Kyle Shevlin

Yesterday, I had an exchange with a colleague that made me feel something very poignant. They didn’t know this was happening to me at the time, but I wrote a short essay and shared it with them afterwards. I have changed their name for the purposes of this essay. Here it is in full:

I just had a moment that I want to share. See if it resonates with anyone. Sorry it’s a bit long.

I was just talking with Joe about an idea I’ve had for a while. Turns out he has a similar, bigger, better idea. We chat back and forth about it. I realize quickly, “Yeah, we’re not going to do my idea.” That’s fine. No big deal.

I started to write, “I’d love to help you with this”, and stopped as I realized “help” doesn’t mean what I want it to mean anymore.

By “help”, I mean, I want to write code. I want to physically contribute to the solution. I want to explore, discover, and learn the lessons about this idea that can only be learned by running into obstacles and getting around them or through them. But that’s not what “help” will mean in this case. Maybe ever again in our industry.

The way I can “help” is to read a doc, write some comments, have a discussion, and hope that maybe it helps Joe write a better spec & prompt. In reality, a bunch of agents are going to write the code. I will have no part in that.

I’ve never been one that gets a lot out of reading architectural documentation. It just doesn’t get my brain going. I can look through Joe’s docs, maybe pick a few things out and give feedback, but doing so won’t feel as helpful to me as actually making code contributions would.

I recognize that this mode of operation, write a doc and hand it off to be done by others, has been the work of managers for forever. But there is a reason (among many) that I’m not a manager.

I can learn a lot of ways, but I’ve always learned best by doing. Typing out the code helps me understand it. Helps me find those things that no amount of planning or prep would have revealed. Just based on recent experience, there’s a lot less “doing” going on, and I find myself understanding less. Learning less.

I think perhaps gone are the days where maybe I would have paired with Joe as he worked on part of this (or vice versa) and I’d get to learn directly from his expertise. But what’s the need for transferring knowledge to other humans if they aren’t going to ever be the one’s using or implementing it? Do I need to know this elegant solution Joe had when it matters more that the agent knows it instead?

I have more I could share, but I’ll stop there. In short, this brief moment I just had, made me very sad. I just didn’t want to feel sad alone.