Are LLMs even worth the hype when it comes to syntax? Most languages have LSPs and editors can easily be configured to autocomplete with them. No (skilled) developer should struggle with syntax in a language that they already understand. The argument that "AI replaces the need to memorize syntax" is one of the most idiotic arguments for using AI assistants.
They're decent for writing boiler plate code, but code snippets are (imho) just as useful.
LLMs have helped me as a research assistant, especially working with new libraries. It's nice to use them to test different approaches. But in the end I've found that the more I use AI, the more time I waste. In a professional setting, generative AI should be used more like salt. Too much salt ruins a dish.
An even better analogy is a leavening agent (yeast/baking soda). Too much and your bread implodes. Unleavened bread is vital for certain religious sacraments.
Honestly you’re not even wrong. When I want back to writing code for the hairy refactor only I could do, I was like, wait, this is what I was afraid of going back to? It’s so easy! And fun!
What sucks is debugging dumb issues. That’s what burns devs out. Claude is a lifesaver for that.
It definitely can be a lifesaver. But the better you are at swimming, the less likely that you'll need a lifesaver when you hop in the water ;)
We'll see where things go. I'm still placing my bets on manual coding being "an elegant weapon for a more civilized time." Luke still used a blaster when he needed it though.
The word online is that you posted a video that *maybe* LLMs are good and you lost half your viewers. If that’s true, I think you should stick to your guns. I’m an AI bear, but I still want to know every side of the argument. And sometimes I wonder if I’m wrong.
If your takes are honest and based in facts, you’re right. And history will be on your side. And maybe you’ll find a whole new audience in the process. Nobody is going to remember those that got it wrong.
Personally I’m drowning in skeptics, and not interested in unrealistic booster takes. It would be refreshing to hear something different from time to time. Presumably there’s room for moments of grey?
LLMs are what, 4 years old? And we’re discussing how well they are coding when them coding wasn’t a thing like 2 years ago? Have some perspective. It’s like seeing the first portable phones and saying there will never be a mobile phone market.
Yup. Actually Codex is much better than Claude at stopping if you tell it to stop in the face of uncertainty. So there is hope. But still, it's not 100%. It's not even deterministic. That may be the root problem.
Are LLMs even worth the hype when it comes to syntax? Most languages have LSPs and editors can easily be configured to autocomplete with them. No (skilled) developer should struggle with syntax in a language that they already understand. The argument that "AI replaces the need to memorize syntax" is one of the most idiotic arguments for using AI assistants.
They're decent for writing boiler plate code, but code snippets are (imho) just as useful.
LLMs have helped me as a research assistant, especially working with new libraries. It's nice to use them to test different approaches. But in the end I've found that the more I use AI, the more time I waste. In a professional setting, generative AI should be used more like salt. Too much salt ruins a dish.
An even better analogy is a leavening agent (yeast/baking soda). Too much and your bread implodes. Unleavened bread is vital for certain religious sacraments.
Honestly you’re not even wrong. When I want back to writing code for the hairy refactor only I could do, I was like, wait, this is what I was afraid of going back to? It’s so easy! And fun!
What sucks is debugging dumb issues. That’s what burns devs out. Claude is a lifesaver for that.
It definitely can be a lifesaver. But the better you are at swimming, the less likely that you'll need a lifesaver when you hop in the water ;)
We'll see where things go. I'm still placing my bets on manual coding being "an elegant weapon for a more civilized time." Luke still used a blaster when he needed it though.
That was my goal within all my coding classes, elegant code. It’s easier to maintain.
When I tell people that vibecoding is not for everyone, this is what I’m talking about.
AI companies convinced everyone they can write code. I have 22 versions of an app that will probably never work.
Not everyone can write code, no matter what AI says.
You still have to tell it to not make dumb shit. Left on its own it makes silly shit.
There’s no skill in taking a flight from NY to LA either, but it’s a damn useful service.
The word online is that you posted a video that *maybe* LLMs are good and you lost half your viewers. If that’s true, I think you should stick to your guns. I’m an AI bear, but I still want to know every side of the argument. And sometimes I wonder if I’m wrong.
If your takes are honest and based in facts, you’re right. And history will be on your side. And maybe you’ll find a whole new audience in the process. Nobody is going to remember those that got it wrong.
Personally I’m drowning in skeptics, and not interested in unrealistic booster takes. It would be refreshing to hear something different from time to time. Presumably there’s room for moments of grey?
LLMs are what, 4 years old? And we’re discussing how well they are coding when them coding wasn’t a thing like 2 years ago? Have some perspective. It’s like seeing the first portable phones and saying there will never be a mobile phone market.
Yup. Actually Codex is much better than Claude at stopping if you tell it to stop in the face of uncertainty. So there is hope. But still, it's not 100%. It's not even deterministic. That may be the root problem.