3 ms·
> AI will never know the why behind the architecture…
That's true only if you don't provide that context. The answer is: Do provide that context. My experience
> AI will never know the why behind the architecture…
That's true only if you don't provide that context. The answer is: Do provide that context. My experience is that LLM output will be influenced and improved by the why's you provide.
it takes longer to explain the context to the model than it does to just write the code based on the context I already understand, especially since code is more terse than natural language
But wouldn’t the context also be useful, in written form, to colleagues?
Definitely, iff you have to provide the context with every task. If agent memory worked better and across your whole team, then providing context might be much easier
if you know that context, you don't need a codemap
this is possible if you have a couple two-pizza teams. beyond that, good luck.
As you just said, codemaps don't include the "why" behind the architecture. That's context you need to add.