Software is architecture defined in code. Minimal common sense reasoning is definitely not enough to write and maintain huge software codebases. And LLMs pass even these reasoning tests with "n-th order of understanding". Writing snippets is one thing, but forward-and reverse-engineering of complex problems is another because the number of possible ways to achieve the same result grows exponentially, but evaluating the optimality of each solution is another task different from what LLM does.