I think it depends on the code you need. If the AI is going to simulate human coding abilities, you are going to have the same problems as with human developers but faster: the more code you have, the bigger and more immutable your solution will become.

Even if you specified the solution in Gherkin and drawn all the diagrams, imagine the coordination effort needed for real developers.

But I think you are way better off than the rest of us because you actually have the concept in your head. I often donβ€˜t. I start trying different solutions to see how they feel. I guess copy-writing is similar.

Reply to this note

Please Login to reply.

Discussion

I actually suffer the opposite problem. The entire system exists in my head, right down to command line help and button shapes and menu items, and I can actually try everything out in my head.

But the software just exists in my head. πŸ˜‚

It's like hacking on imaginary software in a dream.

I am then bizarrely fast, tho, as there are no false-starts or experiments. Already did all of that.

Type the whole thing out rapidly, including the unit and integration tests.

I guess I should start doing katas πŸ˜‡

Shadowboxing, but make it code.