@Jada yeah AI was really shitty, but id still use it for very simple mundane work, like writing some code 100 times or something like that. And i got tired of fixing its code, so i started giving it instructions on how to write code. When i started using claude i used these instructions from the start.
And yeah its not good at super complex work. Even in coding, it fails if ur project is huge and u dont set it up properly and even if u do, u still need to be careful and think about how to use it. At work i have a far more complex project and it fucks up way more often than at my personal projects. It is like writing pseudo code: you need to know what you want written, it just does it faster, you can't rely on it to do research for things you don't know. It can research for things you know you also would have to research and you know the most likely possible outcomes. But if its something totally new or unknow, its far more likely to fail. You have to review its work, always. It will work 90% of the time and then fuck it up completely 10%.
It is a great tool, but thats all it is.
btw copilot used to use chat gpt. now they use diff models, but there is no copilot model.
@Spatail
Ah i see what u mean now. I didnt know it was that bad tbh. I had noticed it was lagging my pc if i have claude run while i play a video game and I knew its from the claude code, but I didnt look deeper into why, because it was a small annoyance and I usually cant really use claude while i game (it requires my attention and focus still), it would happen rarely.
Though I also teach them about the dangers of using AI, which is totally missing from this thread cause of the hype.Really? No one?
what are the dangers really? AI is retarded. You have to verify its work. And you need to know how AI works so you know how to set it up and what to tell it. Like how context works and the quirks of your model, so you can prompt it better. Like making unit tests and reminding it of rules sometimes. Like making comments, having version history, having good documentation, so AI can read it and knows whats what. Or when AI struggles to document it so it wont struggle next time. And giving it more tools, like a headless browser, or ffmpeg, or a pdf reader, so it can do its job more efficiently. Or writing skills and plugins so you dont have to prompt the same thing 100 times. Or when and how to use sub agents...
Like i said, it is its own skill to use AI properly.
And for googling with it, i can write the same paragraph of issues and workarounds and workflows. Tho it is simpler.