Thing cannot write computer programs

I've asserted elsewhere that LLMs are glorified chatbots. One of the comments in TFA mentions the classic chatbot, ELIZA, so it's nice to see someone else come to a similar conclusion. Additionally, it is ironic that LLMs are often used as the backends for actual chatbots; chatbots making chatbots; the circle is complete.

But back on topic, notice that jcoglan is complaining that the LLM can't actually follow his instructions; the LLM is doing everything it can to appear like it is following instructions, but never actually following instructions; the LLM appears to be wasting jcoglan's time in order to get as much money out of him as possible.

How are LLMs supposed to replace employees, when they can't follow instructions? The entire point of the manager-employee arrangement is that the employee tries to understand and follow instructions. I still don't know why any manager would want to deal with chatbots that can't follow instructions, when almost all employees have at least a little capacity to follow instructions.

If I'm wrong, and LLMs do somehow replace all employees, then companies will consist of only managers writing document-length instructions for every trivial task (to paraphrase a commenter in TFA). I dislike managers as much as the next guy, but that sounds like a punishment you wouldn't wish upon your worst enemies.