I don't understand that argument.

I.e. you want 3d party developers who want to do something in their spare time, to grab previous SDK, new SDK, and sit hours and compare manually dates, do diffs, check sizes, loose countless hours just to find out what and where was changed ? Imho that not very good. For us (who know what was changed), it will take just a few hours to do such a file, and we will save hours of productivity of 3d party devs.A diff is 100% accurate.
Of course, notice every single change will be hard for , but key-changes readme are must if we want to make life of 3d party developers easy. I.e. , until i with Trixie not post that info, who ever know what was changed except some smarties who check dates of archives ? Why they should download new SDK on top of old one, if they do not know what was changed ? I mean, its really obvious that with each SDK release there should come "what_new.txt", or something like this, where we can (at least briefly) note what was changed, updated and co. There will be no needs to write any single change in every autodoc and will be enough to say "fixed bucnh of typos in autdocs", but such a "big" updates like fresh mui includes, new dos sdk, new kernel functions and so on are 100% and must be noted, so developers will know what new they can do, without random guess work and loosing hours of time , which they can spend on actual coding with usage of new stuff. Better we will spend few hours one time, in compare that every 3d party developer will do the same and in sum we will have less hours of productivity.
Some note like "that is not full list of changes, we just describe major ones", can be added to the end of readme file, so ppls will know that there not everything noted, just some major moments.Anything a human tells you would be a lie to some degree because they will omit things.