Happy Fakesgiving! (Chicago 56, 73, and 78 removed)
After several reports, I have taken a look at certain Chicago builds
I have determined that Chicago Build 56, 73 (no "g"), and 78 are fake, and the files have been removed.
Since no one else has been arsed to post details about WHY these are fake, here are my finding.
Chicago "4.00.56"
First of all, I should point out that it IS possible for a pre-release software build to have only minor changes between builds if it was rebuilt to address a small important last minute testing bug. But that doesn't look like the case here.
Some file dates have been touched and set to 7/1/1993 3:00pm. Touching the dates seems odd for such an early development pre-release.
Latest dates in 58s are almost a month later (~8/9/1993), so there should be more differences.
Most files differ only in the version string.
Many other files are missing the version resource.
Some files, such as appui.dll, cardfile, cirrus.drv, framebuf.drv, regedit.exe, and s3.drv seem to have a few more random differences, but don't quite look code related.
Missing "mini" and "pluspack" folders.
Shell.inf seems to have an added reference to "msd.com"
Setupc.inf seems to have added references to "wfwnet.drv", and "win31.exe"
setupppp.inf and setup.inf have an added reference to "win31.exe"
winbug.dat removes a reference to DR DOS 6.0.
The command.com is the same size, but the string is not just hex edited. An entire block of text strings is shifted by one, and references to those stings are different.
Includes MSD.EXE (Microsoft Diagnostic) 2.01 instead of 3.00, that can be found in other published Microsoft products.
Of course the boot screen was edited. Microsoft obviously resized the image from a larger source with a random dither, yet in the word "July", the u, l, and y are identically copied from letters on the 58s's "Preliminary Release - August 1993", and the "J" looks hand drawn, lacking dither.
"This looks shopped. I can tell from some of the pixels and from seeing quite a few shops in my time" 
I am positive if I were to look even closer, I would find more damning details.
My conclusion is that it would be improbably lame of Microsoft to produce a build like this and with so few differences. Therefore:
Chicago "4.00.73"
The CD ISO container is not genuine, but neither is the one for 4.00.73g. It is embellished to make it bootable.
Most files are the same as 4.00.73g
No noticeable differences except for resource version strings
Of the few differences -
Different FDISK.EXE.
Buffers and Files line altered in setupc.inf.
Warning.bmp is different, but the image is the same. Re-saved as a compressed bitmap?
Most file dates are very close to 4.00.73g, but not all are the same.
Command.com is only different by one byte - they replace the "g" with a space. Lame.
Contains a home made readme.txt claiming that it is "buggier than build 73g".
No changes to boot logo.
Conclusion: Fake!
Chicago "4.00.78"
Allegedly this build was "incomplete" and "repaired" using pieces of 4.00.73g. That would not necessarily make it fake, but just crap.
However, I don't even see that level of difference.
First, the date stamps are totally borked. It never ceases to annoy me when I see some "preserved" 1980's or 1990s software where all of the files are dated 10/11/2014 8:10:12AM or somesuch. Usually, this is due to idiots posting files on a FTP or torrent without putting them in a zip archive. But in this case it looks like the time stamps were intentionally touched.
Looking at the files, they are almost all the same to 4.00.73g, with a few modifications.
USER.EXE is 100% identical except the version string has been hex edited.
The same with SETUPX.DLL
OLE2 files are different.
A few misc networking files are different.
Most of the DOS files are different, and there is an extra "SUPPLMNT" folder with some versions that don't match. (for example, two different Qbasic.exe files). Not sure off hand where these are from.
Oddly, many of the help files seem have some differences. Can't quite tell what the differences do, if anything.
Again, the warning.bmp file has been saved as a compressed bitmap.
Of course they changed the logo.sys boot logo - in crayon.
From "Preliminary Release - November 1993"
To "Preliminary Release - December 1993"
At a glance, the random dither in the first two "e"s in "December" are identical. Hmmmm.
If there are any genuine "4.00.78" files in there I'm not sure what they even are (the DOS files?)
Conclusion: FAKE!
In summary, I would like to give a big hearty MIDDLE FINGER to what ever stupid kid(s) went to so much trouble to produce these fake builds, tried to pass it off as real, and made me waste my time analyzing this crap.
And a big hearty WTF to certain other archives that hosted these for ages, and the thousands of people who have download this from here or elsewhere without questioning if it is authentic. If I hadn't pulled it apart and questioned it, a certain archive would probably still be clinging to their old fake Windows Premiere Edition.
It is also annoying that whoever previously determined these were fake did not publish more details somewhere where others could see them. Perhaps they thought it would prevent more fake builds? The problem is, I'm not going to remove a file just because someone parrots that it might be "fake" - I need some level of proof.
If anyone sees any other issues with any other pre-release builds, please report them.
So happy Fakesgiving from Winworld, your primary source for vintage software and HOPEFULLY not fake betas.
Comments
I'm glad that these fake builds are gone (even though I have a copy of 56 for testing).
The lengths beta kiddies will go to, eh?
We should have a "Hall of Shame" section containing all the fake stuff we've found so far. (with download links if possible)
Probably not, or some moron will pass them off as real to those gullible enough.
Maybe make it viewable to members only who have had an account for more than 180 days?