Maintenance costs are still real

Back in 2016, when I was a freelancer, clients would always come to me with the craziest ideas, and the perfect replacements for enterprise tools that had everything they needed but fell just short of that one feature that would change everything.

There was one question I kept bringing up to them (and to myself): are we willing to pay the maintenance cost of this?

If that version of me had suddenly gained the ability to vibe code entire applications from scratch while the rest of the market still coded by hand, the question wouldn't have changed.

Even vibe-coded software has to be supported, maintained and improved. The reason why we choose to pay for software is not primarily (or should not be) technical constraints, but time constraints. And while letting AI agents build software for you as you focus on something else is indeed time-saving, it's still not enough to absorb everything else that comes with shipping software to real customers, even if the customers are internal, or just yourself. Things will fail, servers will go down, bugs will be found, features will have to be improved. Technical and design shortcomings are common even if you carefully planned everything beforehand. People will demand support, call you when things break, and require that fixes be shipped urgently if problems are critical enough.

I don't think maintenance costs are going away. So even if we can build an alternative to every service possible out there, we still owe ourselves the question: are we willing to pay the maintenance cost? And is that worth it for our business? Sometimes it is, most times it isn't.