AI is reframing every mistake and success we had over 50 years, and that's a good thing…

Word on the street is that AI is making software development worse. Less stable. Less secure. Less of anything you can think of. It's not all nonsense, and even if many of these claims are blown out of proportion, I think it's good to have these counterarguments to the hype we see elsewhere. I have seen and read many takes over the last couple of years, many of them interesting, or at least worth a conversation. But among all these perspectives, I think there is one more angle that we should take a step back and look at, namely the enablement angle.

Not in building products. No. Pure software engineering. The tech stuff. The coding parts, the framework bits, all the things previously would have taken teams months or even years, are achievable much quicker with nearly — or quite literally — negligible cost.

AI enables something we've never had the luxury of doing before: failing truly fast. Experimentation with virtually zero cost.

You won't have to look far to find software engineers who feel very strongly that AI has ruined software development. It ranges from "it's so not what it used to be" to "this is not what I signed up for". Neither of these feelings are unwarranted. They're not wrong. Software engineering is far from what it used to be just 2–3 years ago, and many of us got into building software because we loved the idea of coding. I remember at 21 seeing the code I typed turn into a website that I could interact with in the browser. It was mind-blowing. I felt like I was Neo from The Matrix. Obviously, I was not, but it felt incredibly cool to be able to do that, and it hooked me forever.

Trying things at no cost

When Angular 1 was discontinued, it was quite the disaster, and that's no dramatisation of what was going on in board-rooms. Countless companies and teams built their software on a framework that got abandoned for a V2 that required a major refactor. Sure, it received security updates for a few more years, but the pain was very real and financially speaking highly detrimental.

When Angular 1 got abandoned, engineering teams suddenly had to explain themselves and find sacks of cash to address the problem.

Businesses plan. That includes engineering budgets. Putting teams in a position to have to explain in a boardroom how Google effed up, and now everyone else is paying the price — a literal price — is no walk in the park. Heads roll, and conversations escalate. I remember us spending nearly three years digging ourselves out of that hole. In USD, that's in the millions. A very costly mistake by Google and everyone who chose Angular's first version.

With AI in the mix, mistakes like these are far less damaging, and don't require escalated conversations, blame-games and millions of dollars. Moving from Angular to React, from React to Svelte or Vue is more or less a single prompt away: migrate this codebase From Angular to React. Don't get me wrong, it's not a 10-minute background task, and it will involve perhaps a week or two of fine-tuning, testing and anything else part of the SDLC, but it's not a 2-year migration, but a one or two-sprint migration. The cost is negligible.

AI allows us to make mistakes like picking the wrong framework, the less than ideal libraries, the wrong language. Betting on the wrong horse, is much less of a disaster than it used to be.

I find that to be a considerable benefit, we haven't really explored yet as an industry, or at least I don't see it discussed much. It's either "AI sucks" or "AI is the best thing since fire". Obviously, it's neither, and its impact varies from use-case to use-case, but I think it can help a lot with allowing ourselves to make bad bets on an engineering level and stress a lot less about financial repercussions. That's not to say we can take security lightly — looking at you Claude and Open Claw — nor that we can merge anything and everything into master without some form of reliable review system. It rather means that we don't have to figure out everything upfront, we don't need to lose sleep over making the wrong technological bet. It's reversible, migratable and at a much lower cost than before. And guess what? Now we have another good excuse not to optimise early everything.

Experimentation at a different scale

Beyond being allowed to make mistakes, however, there is another aspect I feel very positively about with AI as a development partner: experimentation. One of the bigger challenges I see in software teams is engineers not getting enough time to truly act as engineers.

Engineers don't get enough time to be engineers, just build-build-build, feature after feature. AI is changing that. You can try more with less.

If you have ever worked on a pet project, you know what it's like to abandon it half-way and rebuild it in a different language, framework, architecture, etc. Heck, it's a running joke that every engineer has built at least 10 To-Do apps in 10 different languages and none of them are actually finished. Each of those abandoned projects took at least a few days. Time, even your own, does not come free, and as much as we enjoy the process of building these projects, learning a new language, architecture or framework, all that time does add up.

A few months ago, I "vibe-coded" a macOS app. I wanted to turn some of my Medium stories into Instagram Reels. I spent a total of about 8 hours on it, though, half of that was really just bug-fixing and fighting the video rendering engine. Pre-AI, that project might have not even happened. I am not a macOS developer, I know little to nothing about video rendering engines. I would have probably opted to build it with Electron instead, only to run into other issues. Long story short, the project idea itself would have been intimidating enough to just abandon before even starting, and if I did decide to dive into it, it would have taken me several weekends to develop the same application.

With AI, there aren't intimidating projects anymore, and you don't even have to decide upfront which avenue you want to take. Take them all, and find out which fits the bill better.

And you don't have to do it "blindly" either. You can use spec-driven development to start in parallel more versions of the same idea. For instance, in my case, I could have developed an Electron app too, and another one using a template system rather than a pixel-by-pixel video rendering solution. Heck, I could have even asked AI to give me additional ideas, and try all of those as well. At the same time! The bottom line is, you have an idea? You can try it, and if it doesn't work as well as you hoped, you haven't lost much, if anything. I find that incredibly exciting. That's a luxury we have never had before in software engineering. Time was always a massive constraint. It isn't anymore.

And who even knows what else…

AI-driven software development is still a developing story. What comes next is difficult to predict. Partly because we have not yet reached peak AI, and partly because human creativity is virtually endless, and engineers will find very creative uses for this emerging technology. But overall, it's exciting, and I really do feel it gives software engineering another chance, it breathes a whole new life into software development.

We finally get to try the things we didn't try before. Mistakes are easier to shrug off. Things that work, we find out quickly. Risks are much lower. AI emboldens us. Adopting one tech over another doesn't feel like selling your soul.

That's inspiring in ways we have not felt inspired in a long time. Because if you look at anything in STEAM it was never about just building things, but trying, failing repeatedly, learning, then succeeding and AI makes it a lot more affordable. Truly looking forward to the next decade of software engineering.

Attila Vago — Software Engineer improving the world one line of code at a time. Cool nerd since forever, writer of codes, blogs and books. Author. Web accessibility advocate, LEGO fan, vinyl record collector. Loves craft beer! Read my Hello story here! Subscribe for more stories about LEGO, tech, coding and accessibility! For my less regular readers, I also write about random bits and writing.