I used to quietly believe that being a good engineer meant being the fastest one to solve the ticket. Hand me a well-defined problem and I would race you to the green checkmark. That was my whole identity for a while.
Then the machines got really, really good at solving well-defined problems. I watched Claude close a bug in about the time it took me to finish reading the description. Small existential wobble. If the thing I was proud of is now a commodity, what exactly am I bringing?
Then something clicked, and it reorganized my head. The work that stays valuable is the work that cannot be graded by a loss function.
School optimized us for the wrong thing
Think about it. School is sixteen years of well-defined problems with an answer key in the back. You get very good at solving. Nobody grades you on picking which problem was worth solving in the first place, because the teacher already picked it for you.
Then you get a job and the answer key is gone. The hard part was never the solving. It was deciding what deserved to be solved at all.
Finding beats solving now
This is the big one. AI is a solving machine. Point it at a clearly stated problem and it will grind out something workable. So the scarce skill flips. The person who knows which bug actually matters, which feature will move the needle, which vague complaint from a customer is really an architecture problem, that person is worth more than ever. You become the one holding the compass, not the one swinging the pickaxe.
The last mile is the whole game
Here is the uncomfortable part. AI hands you a competent, slightly soulless median. Eighty percent there, and that eighty percent is now free for everyone. So the entire difference lives in the last mile. The taste. The polish. The “this still feels off” that you cannot quite articulate but chase down anyway. That final stretch is unglamorous and mostly grind, and it is the only part nobody can copy paste.
What’s actually scarce
Not code. Not compute. Those got cheap. What stayed scarce is your time, the trust of the people you work with, and your reputation. I spent years guarding my code like it was precious and treating relationships as a nice-to-have. I had it exactly backwards. I wrote before about how being cheap with your tools is a tax on that same time, and this is the bigger version of that idea.
The machines took the part of the job that had an answer key. Honestly, good riddance. What they left behind is the interesting part, the part that was always the actual job.
