Can LLMs write better code if you keep asking them to “write better code”?

“After benchmarking, the algorithm is extremely fast, resulting in about 6 milliseconds a run, or a 100x speedup.”

Repeating “write better code” took a Python function from 657 milliseconds to 6. The casual approach also produced more bugs than a properly engineered prompt. You still need an engineer to tell which changes are improvements.