I feel that a large divide in a programmer's perception of the utility of LLMs as an efficiency boost for coding is based on a large divide in specific skills. A very capable and experienced programmer, as I have often found in industry, may not be that fast at comprehending code that they did not write. On the other hand, I feel I am able to rapidly proof read code -- be it from my colleague in code review, stack overflow, or the output of an LLM. As a result, Copilot has let me write many Bash and Python scripts at breakneck speed. On the other hand, I have allowed LLMs to decide on the architecture of small programs before, and it has often created a garbage, unintuitive code structure. It is a tool that you must use to its strengths and your own personal strengths. If proof reading the code is slower than writing it for you personally, then it sounds wise to avoid the tool.