As someone else in comments mentioned though, one alternative strat is to describe your problem to the LLM using pseudocode or napkin code that you at least don

As someone else in comments mentioned though, one alternative strat is to describe your problem to the LLM using pseudocode or napkin code that you at least don't have to link every last inch of to get it to function. And then of course the LLM's reply is going to revolve around an iteration upon the code that you wrote (often with most of that linting done for you for free).