These comments may seem silly at a glance, like joking about using WebMD instead of going to the doctor. But if this is the kind of unhealthy over-reliance on AI that OpenAI considers “cool,” we should all be worried.
These comments may seem silly at a glance, like joking about using WebMD instead of going to the doctor. But if this is the kind of unhealthy over-reliance on AI that OpenAI considers “cool,” we should all be worried.
it’s better if you have it working iteratively. The newer command line tools allow it to compile the code and fix errors itself. There are also some techniques you can use so it does a better job of keeping track of what it’s doing. Generally they still suck at large existing codebases but they’re pretty good if you are doing something greenfield like a prototype or using the existing project as a dependency rather than trying to modify the existing project.
I’ll have to have a look into that. If nothing else, it has been good at collecting my thoughts and giving me a clear idea of how not to do things.
It has some good suggestions at time also. Don’t just discount it “because”.
You need to use your critical thinking skill with this as much as everywhere else.
So I’ve tried it, don’t think it is worthwhile, and don’t think it is ultimately a good use of humanity’s resources, and you’re telling me that I’m not “critically thinking” because I’m not a hardcore true believer?
Jesus, what the fuck happened to Lemmygrad since I left? I knew it was pushing in the direction of AI stuff, but I didn’t think it would get so full of smug techbro bullshit so soon.
the thing i’ve found that works best is telling it to keep a scratchpad of what it’s doing and what it’s learned and a few other things. You have to tell it explicitly it has to keep it updated otherwise it’ll neglect the scratchpad and it’ll be a disaster when you go to recover from it forgetting everything.
You can ask it to make a memory dump in text form also.
Chatgtp also has a memory feature… It only gets full easily!