i appreciate the article and the full examples. But I have to say this all looks like a nightmare to me. Going back and forth in English with a slightly dumb computer that needs to be pestered constantly and hand-held through a process? This sounds really really painful.
Not to mention that the author is not really learning the underlying tech in a useful way. They may learn how to prompt to correct the mistakes the LLM makes, but if it was a nightmare to go through this process once, then dealing with repeating the same laborious walkthrough each time you want to do something with Docker or build a trivial REST API sounds like living in hell to me.
Glad this works for some folks. But this is not the way I want to interact with computers and build software.
> You're gonna get left in the dust by everyone else embracing LLMs.
Probably not, there's a very long tail to this sort of stuff, and there's plenty of programming to go around.
I'll chime in with your enthusiasm though. Like the author of the post, I've been using LLMs productively for quite a while now and in a similar style (and similarly skeptical about previous hype cycles).
LLMs are so useful, and it's fascinating to see how far people swing the opposite way on them. Such variable experiences, we're really at the absolute beginning of this whole thing (and the last time I said that to a group of friends there was a range of agreement/disagreement on that too!)
They’re certainly useful if you know what you’re doing. An example: if I try to create an application in .NET for Windows, I’ll have a hard time using an LLM cause I’ll have no way to know if the solutions are the best, what’s possible and what isn’t, etc.
But I’m an iOS developer who doesn’t have experience with SwiftUI. I’ve been creating an app clone for the purpose of learning it and I’ve been using ChatGPT extensively like one would use StackIverflow when you’re still picking up a new framework. It works very well and I’ve advanced very fast because I’ve read and watched multiple content about it, just never got into actually using it. It’s easy to know and even try out variations of what the LLM gives me. It feels like having a friend that knows SwIftUI which I can ask stupid questions as I try it out.
Not to mention that the author is not really learning the underlying tech in a useful way. They may learn how to prompt to correct the mistakes the LLM makes, but if it was a nightmare to go through this process once, then dealing with repeating the same laborious walkthrough each time you want to do something with Docker or build a trivial REST API sounds like living in hell to me.
Glad this works for some folks. But this is not the way I want to interact with computers and build software.