Yeah we're trying to do the same, document things organically as we learn them on an internal Confluence. It helps, but we're piecing together fragments of kno

Yeah we're trying to do the same, document things organically as we learn them on an internal Confluence. It helps, but we're piecing together fragments of knowledge without context, so information density is low outside of specific tasks. It's enough to solve the same problem again, but not enough to gain much holistic knowledge of the software. And same as your situation most of the veterans don't bother with it (they already have the knowledge and are extremely busy). Doesn't help that some of the engine code dates back to the late 80s, was originally written on a mainframe for a compiler where variable names had character limits, so even if you know the surrounding conceptual terminology it's functionally indecipherable unless you have a lot of experience with it. It's hard to tell what it's doing or why, or even if it's relevant to the problem you're trying to debug. What we really need is something like a dedicated sprint where we stop the presses outside of critical maintenance and just write documentation, free up the veterans to write down their knowledge. But that would require the company to actually invest in what is otherwise seen (wrongfully IMO) as a cost-center. :P And honestly if we had that kind of money there's other software redundancies/inefficiencies that need to be resolved first, that would lower the documentation requirements in the first place.