Gen AI and the Rise of Disposable Code

I’ve noticed something fascinating while using generative AI tools like GitHub Copilot and WindSurf: I often generate code for highly specific, one-off tasks, use it once, and then delete it. I don’t even bother committing the code to a repository.

As an ex-software engineer, I was trained to treat code as a sacred resource—something to be maintained, structured neatly, and stored using specialised tools while adhering to best practices. Treating code as disposable feels like a huge shift.

Writing code used to be expensive—it took time, expertise, and collaboration. That’s why reusability was such a focus: build it once, use it everywhere to save costs.

But now? Writing code isn’t always expensive anymore. With AI, I can generate exactly what I need, tailored to the task, in seconds (most of the time).

Once the job’s done, I delete the code. No maintenance, no clutter, no need to future-proof it.

This doesn’t mean that all code will become disposable. Critical operations and infrastructure will still require well-maintained, reusable, and structured code. But as AI makes code generation faster and cheaper, we’ll naturally start reducing the amount of code we store and maintain long-term.

In the SaaS and internal tooling context, the layer closest to the user will most likely be produced directly by the user. For many users, it will be easier to generate new code from scratch than to modify existing code they generated previously.

Do you see this pattern emerging in how you use generative AI for code?