Sustainable Software in the Age of AI Hype

blog ai sustainability software-engineering tooling

Somewhere between the AI-powered IDEs and the "GPT in every feature" startups,
we need to talk about sustainable software again.

Because the pace is accelerating. Tools like GitHub Copilot, ChatGPT, and
Cursor are dramatically reshaping how we write code. Teams are shipping faster,
writing more, and spending less time on boilerplate. It feels like we are in a
golden age of developer productivity.

But here is the thing: speed is not sustainability. And if we are not
careful, we will end up with a mountain of code no one understands, scaling the
unsustainable faster than ever before.

And yes, before you ask — this post was drafted with the help of ChatGPT.
I am part of the contradiction too.

The irony: I am using AI too

Let’s get the obvious bit out of the way: I think AI coding assistants are a
game changer.

I have been using Cursor as my daily driver, and it has genuinely improved how
I work. Faster refactors. Fewer context switches. Smart suggestions that
respect my codebase. It is brilliant.

But even as I lean on tools like ChatGPT and Cursor, I am aware of the trap:
offloading too much thinking to the machine. Not everything it suggests is
wrong — but not everything is right, either. And it is very easy to stop
noticing where the line is.

Hype does not equal help

The industry is caught in a full-blown AI hype cycle. Every team wants to plug
a model into their product. Every engineer wants to ship faster, write less,
and "use the magic."

And that is not inherently bad. But hype papers over nuance. It tells us to
move fast and automate everything, but it rarely asks whether what we are
building is something anyone will want to maintain a year from now.

AI can accelerate decay as easily as delivery

With enough prompting, a language model can write code that compiles, passes
tests, and even looks clean. But code that works is not always code that
makes sense.

I have seen PRs where Copilot generated a wall of logic that no one could
explain. It did the job, but no one understood how or why. That is not
productivity — that is entropy in fast-forward.

What sustainability actually means in this moment

Sustainable software means different things to different people, but in the age
of AI, I think it comes down to three principles:

  • Code should be legible, not just executable.
  • Developers should understand what they are shipping.
  • Teams should be able to adapt systems, not just deploy them.

In other words, the future should not be unreadable just because it arrived
quickly.

Using AI tools sustainably

I am not here to preach abstinence from AI tooling. I am using it myself — and
honestly, I am not giving it up. But I use it like a power tool: with intention,
with care, and with a healthy amount of human oversight.

Some personal heuristics I try to follow:

  • Let AI handle repetition, not responsibility.
  • Always read and edit what it suggests.
  • Keep review and design human-led.
  • Prioritise code clarity over cleverness.
  • Document the decision, not just the implementation.

These tools are assistants, not authors. They do not replace judgment — they
amplify it.

Final thought

Sustainability in software has never been just about carbon or compute. It has
always been about what lasts. What can be understood. What can be safely
handed to the next developer, or the next team, or the next generation.

The rise of AI does not change that. If anything, it makes it more urgent.

Because sustainable software is still human-shaped, even if we are building it
with a machine.