Comment by throwaway85825

That's the push factor. The pull is, say there's a giant repo like windows or Adobe. LLMs produce spaghetti code because the repo doesnt fit in the context and it wasn't in the training set. If Adobe fine tuned it the LLM would be much more useful.