We're in this situation because of laziness. I doubt doubling down on it is the solution.
replyProgrammers will just /goal the LLM to make the code more memory efficient. Then let it run it run overnight.
replyAnd it will spend 1 million tokens finding the most memory efficient program:
replyint main(){}