alright, /g/. I had another brainfartdesign patterns are supposed to make code easier to maintain, read and understand - for humans and indians alikebut future code is gonna mainly be maintained, read and understood by clankersso my question is, do we keep the old patterns because those are what the models are trained on and what we understand, or do we(or they) make up new patterns that might be horrible for a human to work with but more optimized for llm's?
>>109394882>but future code is gonna mainly be maintained, read and understood by clankersOver my dead body.
>>109394889That’s the idea, yeah.
>>109394882>Design patterns are a huge mistake. NEVER decide on how you should solve a problem before you have solved that problem. Also, knowledge about design patters don't translate to good code. (Image of Casey Muratory)Btw, LLMs prefer spaghetti code that looks like decompiled bullshit, or code that looks like free software code, stripped from github without the comments.
>>109394882what a bizarre image
>>109394918Got eeem
>>109394882Heaps of stuff in computer science is good at solving big data problems, but horrible to understand. Binary trees? Hash tables?It comes down to the deeper question, can LLMs invent anything new, or just mash together the collection of human knowledge and see what falls out? I think it remains to be seen.Optimised for LLMs just means in the lexicon and small enough to fit in the context I guess?
>>109394882
