Training Up

At some point you end up managing someone who is better than you will ever be at the thing you are both meant to be good at. For me it has usually been a staff engineer. I can open one of their pull requests, read every line, and know that anything I might add to the code itself would just be noise. The insecure move is to leave a comment about a variable name anyway, just to have made a mark. I don't do that any more. Knowing exactly where my usefulness ends isn't the weak part of this job. Pretending it ends later than it does would be. The harder question is what that leaves me with. If I can't make this person better at their craft, and I can't, then what am I here for? This isn't really a new problem, though; it just doesn't usually happen in an office. Coaches are rarely the best players, and the ones who keep trying to prove they still have it get in the way. Whatever a good coach does, it isn't out-playing the person they coach. It is seeing what that person can't see from inside their own game, and handing them what they can't hand themselves. If I am honest about what I was ever really giving my strongest people, very little of it was ever the craft, whatever the title implied. It was mostly context they couldn't get from their own desk, an opinion on which problems were worth their time, and the occasional uncomfortable conversation about where they stood. I have understood that for a long time. What AI changed is that it removed the parts of the job that ran on me knowing more, so what was always the real work is now the only work. I still open their pull requests, but reviewing the work of someone faster than me, with a model at their elbow, is a habit now rather than a contribution. They used to get stuck and come to my desk. Now they ask the model. And the distance between us grows every year instead of shrinking, because the thing making them faster does nothing for the part of the job I had been leaning on. What is left is harder, and less flattering to me. It comes down to the responsibility that was always the real job underneath the technical one: making it possible for someone better than me to do the best work they are capable of. I have run my teams on that belief for years. Some of it is simply telling them the truth. Someone very good, carried along by their own success and a model that tells them every question is a good one, genuinely cannot tell how good they are, or what the next level would ask of them. I can, and making myself say it, before their success hardens into a blind spot, is a real part of what the job now is. Said plainly, and measured against the field rather than the team, it is the most useful thing I have, and the least comfortable. The larger part is enablement, which is a flat word for something that matters more than the title work ever did. It means clearing the obstacles they shouldn't have to spend their attention on, and going out to find the opportunities that will stretch them rather than waiting for them to ask for one. It is speaking for them in the rooms they are not in, making sure the work gets seen, and taking the hit myself when a risk I talked them into doesn't come off. None of it works unless you actually care what becomes of the person, and not only their output; that last part is the one a lot of managers never really bothered with. And when someone can build almost anything in an afternoon, I worry less about what they build than about whether it was worth building at all. I have got this wrong in both directions. I have tried to stay technically relevant enough to keep reviewing their work, which fooled no one and mostly wasted their time. And I have gone the other way, deciding someone was too good to need much from me, which turned out to be its own kind of neglect. Being better than me never meant they needed nothing from me. It usually meant they needed the things I was avoiding, because those cost more than knowing the answer. AI is about to make obvious what has been true the whole time: a lot of what gets called developing people was never that. It was correcting their work with things they hadn't learned yet, and calling the gap growth. Once the gap closes and everyone has a model, it gets very easy to see who was actually growing people and who was only ever the one who knew the most, mistaking that for leadership. That has never been comfortable to watch, and it is about to get a great deal easier to see. Managing someone better than me strips the job down to the part I can neither fake nor hand to a machine: seeing what they can't, saying what no one else will, and staying with it long enough to get them past where I got.