I am (or at least
was) an experienced assembly programmer. I'm familiar with numerous obscure x86 opcodes that take dozens of machine cycles to execute.
This knowledge comes in handy when reading other peoples' code, but I don't use those instructions myself. Why? Because I know how to achieve the same effect through carefully-scheduled combinations of smaller, faster opcodes.
Exactly the same principle applies in English composition. Use an awkward or unusual word, or a flowery modifier? That's an L1 cache miss. It'll cost a few cycles of wasted execution time, if viewed in terms of the reader's attention span. How about a word that sends the reader to the dictionary? That's an L2 miss. Those can waste thousands of cycles.
You can tolerate a few L1 misses here and there, but those L2 misses that have to be fetched from DRAM will make your "code" run like shit.