On ChatGPT reliability:
Honestly, it's hit or miss. It's great for getting ideas or writing basic structure, but it definitely gets things wrong sometimes; especially with WE-specific stuff like rawcodes, patch differences, or weird GUI interactions. I've seen it recommend functions that don't exist in older patches or suggest things that sound logical but break in-game.
I usually use it as a starting point-idea brainstorming, then test and tweak. You still need to understand what it's doing, otherwise you'll end up debugging its mistakes xd Also DeepSeek is a good alternative.
It's more about guiding it. You can give it context upfront (patch, GUI/JASS, what you want) and if it gives a bad answer, correct it and ask again. It does learn within the conversation. The more specific you are, the better it gets.
So yeah, it's a useful tool, but don't blindly trust it. Test everything.
Beware the living(and AI)!