Thinking about custom software in a new way

At this point, anybody saying things about AI’s usefulness in software development like notch did on X the other day is either a) burying their head in the sand, b) an old man yelling at clouds, or c) seriously entrenched for some other reason.

Here’s notch’s take, so you don’t have to bounce:

Reminder that using AI to write code is an incredibly bad idea still, and anyone advocating for it is either incompetent or evil.

It’s just as dumb as letting AI write the laws. It’s about logic, not about typing.

Well, call me incompetent. Or evil. Call me whatever makes you feel good about yourself. But I believe you should be using AI to write code today. And tomorrow. And for the foreseeable future.

Two things can be true at the same time:

The AI hype is (still) off the charts
Claude Code has changed how I think about custom software

After many years of software development, both professionally and personally, my default refrain when presented with the prospect of throwing custom software at a problem was: perish the thought!

In other words: No, let’s not. Unless it is absolutely necessary and obviously worth the (total) cost (of ownership).

That’s because I was intimately familiar with exactly how expensive a proposition custom software is, so I avoided it if at all possible.

Now I’m thinking about custom software in a new way.

The price to prove a concept, to spike out a prototype, to build a thing and see how it works, to “go West, young man” has dropped precipitously thanks to these new tools. The new price has moved me.

From “perish the thought” to “entertain the thought”

From “but why?” to “why not?”

In other words: Yes, let’s. Unless it is absolutely unnecessary and obviously not worth the (total) cost (of ownership).