It wasn’t writing the code generated by AI, but explaining it afterwards.
by•
AI allowed us to transform our ideas into prototype significantly faster than ever before.
The issue arose a couple of weeks later when we started making changes.
The simple feature change needed an explanation of the relation between several pieces of code that didn’t seem related to each other at all.
The code did what it was supposed to. The context wasn’t clear.
We now make documentation about the reasoning of unusual choices made by AI, rather than about the lines of code. Just one note on the purpose of something.
It’s already helping us with future changes since we preserve the logic, not just the solution.
If you are actively developing software using AI, how are you keeping your reasoning from being lost?
16 views
Replies
the missing context can become a bigger problem than the code itself. I’ve found that understanding why something exists makes future changes much easier than simply knowing what each function does.
@isabella_wandrei Documenting unusual AI decisions sounds useful to me. A short note explaining the intention could save a lot of time when someone revisits that code weeks later.
I have started keeping short notes on the reasoning behind AI generated code too It makes future changes much easier when I need to understand why something was built that way