I've been responsible for documentation in similar ways.
What I think is interesting here, is that with so much detail specified, tools could be created to write and check technical documents. I suspect with LLMs and the like becoming broadly available, that could also help.
On the other hand, I've seen technical documents that were accurate and correct, but they obfuscated more than clarified. One document I remember specified linker records with items like 0x54 instead of 'C' and end-of-record instead of '\n'. It seemed a little too pedantic instead of common sense.
From my experience technical documentation for aviation was made challenging by the corpus of engineering plans that were signed, initialed, and scanned into a PDF as an image. Modern machine learning to ingest that data and produce a diff would be very useful even if it still needed human review.
Comments
I've been responsible for documentation in similar ways.
What I think is interesting here, is that with so much detail specified, tools could be created to write and check technical documents. I suspect with LLMs and the like becoming broadly available, that could also help.
On the other hand, I've seen technical documents that were accurate and correct, but they obfuscated more than clarified. One document I remember specified linker records with items like 0x54 instead of 'C' and end-of-record instead of '\n'. It seemed a little too pedantic instead of common sense.
From my experience technical documentation for aviation was made challenging by the corpus of engineering plans that were signed, initialed, and scanned into a PDF as an image. Modern machine learning to ingest that data and produce a diff would be very useful even if it still needed human review.
edit: and by 0x54 I mean 'T' lol