In my experience, GPT-4 does get lots of textbook/common problems right in common languages on first try. But when I tried to solve a just mildly complex proble

In my experience, GPT-4 does get lots of textbook/common problems right in common languages on first try. But when I tried to solve a just mildly complex problem (finding a ring of cells in a Godot GridMap using GDScript), it produced nonsense: Solved the problem wrong, called functions that don't exist and so on. It did correct some of the mistakes after I pointed them out, but I couldn't get something working without manual editing. Then I asked it to optimise the solution and got nonsense again. For simple data wrangling in Python it worked a lot better. I have a suspicion there might be some... showmanship at work where all kinds of test problems and common tasks are made sure to work well. Also it generates code slower than I would type (if I know what to write), which makes me mad for some reason.