Yeah, it can help with some things, but if you ask it to, say, convert a python app to Go, even with plenty of context it will give you hints without full implementation.
I suppose they favored concise answers in the fine tunning process and reused the model to train the 32k.
Things that I have been doing: paste full table DDL with entities and other classes involved and asking for some refactoring of a class. Also that and some stacktraces. It definitely helps, but the price (at least on azure) make me think twice before using.
Also azure playground is awful.
