interesting attempt but would need the next generation of LLMs to be trained on memelang code and it's documentations so we don't waste more tokens in context windows trying to instruct the model on how to use memelang to save a smaller amount than we waste.