Copilot - not convinced. My experience after a few months is that the suggested code disturbs my line of thought, and for complex code I'd rather it just fucked off and stopped "helping".

It will dumb down developers, mark my words.

Does it act like word suggestions on texting app?

Yup, that’s exactly my worries that you confirm here.

Coding is thinking thousand lines ahead. copilot only does a few. It's really impressive stuff, but it makes you from a typer a reviewer. The only thing it really does - is making you type less in SOME cases. But I like to type, it's smth to do while i think. Also, my typing is more accurate than the AI proposals while they're often surprisingly good, like very, based on your code style and stuff.

@demolishun I advise you to try codeium. It autocompletes setting every new field to zero after a malloc and completes the same malloc line you write everytime in your style used in rest of project.

I stopped using it - as I said - typing is smth to do when i think. I can imagine other people liking it

Dumbing down developers would be terrible since we're already on the level that people don't care how the source works anymore. We already accepted type stuff we don't understand. Your frontend has a freaking build process - you can only use that happily if you don't care. React would've not still be used if the average programmer wasn't an idiot. I really mean this. People feel themselves skilled due React because - YES - IT WAS A LOT TO LEARN. That's the result of bad design!

It very much will dumb down developers and prevent upcoming devs from learning anything, making them useless when it comes to troubleshooting and bugfixing.

It will limit the ones that rely on it to only being able to do what Microsoft and Google allow them to by preventing them from knowing how to do anything more.

It's a very slick way of killing competition. Lobotomize the next generation into not even being able to compete.

The few times I've used it, it was an impediment. Even if it wasn't, I'd have to look through every line of code to make sure it was right, so why not write it myself to begin with?

@cuddlyogre that's why maybe AI will always be limited. We need to control if the output is valid. If we can't, we don't use it. The moment, we take its advice without reviewing it, maybe AI can grow bit past us with certain stuff.. But we wouldn't know. We call it hallucinating probably 😂

Maybe pre-AI dev will become smth that will raise your salary or the opposite

You can hit escape. 🤷‍♂️

I think there’s too much hype.

Skilled will still recognize skilled, and half ass will still look half ass. All at some new productivity offset.

I find reading code way longer than writing code. On my IDE, I only use the features that have a deterministic outcome because otherwise I end up loosing more time than anything else.
I saw a few colleagues using it and for every line it’s writing you have to verify if there is not a typo. Then you get lost because you don’t know what lines you changed and what lines copilot you changed so you have to reverify all over again.

It can happen that I use ChatGPT on the side for things that I am too lazy searching in docs, but it’s more to have an idea. Sometimes it’s also lost time. Generally, when you know what you are doing, AI are just another hassle you have to deal with.