I'm still surprised sometimes at how rapidly I can prototype something. Is it always shippable? No. But something that would've taken me an amount of time I can't afford to is much quicker.
This past weekend I decided to build an iOS app as a personal project. Normally I wouldn't have the time to work on one like that but I was able to get a working POC that is actually usable for my trivial use case in a matter of hours.
Because there's people who care about the craft, and there's people who don't.
I don't see any kind of AI-assisted coding as making anything. You didn't build that iOS app, Claude did. Maybe you can delude yourself into feeling some kind of pride or contentment in it and somehow think you built it but a lot of us who actually have a passion for the domain can't.
Imagine you're a carpenter. You've spent decades honing your skills, countless incomplete or borked projects that were recycled for newer better ones, the skin on your fingers and hands have become tougher and rough like sand paper due to how the majority of your waking hours are spent manipulating wood.
Now, someone gives that carpenter a machine that you can tell it your idea for any wood object and it just spits it out in a couple minutes and charges you a few bucks. The carpenter says, out of curiosity, "make me the roof in the Chapel of San Idelfonso", and after a little while, it spits out a nearly perfect copy of that magnum opus.
Any real carpenter would probably be disgusted, or at least feel a deep void inside them and nothing but disinterest in the machine. They'd just go back to their workshop and leave the machine outside. But that's probably because nobody becomes a carpenter to chase the current hype thing and to make money; they have passion, just like our engineers used to.
Disingenuous analogy, because you're taking the work of a carpenter and then jumping all the way to a Star Trek style replicator, which is daft.
The jump from writing code by hand to having a machine learning algorithm compute code for you is not the same as that. We had tags, syntax highlighting, auto-completion, word suggestion (via dictionary), intellisense, enhanced auto-completions via snippets, natural language processing, language servers, etc. etc. throughout the decades and an LLM is still not a Star Trek style replicator.
Now, it might have taken a blacksmith a hell of a long time to work various kinds of metal to order, but then someone got the idea of creating a mould so you could make more of them at once. Then later on technology advances and you can now machine it to ridiculously high precision with CNC. Now the person who writes CNC code for a high precision part might not be the same person who has to work another metal part manually.
Did the machines do that or did they? And why exactly does that distinction matter? Half the code I write depends on stuff hundreds or thousands of others have built before me.
spike021 · · focus · HN ↗
I'm still surprised sometimes at how rapidly I can prototype something. Is it always shippable? No. But something that would've taken me an amount of time I can't afford to is much quicker.
This past weekend I decided to build an iOS app as a personal project. Normally I wouldn't have the time to work on one like that but I was able to get a working POC that is actually usable for my trivial use case in a matter of hours.
I'd call that a victory.
kode-targz · · focus · HN ↗
I don't see any kind of AI-assisted coding as making anything. You didn't build that iOS app, Claude did. Maybe you can delude yourself into feeling some kind of pride or contentment in it and somehow think you built it but a lot of us who actually have a passion for the domain can't.
Imagine you're a carpenter. You've spent decades honing your skills, countless incomplete or borked projects that were recycled for newer better ones, the skin on your fingers and hands have become tougher and rough like sand paper due to how the majority of your waking hours are spent manipulating wood. Now, someone gives that carpenter a machine that you can tell it your idea for any wood object and it just spits it out in a couple minutes and charges you a few bucks. The carpenter says, out of curiosity, "make me the roof in the Chapel of San Idelfonso", and after a little while, it spits out a nearly perfect copy of that magnum opus. Any real carpenter would probably be disgusted, or at least feel a deep void inside them and nothing but disinterest in the machine. They'd just go back to their workshop and leave the machine outside. But that's probably because nobody becomes a carpenter to chase the current hype thing and to make money; they have passion, just like our engineers used to.
ljm · · focus · HN ↗
The jump from writing code by hand to having a machine learning algorithm compute code for you is not the same as that. We had tags, syntax highlighting, auto-completion, word suggestion (via dictionary), intellisense, enhanced auto-completions via snippets, natural language processing, language servers, etc. etc. throughout the decades and an LLM is still not a Star Trek style replicator.
Now, it might have taken a blacksmith a hell of a long time to work various kinds of metal to order, but then someone got the idea of creating a mould so you could make more of them at once. Then later on technology advances and you can now machine it to ridiculously high precision with CNC. Now the person who writes CNC code for a high precision part might not be the same person who has to work another metal part manually.
Did the machines do that or did they? And why exactly does that distinction matter? Half the code I write depends on stuff hundreds or thousands of others have built before me.