I made my living for 15 years by telling computers exactly how to do things, step-by-step. That was called “programming”. Yesterday “I” rewrote my backup script. I had written it the “old school” way, relying on fairly detailed knowledge of bash scripting, which is notoriously full of pitfalls and irregularities that bite you. I went several rounds with “the AIs” and re-wrote it, ground up. I told it what I wanted it to do and got a clean, full featured, well commented script with error checking, logging and documented usage. Poof. I had tried to have it clean up my original script, and it did OK, but there was too much confusion there. Starting over with “what” and not “how” was the right call. I did carefully review the generated script before turning it loose on my files, watched the first few runs closely and manually verified that the right files were indeed backed up.

I think this is the shape of a lot of the AI interaction to come. Gone are the days where it makes sense for a lot of people to memorize DEC10 assembler, learn BLISS, be able to quote chapter and verse from “K&R C” or the Perl “camel book”. To be sure, there are still many compulsive, detail-oriented people in the world (me) who were and are good at that sort of thing, but there are growing areas where we no longer have to focus that human talent. In my personal life, I’m trying to pivot to find more productive ways to apply that obsessive focus and good memory for minutiae. “Let’s see, how do I pass variables in FORTRAN-77 common blocks again?…”

Early in my career, people called what we did a “Small Matter Of Programming” (SMOP). This usually came from people who did not have the chops to do the focused, detail-oriented work we specialized in. We knew that getting the “how” right was not small. But now with AI becoming much better at some of the “how”, it’s looking like the “SMOP” charge is becoming more real than we like to admit. Maybe now we will be able to free up more of the world’s creative talent to work at a higher level, thinking up the important “what” questions and leaving more of “how” to the machines that are better at it, while we continue to move ever onward in the quest to learn more of what it means to be human.

All this from a broken bash script? Are there “what” questions you’ve been putting off because you were not sure you knew enough “how”?