You can compile Mini vMac without checksum verification (either yourself or with the variations service[1]), which will allow you to use unknown or completely custom ROMs, though you need to be aware that it patches the ROM (it doesn’t emulate the original floppy hardware; instead it pokes a custom driver in where the original one ought to be) so if your ROM doesn’t line up with the original you will have problems with random chunks being overwritten.
I’ve done this so I could use Mini vMac to learn assembly language: the Mac has the convenient property that pixels on the CRT are 1:1 the contents of a chunk of RAM at a fixed offset, so you can get very immediate visual feedback about what your program is doing. I just set up an assembler to dump raw machine code and named it "Mac128K.ROM" and Mini vMac picked up on it fine.
[1] https://www.gryphel.com/c/minivmac/var_serv.html - though since Paul Pratt disappeared a few years ago nobody is quite sure how the server is staying up
It was interesting to discover that the Apple II Plus' ROM didn't support Kanji, but there were third-party add-on cards that added Kanji support. The Apple II Plus I found had a Multitech Kanji Card with it. Multitech later became Acer.
If you ever decide to part with it let me know :)
I wonder if one could put this larger ROM, and the other files into a custom built image so no swaps are required.
They had enough room left in the 512KB ROM to fit a 357KB boot disk a stripped down System 6.0.3 and a few useful tools (MacsBug and AppleShare Prep)
They are also quite good at translating poorly written and only semi coherent writing, which can be incredibly useful if the person you are communicating with is quite sloppy.
The whole LLM scene today came about because context was really important to translations. The "attention is all you need" paper was by the Google Translation team as they came up with ideas to improve how to map context of words and carry them across in translations.
At some point people started asking the translation to "translate from English to English as if you're an AI assistant".
Anyway it shouldn't surprise anyone that LLMs are good at translation. The real surprise to everyone is how powerful translation engines that understood context could be!
The translation transformer also was able to peek ahead in the context window while (most?) LLM's now only consider previous tokens.
You see this with recent automated translation on YouTube. If the creator of (say) an English-language video doesn't upload subtitles, YouTube automatically creates them based on the audio, but they lack punctuation and have nonsense phrases. The AI-driven translation of those subtitles to other languages cleans up the text along the way, so the end result is that non-English speakers get better subtitles than English speakers.