3 ms·
> There is too much code and complexity to check if the changes are ok To me, the challenge in software is to reduce this complexity to keep it manageable. Pro
by DWe1 1mo ago
> There is too much code and complexity to check if the changes are ok
To me, the challenge in software is to reduce this complexity to keep it manageable. Probably AI can manage more complexity than a human being, but it seems to be difficult to prevent or reduce that complexity in the first place. How is your experience once the complexity is too hard for a frontier LLM to handle, or for you to prompt the correct direction towards it?
- XCSme 1mo agoI basically stop caring about the code, never look at it, and just tell it what to be aware of from my experience. Tell it stuff like "use caching", "use library X", "make sure that it works when offline", etc. And then basically live test it and see if it works as expected, and tell it more improvements/changes to make.