"sufficient detail, and if done correctly" -> the machine will once again be able to interpret your intention ...

This does not actually follow from the way LLMs work.