A Developer’s Take on 2026

After years of honing my craft, learning how real systems fail, and continually evolving my skills, myself and millions of other developers were told that our jobs are near extinction this year. In hindsight, the idea that fully autonomous coding agents would take my job by completely replacing senior developers feels like a toxic and dangerously misleading lie meant to do nothing but market the next hot AI product. Over the last few months I've accepted that coding will inevitably change and is never going to be the same, but the narrative that "learning to code" is a relic of the past is still something I can't get behind.

Lately there's been a push for people to give up critical thinking and accept that knowledge, skills, and coding itself can be offloaded to agents. I can't help but feel a sense of unease about Sam Altman's remarks about intelligence becoming a standard utility that we pay for, like electricity or water. While I find myself stuck in this strange middle ground of loving and hating AI, I still can't believe how much I have started to immerse myself in these tools over the past six months.

You can use these tools with a certain level of efficacy without understanding the underlying principles. But pure vibe coding and agentic engineering with a toolbelt of real skillsets are two very different practices. When you build an application that you plan to maintain, wich is pretty much an essential requirement of anything useful, you need to understand how it works inside and out. The non-deterministic nature of AI programming makes it statistically impossible to build something you don’t understand while also expecting it to work reliably every time you need it to. The idea that “anyone can build something” is a dangerous oversimplification that ignores the complexity of software development and the skills it requires.

The Promises and the Lies

The narrative that all developers will face unemployment soon is thankfully fizzling out, but I can’t help feeling a bit of anger at how readily it was accepted and cheered on by many CEO’s. The dystopian angle still gets plenty of attention, but the real issue is how normalized it’s become to talk about a world where AI runs everything by default and our skills are basically not needed. If everyone can build something, then that something simply loses its value.

Yet software companies and SaaS infrastructure remain a pivotal part of what holds our economy together. One reason for this, at least from my perspective, is that vertical SaaS companies have been forced to pivot outside of their general-purpose restrictions. The fear went from an agent doing someone’s job to an agent serving the same results as an entire SaaS platform. In many ways I think this creates a healthy restructuring and a new era of hyper-specialized products. Instead of general knowledge and horizontal focus, we’re seeing tools that target very niche domains and vertical scopes.

The Year of the Product Engineer

This has opened the floodgates to all kinds of products, and it has closed as many doors for developers as it has opened. That’s why the narrative that “programmers are being replaced by AI” isn’t as real as it seems. The reality looks more like H-1B visa games, developers doing what used to be three people’s jobs, and interestingly a moment of glory for product designers and engineers. The question is no longer how to build. It’s what to build, and why.

That framing isn’t mine alone. Tobias Cornille makes a similar argument in The Golden Age of Product Engineers.

Filling the Gaps LLMs Can’t

A few months ago it was often easier to just build parts of my software ony my own. But ever since that historic winter break where everyone came back to work and the agents were suddenly “just working,” it has become much easier to simply tell a model what you want.

These agents are fully capable of writing code now, but they still lack taste and a human understanding of what we actually need. I think this is one of the core reasons they will continue to be so awful at UX and frontend design. Without a fully thought-out, well-structured wireframe, the slop factor is obvious to anyone in the field. Maybe it is just a lack of taste, but in many ways it feels like the best product engineers are the ones who understand the needs of a customer, and this is something an AI will never technically truly be able to grasp and feel.

I also think this is the defining factor in why an engineer with years of experience will build something far better than a amateur vibe coder. We can debug and we know when something is really working. We understand the difference between a JWT and an OAuth flow. We know why Astro is often a better tool than Next.js. But more importantly, we know what customers want and we have many personal stories of times a product did not hit that mark.

My Experience this Year

I see this strange merge happening in my own career, where my interests keep shifting toward product engineering. In a world where raw implementation is easier and cheaper, it makes sense that the challenges of design, translation, and connecting business constraints into one coherent picture are now the actual focus-point.

My brain gets more time now to zoom out on the larger landscape. In other words I get to observe and impact a product’s lifecycle from inception to distribution. I don’t code by hand as much as I used to, and I do miss the reward of solving gnarly bugs myself, but there’s a different excitement in thinking about the factors I was often left in the dark on.

Where We Are Going

I learned this lesson a long time ago, and it’s worth saying plainly: the newest shiny tools aren’t always the most optimal.

The ecosystem is moving at an almost incomprehensible pace. The pros are real: access to tools, ideas, and workflows that didn’t exist a year ago. The cons are just as real: noise, abandonment, too much choice, half-baked products, and a culture that rewards hype over depth. Yet on top of these transformations, I’m also seeing benefits from the AI craze I would have never predicted. I’m finding more opportunities to carve out time for interests I always wanted to explore but never had the time or a fully committed tutor for. I am losing track of the personal projects I have explored and learned at home where Claude or Codex has helped me every step of the way.

In many ways it feels like AI has reduced the penalty for deeper exploration and curiosity. I’m still very much invested in my job as a full-time developer, but instead of learning the syntax of one specific database that deeply, I can actually explore two or three, what they might look like and actually understand the tradeoffs through a conversation with an agent.

Fun Explorations

Yesterday I booted a t2linux-patched Arch ISO from USB behind an Apple bridge controller, partitioned ext4 with a GRUB UEFI bootloader, and brought Hyprland with SDDM as the display manager into the mix. I would never have figured out how to partition my Mac drive without an LLM coaching me through it the way it did. I learned more than I would have surfing the web and sifting through toxic Arch forums where people bicker about this tool vs that. I ended the day with a little more experience than I had the day before, and I only see that trajectory improving as these models grow with us.

One thing definitely hasn’t changed, and it reminds me of my early days programming. I still ask myself: how did a kid who wanted to be a DJ, with zero affinity for math or science (or tech for that matter) end up becoming a fairly successful developer? The answer was exploration, curiosity, and persistence. I think you’ll see those same strengths reward the current and next generation of hybrid programmers, product engineers, developers and anyone else who is an integral part of the journey to production.

Where This Leaves Us

It’s impossible to say where this is all going over the next few years, but I find it hard to label myself as just a developer these days. I think the industry is going to reward specialization and vertical focus on deep niches an AI can’t solve directly. On top of that a broad understanding of the tech around us, and how AI integrates with it, is what keeps you ahead. Ignore all the crap about “being left behind.” I should have been left behind a long time ago by those calculations, and guess what? I am still as eager and motivated as ever.

I will always consider myself an engineer, and the imposter syndrome will never go away, but I don’t think AI made me feel that way. Over the next few years I expect that most software that breaks boundaries will still be written by extremely experienced developers who know what they’re doing.

I had many opportunities to jump ship and change careers, and I’m really happy I stayed. I love building things, and I love that a handful of people can now build what was once the domain of large enterprise teams. In a strange way, as embedded as these AI systems have become with corporate investment, the barrier to entry and the cost of building something has handed power back to the individual.

I don’t see that changing. The best thing we can do is stay informed, keep learning, and not forget our roots. I’ll never be able to just shred through an application without understanding how it works. Thankfully, I think that’s still going to be a requirement for building something that lasts.