So interestingly that shows the Venus was IFF ILBM, which matches its origin as a demo image for Deluxe Paint rather than GraphiCraft.
The format for most of the others is described as "PLBM", which makes sense as P(L)BM is basically a cruder variation that'd allow dumping the screen/window content to the disc in planar format directly (native for all Amiga on-screen display prior to AGA or graphics cards) with minimal framing, while ILBM interleaves the bitplanes row by row (and can optionally use run-length encoding too), and so needs a bit more finesse.
It's kinda funny how they play up that the person who found this out "specializes in and is world-renowned for hacking and reverse engineering" - if you're remotely familiar with image file formats, it takes about 5 seconds of looking at these files in a hex editor to see that they're an IFF related format, and the chunk names are plain ASCII. I'm sure the person in smart - I'm not disputing that; but they're overplaying how much expertise is required for this.
Both ILBM and P(L)BM are formats you can write a basic decoder for in an hour or two - IFF contains chunk headers and length markers to let you extract each chunk, and you need the palette information and the bitplanes. ILBM also allows for basic run-length encoding. There's no complicated compression or anything else to make it hard.
The version of GraphiCraft used for the demo was not a finished version, and was known to be buggy. Recollections of people involved includes claims that they were terrified when Warhol used the flood fill, for example, because there was a known bug in flood fill that would cause it to fill memory outside the screen bitmap. So it's quite likely that there other bugs too.
As a wild guess, one possible source of problems might be that GraphiCraft might have taken shortcuts in saving that might have e.g. assumed a certain offset into some data structure to copy the palette information from the file to that changed before release, as some of these early programs were known to basically pretty much dump memory straight to file and load it back in straight into some data structure with minimal parsing and validation.
The format for most of the others is described as "PLBM", which makes sense as P(L)BM is basically a cruder variation that'd allow dumping the screen/window content to the disc in planar format directly (native for all Amiga on-screen display prior to AGA or graphics cards) with minimal framing, while ILBM interleaves the bitplanes row by row (and can optionally use run-length encoding too), and so needs a bit more finesse.
It's kinda funny how they play up that the person who found this out "specializes in and is world-renowned for hacking and reverse engineering" - if you're remotely familiar with image file formats, it takes about 5 seconds of looking at these files in a hex editor to see that they're an IFF related format, and the chunk names are plain ASCII. I'm sure the person in smart - I'm not disputing that; but they're overplaying how much expertise is required for this.
Both ILBM and P(L)BM are formats you can write a basic decoder for in an hour or two - IFF contains chunk headers and length markers to let you extract each chunk, and you need the palette information and the bitplanes. ILBM also allows for basic run-length encoding. There's no complicated compression or anything else to make it hard.