I built a 2x faster lexer, then discovered I/O was the real bottleneck

Nice story; that's exactly why people always warn from the eval called "premature optimization without measurement".