[bcjdjsndon]

> Even extremely slow LLMs can generate Part B faster than I can audit Part A.

Depends what part a and b are...because workloads are different