I’m An AI Manager

These days, I’m working by myself… in the traditional sense. I don’t have human teammates—but I am collaborating all the time with AI coding assistants.

I have no illusions. I don’t believe that AI “thinks”. These models are not people. Yet, to get the best results out of them, I need to use people manager skills.

The two AI coding assistants I use are Claude Code (with Opus) and Cursor (usually with GPT-5.2). They don’t have different personalities, but they definitely have attributes that lead me to turn to each of them for different tasks.

The simple breakdown: I have Claude Code write code and have Cursor/GPT-5.2 review it. I bounce back between them all the time. I’ve found it interesting and instructive to read what these two AIs say about each other’s work, and they way they respond to each other with patches and diffs. I pepper edits and comments of my own. After a few weeks of working like this, it feels to me that Claude Code is better at implementations and details and Cursor/GPT-5.2 is better at higher-level and more goal-oriented evaluations. When I develop plans, which is an essential part of getting code written the way I want it, I use both AIs. Again, I run a three-way cycle between us, with me sitting in the middle directing the work.

The people manager aspect is about taking the technical goals I’ve chosen, with their priorities and organized tasks, and then communicating this information to make the expectations clear to my AI my “co-workers”. It’s incumbent on me to understand the relative strengths of my team members, and make the work assignments in a way that sets up everyone for success. All the while, I need to recognize and provide the background and context they might need without them having to ask for it. That’s especially true in this case, since I can’t hope the will have any meta-view on what they’re doing or why. Obviously not. They aren’t people. I also need to give feedback (and sometimes pointed negative feedback) when work isn’t done up to my standards.

I have yet to find any tangible benefit for giving praise. This makes me sad. One of the happiest things I can think of doing in a work situation is telling another person, “Hey, that work you did is great!” Yet with an AI “teammate”, this will do little more than waste tokens.

Even so, I’m finding that if I do a better job supporting the AIs I collaborate with, they do a better job at accomplishing the tasks I give them, and I am more likely to achieve the goals I have for my projects.

I’m a people an AI manager.