Also infuriating is when the OP gets told by a mod to stop posting AI slop comments and OP responds:

I'm explaining the project in as detailed and clear a manner as possible - using the LLM to do it. The same one that helped me build it. Shut it down then man, why warn me?

It's hard to imagine turning an image into JSON would manage to compress it. 🤔

Here's my "symbolic image compressor":

American Gothic is a painting of two people standing in front of a farmhouse. One has a pitchfork.

Are you impressed? I used symbols to compress the image down to under a kilobyte. And I didn't need an AI.

Lol. It's funny to see vibecoders stumble

Reading OP and thinking about their misinformed understanding of what they are doing, I came upon an idea I propose to all of you: the almighty Babylonian Compression Algorythm.

As long as we have all combinations of (say, 256x256px) images in the database, we can cut down image size to just a reference to a file in said database.

It produces a bit-by-bit copy of any image without any compression, so it puts OOP's project to shame. Little, almost non-existent problem is having access to said database, bloated with every existing but also not-yet-existing image. But since OOP's solution depends on proprietary ChatGPT on someone else's server, we are on par there.

AI generated

"recursive symbolic"

Sounds about right