The Promise of AI is Nice
Kickoff
I'm about to start work on a decent sized coding project. The next version of my Neopoligenneo website app. The new design is different enough that I'm starting from scratch. It's gonna take a lot of work.
The New Vibe
Using AI services to build apps is called "Vibe Coding". The idea being that you write prompts detailing what you want an app to do and the AI generates everything for you.
I've seen folks doing it. To a degree, the results are impressive. The fact that the systems can produce working code at all kinda blows my mind. The power of stats and the huge data sets they work off is incredibly impressive. But, they fall short.
Going The Distance
The current slate of AI tools is not for me. They get lots of things right. Sometimes they get things way wrong. Worse, sometimes they get things mostly right. Leaving you with something broken in subtle ways. Errors that are easy to miss.
People are just as capable of creating semi-working code. The difference is that you work with the code during development. It breaks down like this:
- Create a mental model of the system you're working with.
- Create a mental model of what you want to change.
- Write code to make the change.
- Evaluate the change to see if it did what you thought it was gonna do. Start over if it didn't.
Your mental model gets refined on every loop. Both narrowing down the specific solution and expanding your view of the possible cases. Critical thinking that AI tools don't provide.
The time spent writing code is a feature, not a bug.
Maybe this will change.
Code Kit
I love programming. Thinking in systems and problem solving provides wonderful dopamine hits. But, I'm not religious about the tools. My goal is make things, not to obsess over what I use to make them.
A few years ago, writing apps wasn't a possibility for most folks. Learning a programming language is hard work. That put the barrier to entry higher than most folks had the resources to reach, even if they had the interest.
AI is lowering the barrier. In the same way adding cameras to phones lowered the barrier to entry for photography, more folks can now make apps. The question becomes one of quality. In the right hands and the right circumstances, a phone's camera can create amazing images. TBD on if AI will have a similar trait.
Moving Forward
I love the idea that AI tools can open provide more folks with the ability to control their devices in the same way shifting from film to digital allowed more folks to create photos.
I don't like the idea of the developer industry getting decimated the same way photo industry did. But, if that's what it takes to give us control over the devices that are so pivotal to our lives, the cost will be worth it.
And, if the quality is there, there will be no way to stop it.
-a
Endnotes
This piece took a bit of a turn from what I originally intended. A natural consequence of using writing to help me think.
It was going to be more of a personal note. Talking about how I like the idea that I can just prompt an AI and have it generate exactly what I want. How appealing that promise is. How I hope we'll get there some day, even as much as I doubt it.
The question I'm interested in now is what the future of non-developers making apps will look like. I've heard anecdotes of folks with no real programming experience vibe coding apps.
In 10 years, how close will we be to Star Trek level machines?
Beyond the quality of the code, all the environmental/resource issues would need to be solved as well for long term viable use. Those details feel lost in the hype for most folks.
I've never vibe coded. I've used AI tools a few times as a stand-in search engine. That's about it.
I'm interested to give it a try though. I want to see what it's actually like.
Footnotes
My website building app.