>>112265
If you're an aspiring programmer, the most important part is learning to actually program. An AI can churn out code, but if you don't know how to parse what it spits out, you won't be able to identify the nature of the bugs you find or figure out how to fix them, short of trying to feed back the broken code to the AI that dumped it on you in the first place and hoping it somehow does better this time.
SH's code isn't necessarily bad, but it's a bunch of raw javascript and HTML with some semblance of formatting. No pre-existing frameworks, no javascript libraries, etc. There's some oddball design decisions, like having every key/value pair regarding personalities or jobs or interests using the full string as the key, rather than variables. This'd make localizing to other languages a pain in the ass, and even works against the dev as changing one string (changing the personality name of 'Bubbly' to 'Upbeat' or something similar) would mean changing it more than 200 times in the code, rather than just changing the text string associated with a matching variable once and leaving the variable name as-is across all the other references.
The portrait art system is actually a clever solution, even if it's not something most people would think to do. If you want to insert and layer premade images, you're including 13 different face and body sizes, thirty hair types (some with front and back layers), five eye types (with each eye needing separate handling as the game allows heterochromia) , plus other stuff like dimples, freckles, glasses, piercings, etc. That's close to a hundred images, and if they're pre-colored that quickly becomes thousands due to the number of color/tan options for skin and color options for anything else. Even if you instead had them all in grayscale and attempted to tint each layer as you go, and presented the final result of that, you'd still want to manually test a large number of the combinations to ensure that the tints come out right, don't conflict, etc., and depending on the size of each image it could still be a pretty big bloat on file size as they'd likely be encoded as a base64 data string, which would increase the final size of each image by more than a third - though realistically, at the current 40x40 render size, some grayscale images might still be smaller than the ASCII art, depending on encoding.