More than that I feel there is hidden/lost knowledge. For example real world software design documents are so hard to come by, I think they must be one of the closest guarded things in software companies. The history of major software development projects in the industry has simply been lost to us.
I’m pretty sure the main reasons such documents are not public are: Either nobody documented it or the document is badly written. So it is just too embarrassing.
> ...the document is badly written. So it is just too embarrassing.
At risk of being accused of being on the A.I. hype train, it seems like having a writing bot that could distill the important points out of poorly written documents, and state them more clearly would be a huge boon. Hopefully it would allow an iterative process where someone could hone their offering to a point where they felt okay releasing it.
For software documentation, “badly written” often means that it’s ambiguous, is missing information, and/or is contradictory. AI can’t really help with that.
I would think that a good AI could actually point out missing and ambiguous information and allow an iterative process where a programmer could more easily fill in the blanks before resubmitting it for another round.
Yes, ChatGPT is good enough to help writing this stuff. It is still work though. Work which is not valued much and which most engineers like to avoid.
Personally, I like it. For me, there is no better way to achieve clarity than writing about something. So I don’t really understand why they like to avoid it. They apparently value clarity less than I do but that doesn’t explain it. “Everybody is different“ doesn’t really explain anything.
I feel confident that an AI trying to read that info back would suffer the same limitations as any human. What is important depends completely on the audience and why they are reading. The sum of all important points to all people at all times is equal to the content of the computer program itself.
