AI will (does) allow for much faster application prototyping...

I think I got that one right. Last night, I whipped up a Raycast-like SwiftUI app entirely by prompt in the Claude desktop app without ever opening Xcode. So the much faster application prototyping thought definitely holds true here.

I also shared this:

At the same time, good design docs will become less common as project managers question why they’re even bothering to create them. Heck, they might even begin asking ChatGPT to write their design specs based on the code of the prototype.

I think I slightly missed the mark on this one. Before writing any prompt, I took the time to craft a good design doc, giving all of the features I wanted in great detail.

I shared that SPEC.md doc with Claude Code and asked it to interview me about the features I asked for and any considerations I might have missed. It took me through four or five clarifying questions, then made a plan for me to review. I commented on it to revise a couple of things, then told it to build the app.

In a little over an hour, with very little intervention, I had a Raycast replacement I was quite pleased with, and plan to continue using, because well, I can.

I believe what was key in this success was having a well written design spec. If I'd simply prompted my way through it without the many years of enterprise app design experience I already have, I don't think it would have come close.

However...

I also believe this is a reason for the huge influx of new apps from "promptvelopers"[^1] we now see. If you visit the macapps SubReddit, you'll find an endless "I developed an app that does x" posts. Some of them are useful, some are pointless copies of other notch apps and (ahem 😁) Raycast clones.

Now, I won't deny that I've used AI to build features for my own apps. I'm not that great of a developer if truth be told, but I'm a fantastic product owner and I know what works and what doesn't. AI has saved me a lot of iteration time, and I've found it to be a useful tool.

But as I also said in my earlier post:

As with any tool, what a user produces will reflect the skill of the person. A master craftsman will produce excellent results in a fraction of the time. The novice will produce janky unsecure slop – also in a fraction of the time.

And...

The results of a novice will improve very little. Results of a master craftsman will improve quickly. To put it another way, AI will put a magnifying glass on the efforts of those who use it. Lazy people will forever produce slop and simply not care; thoughtful creators will intentionally spend time polishing their use of the tool.

I definitely think these are right. I believe many promptvelopers don't take the time to seriously consider security in their apps. They're just trying to get something that looks flashy and works the way they want it to out the door. How soon before we start hearing of some glaring hole in a popular App Store app that steals personal information? Any day now, if we haven't already and I don't yet know about it.

So I guess the question is: Is it wrong to use AI to build apps? Some will vehemently say so and flame me to high-hell for this post. Personally, I don't think it is. Is it wrong to charge money for apps prompted into existence? On this one, I think it is -- but that's my own opinion. No judgement here.

In the end, some will continue to take a stand against AI in all forms. Others will embrace it at different levels, while rejecting other parts. Still others will embrace it without any skepticism. What I do foresee is that the AI landscape is going to be much different a year from now. Jobs will definitely be lost, but new ones will be created too.

Anyway. It's just a couple of thoughts. I find it a useful tool that I've been able to apply to some areas of my life -- but not all of them. If you disagree with me, please do so respectfully and leave your pitchfork at the door.

Thanks for reading!

[^1]: I came up with this label as I was writing that sentence. I don't think I've heard it used before, but it's possible I have. Either way, I hope it somehow sticks for these type of "developers."