I have yet to see an LLM build and debug something complex. Barely stringing TODO apps, sure. Still need a competent human to oversee. Hallucination are a seri

I have yet to see an LLM build and debug something complex. Barely stringing TODO apps, sure. Still need a competent human to oversee. Hallucination are a serious problem. Without symbolic reasoning, LLMs quickly start to fall apart due to context limits and being able to know what exactly is wrong and needs to be changed.