WYSIWYG editors, with almost all code abstracted away.
This same narrative has re-emerged in recent years, as AI tools pop up everywhere, each one promising (or threatening) to be the one that changes development forever.
Big players like GitHub have predicted a near-future where we have one billion developers, with tools like Copilot helping people who know how to code - and those that don’t - build something, regardless of their skill, and Vercel’s v0 (and any number of similar tools) suggesting anyone can now vibe-code a full-stack, production-ready application.
Gunnar Grosch, Principal Developer Advocate at AWS, reminded us in a recent Coffee with Developers episode that developers hear doom-and-gloom all the time, and reality rarely turns out that way. With this in mind, we took some time to consider what engineering really is in the age of AI, and how developers can make sure they stay ready for whatever comes their way.
More AI Means More Developers
We so often hear it said that AI will write more and more code in future, and the natural conclusion to this is that humans will play less of a significant role in the process.
However, the reality might be the exact opposite.
Gunnar pointed to the Jevons paradox to explain his point, saying that as AI becomes more advanced and efficient, it frees up developers to complete more meaningful tasks, which in-turn results in greater productivity. So rather than removing the human from the loop, there’s a good chance we’re going to remain a part of it, with AI taking care of tasks we’d rather not do by hand.
“AI is just the latest (hype wave) that says, it’s the end of the developer, but what we see every time is that demand for developers actually increases".
The Art of Keeping Up
Right now, the pace of change is dizzying, and perhaps quite intimidating for developers.
New models drop weekly, each claiming to embarrass the last on benchmark tests, with clever marketing creating a sense of FOMO among developers who haven’t even tried the last model, let alone the brand new sparkly one.
The issue here is it’s harder to hear the signal in so much noise.
“It’s an avalanche,” Grosch said. “There’s so much happening, but a lot of it isn’t really that important for us as developers.”
He’s right. Not every headline or tool release deserves our attention. What matters isn’t which model we use, but how we use it. AI is becoming less about the shiny new thing and more about integrating it into the messy, practical, and deeply human process of building software.
Sure, AI can generate lots of code, and quickly. But secure, maintainable, scalable code? Not so much.
Ultimately, the hardest part of engineering has never been writing code.
“Providing code was never the issue,” Grosch said. “It’s about creating something that works in production; something reliable, maintainable, and valuable.”
AI can write a thousand lines of code, but making it secure, scalable, and user-friendly are still very much human jobs.
The Junior Developer That Never Sleeps
Looking past the hype and excitement around new models, there is still a long way to go when it comes to the intelligence part of AI. They make silly, obvious mistakes. They hallucinate entire concepts. They’re confidently incorrect, even when presented with evidence, and occasionally brilliant.
If a human employee was described in this way, they’d be a great asset to any software engineering team, but would need plenty of supervision.
One of Grosch’s best analogies is that AI coding tools are like a junior developer who’s memorised the documentation but never deployed to production. They’re fast and eager, but also worryingly overconfident, even when making massive mistakes.
“It’s confident even when it’s wrong,” he laughed. “And then it’s confident in its next answer too.”
That’s why the skill of verification, knowing how to test, review, and validate AI output—is becoming a new pillar of engineering. Code review is different now: AI tends to overproduce, hallucinate, or add unnecessary complexity. Developers need to rein it in, apply guardrails, and ensure observability into what these systems are doing.
This doesn’t represent a change in software engineering, but actually a shift back to engineering fundamentals.
Specifications, design docs, version control, and documentation aren’t old-school relics of a prior version of engineering, they’re core safety nets. AI can help us generate a spec or a first draft, but we still need to shape, reason about, and own the final outcome.
“We can get AI to work for us,” he said, “instead of us working for the AI.”
The Shrinking Middle
So according to Gunnar, we should treat AI like a junior developer, but it’s a very different thing to actually replace your junior devs with LLMs.
Even so, it’s happening everywhere, and it’s a worrying trend. Over the past 18 months, there’s been a noticeable decline in the number of junior roles. Many job listings focus on Senior roles, asking for “AI-ready” developers. Companies want people who can wrangle these new tools immediately, with a wealth of experience and technical skill to fall back on when the AI creates an issue.
The issue is that if we’re replacing juniors with AI, who will take charge when they move on? As Grosch warned, “In two years, when your seniors move on, who’s left to lead?”
AI shouldn’t replace juniors, but it should help train them. A good developer grows through curiosity, through debugging, and through failure. Those experiences build intuition and judgment, and while AI could remove some errors in the code, developing skills runs much deeper than simply removing errors from your IDE as quickly as possible.
Treating AI like a helpful colleague rather than a shortcut lets new engineers learn faster without skipping the foundations.
Ride the Wave
We are, undeniably, in a hype cycle. Some call it a bubble, with an increasing number of worried sounding voices saying it’s about to burst. Some call it a revolution. The truth is probably somewhere in between. As Grosch put it, “There’s going to be hype on one hand and scare on the other. Where we want to end up is somewhere in between.”
The best developers will be the ones who can ride the AI wave, adapting to the changes that it brings but without outsourcing their thinking.
As Wired put it, “AI may write more code, but humans still have to live with it.”
The engineering mindset - verifying, observing, designing, and questioning - remains as crucial as ever. No matter how fancy our tools get, if we stick with this mindset, development will continue to thrive, even if it looks a little different than it does today
AI won’t end software development, it’ll just make the job more interesting.