For a longer changes (especially refactorings) it is useful to tell the llm to write a plan. I iterate over it until it is reasonable and save it as a kind of meta program. Then I instruct it to implement it step by step.
Sometimes I need to update the plan or even there are cases where the change I imagined is not really useful and I abandon the plan.
I have feeling that this meta-programming will get important. There might be many techniques to it - like writing pseudocode and then rewriting it in real code or various ways of adding tests.
For a longer changes (especially refactorings) it is useful to tell the llm to write a plan. I iterate over it until it is reasonable and save it as a kind of meta program. Then I instruct it to implement it step by step.
Sometimes I need to update the plan or even there are cases where the change I imagined is not really useful and I abandon the plan.
I have feeling that this meta-programming will get important. There might be many techniques to it - like writing pseudocode and then rewriting it in real code or various ways of adding tests.
After listening to your conversation https://newsletter.pragmaticengineer.com/p/ai-tools-for-software-engineers-simon-willison
I wanted to write this exact how to but found you had already done an excellent job. Thank you!
Thank you. I should document too