Whether AI Will Be Catastrophic for Software Development Depends on Maintenance—and a Decision

Published on Oct 29, 2026, filed under development, ai, philosophy. (Share this post, e.g., on Mastodon.)

AI means a seismic shift for software development. In my view, it’s likely to be catastrophic for most people in the field only once we train, prompt, and automate it to do the one thing well that makes all the difference: maintenance.

The reason for this is that close to all software projects decay over time. Good software, and that includes websites, is therefore not “fire and forget”—“fire and forget” software has a timer running for when it becomes stale, brittle, or insecure.

If AI cannot or is not used to maintain software, there’ll be a need for humans and conventional tooling to take care of it. With more and more software being generated through AI, this may even grow maintenance-related parts of the field.

But if AI is able to handle maintenance well—this seems likely, but we don’t seem to be there yet and I’d generally like to leave the question open out of respect for what it means to maintain code over years, if not decades—and if it’s set up to do so, then there may be very little AI might not affect. Personally, I wouldn’t place bets, not even on AI limits around technical innovation.

The Good News and the Bad News

Is this “bad”?

For us as people, it’s certainly uncomfortable. AI may be even more disruptive for our field and our careers than it already is.

But there may be another side to exactly why AI can never be ethical or safe (context and intent). Maintenance is not only mechanics; it’s deciding what software should do, what to optimize, what to remove as needs and requirements change. This may create a space for us where, alongside creativity and a drive towards mastery, we have to guide and oversee AI (we need to do so in general, but this will create some new opportunities).

That’s the good news.

Here’s the bad news:

Our model of the world is insufficient (I discuss this in other writings). Informed by this insufficient model, our intentions and actions are not constructive. ^* Yet when we guide AI, it’s our intentions that it carries out. Taking responsibility for AI in the very area in which it’s fundamentally limited therefore requires us to collectively take a major step forward.

Fortunately, whether we take that step depends less on a more accurate and constructive model of the world than a decision—whether to keep doing and accepting that others do things that harm our environment and anyone in it, or not.

To close the argument—which may be an example of how philosophy is quite practical—, when we decide to be more constructive, we can also prevent a technology like AI from being destructive (which does not only apply to software development). It’s obvious when you think about it—but we need to start making the connection.

^* Think about us destroying our habitat and killing vast numbers of members of our and other species. This is the definition of being destructive.

About Me

I’m Jens (long: Jens Oliver Meiert), and I’m an engineering lead, guerrilla philosopher, and indie publisher. I’ve worked as a technical lead and manager at various companies (e.g., Google); I’m an active web and tool developer (code optimization, digital defense), a contributor to web standards (like HTML, CSS, WCAG), and a book author (O’Reilly, Frontend Dogma). Stay in touch.