Bespoke, Disposable, and Written by Machines
We’ve stepped into a new chapter of software development, and it feels a lot like the fast fashion era. Not because everything is tacky or flimsy, but because of the pace and the ephemerality. I can ask an AI for a bespoke one-shot app, get it in seconds, use it once, and toss it aside. A tiny program tailored just for me. Worn once. Retired immediately.
But before anyone panics: the old paradigm isn’t dead.
It isn’t even fading.
If anything, its importance has been thrown into sharper focus.
Enterprise systems still matter.
Critical infrastructure still needs slow, sturdy, hand-crafted code.
There is still a place for architecture diagrams, integration tests, CI pipelines, lint rules, team reviews, and good engineers who design things that last.
And here’s the part we shouldn’t forget:
the AI models stitching together my disposable little tools learned from that legacy.
The craft, the patterns, the conventions, the readability, the care — that’s the foundation. The fast fashion era is built on the shoulders of the slow fashion giants.
When Things Were Built to Last
Before AI, you built software the way you built furniture. Heavy, sturdy, intentional. You carried the burden of “future maintainers,” even when those maintainers were just future you. You worried about edge cases. You argued naming conventions. You wrote documentation for things you hoped someone would read someday. And because the cost of building was high, the default mindset was permanence.
We still need that.
We always will.
You don’t want your power grid running on a one-off app written at 11pm by a language model. You don’t want safety systems or financial systems or GIS infrastructure built like a Shein haul.
But Now…
Now I can say something like:
“Make a tiny browser tool that takes this AGOL metadata and converts it into a GeoJSON schema table.”
And five thousand tiny decisions get made for me, instantly, by a model that absorbed decades of engineering tradition. It produces a small React app, a CSS file that’s slightly confused about colors, and a working solution that exists purely to unblock me right now. I use it once, get what I came for, and close the tab forever.
That’s the new pattern: hyper-specific, hyper-local, single-purpose software.
Apps built for an audience of one.
Apps that don’t need to be saved.
Apps that live and die inside a single moment of flow.
The Two Worlds Don’t Replace Each Other
This isn’t a battle.
It’s a dual ecosystem.
We’ll always need slow, careful engineering for the things the world runs on. The boring stuff. The reliable stuff. The stuff that can’t break.
But we’re also entering an era where it’s normal to spin up dozens of tiny apps a week. Each one stitched together by a model trained on the careful work of thousands of engineers who did care about readability and patterns and testing.
Fast fashion thrives because slow fashion exists.
Same in software.
The Closet and the Cathedral
My digital closet is overflowing with single-use AI apps. Half-finished. Half-purposeful. Still useful. Meanwhile, entire engineering teams are crafting the digital cathedrals the world depends on. Both matter. Both deserve respect. One gives me speed. The other gives me stability.
And the best part?
They’re starting to coexist. AI tools are helping engineers write better long-term systems, and long-term systems are giving AI the patterns it needs to generate better short-term tools. It’s an odd, symbiotic loop.
The Point
We’re not replacing anything.
We’re expanding the spectrum.
On one end, careful engineers building things meant to outlast them.
On the other end, me generating a janky tool to fix one CSV column.
Both have value.
Both are part of the new normal.
And one could not exist without the other.
The fast fashion era is here.
But so is the timeless wardrobe.
And good software lives somewhere between the two.
Top comments (0)