No matter how many times I scold it and despite the fact that I have it in my preferences in the main settings, GPT continues to put #comments in code. Yet it will do it on command when I repeat myself.
No matter how many times I scold it and despite the fact that I have it in my preferences in the main settings, GPT continues to remove my API keys from code. Yet it will do it on command when I repeat myself.
Consistently removes colorama font and emojis from code for no reason, knowing it was there despite being instructed not to remove or omit any code that wasn’t discussed.
These things are frustrating and waste a fair amount of time. I don’t need to be policed by it.
It’s important to note that most LLMs often struggle with understanding negative logic, such as “do not” or “don’t” in prompts. When encountering such issues, it’s advisable to replace these phrases with “avoid.”
Regarding code changes with LLMs, there isn’t much that a prompt can influence at this stage. This limitation seems to be inherent at the moment and may change in the future, perhaps in a year or more. Note: I am not an OpenAI employee.
Coding simple tasks with LLMs, despite the presence of alpha and beta bugs and missing features, is still considered a step forward. The workarounds required are often worth the benefit of faster coding.
That’s an interesting insight. It’s just odd that it will do it on command but not in general. I managed to make a complex day trading bot and now I am working on a project for Kraken/crypto. I learned about the stock market backwards using GPT to create it in VS Code from scratch. It’s been quite an experience.
