As people created the most abstract way of programming, syntax knowledge of programming language importance is decreasing, still, you need to understand WHAT code does, but the more important thing, now you need to build architectures instead of raw code, and the best way to make efficient and fast system, is to understand how computer works inside. How does that relate to AI and LLMs? From my experiences, LLMs are extremely bad with huge code-bases, but frighteningly efficient with small tasks, good old divide and conquer, if you separate tasks and create modular and abstract enough architecture, that even newbie will understand, LLM can create perfect, edge-case proof code. Does that mean, that we went full cycle and returned back to need of software engineering, instead of direct code writing?
[link] [comments]