Working with agents has changed the shape of my day¶
I used to think the better agents got at doing things, the lighter my own work would get.
What I've found is something different. I am not busier in the old sense, but the shape of my day has changed, and the new shape asks more of me, not less.
I work with five or six agents daily. They write code, dig up references, summarize meetings, run tests, draft documents. Each one responds fast and eagerly. Taken together, they produce a kind of constant forward motion. Tasks move. Cards close. Branches merge. It feels productive.
That feeling is where I have to be careful. Agents are so quick, and so willing, that they can produce an illusion of prosperity. A lot of output is not the same thing as the right output. If I am not paying attention, I can spend a whole morning feeling good about what "we" got done, without having asked whether any of it mattered most.
So my job has shifted. It is less about doing the work myself, and more about deciding what work should happen next, in what order, and why.
That sounds obvious written down. In practice it is harder than before. When I was slower at execution, prioritization happened naturally, because I could only do one thing at a time, and I had to pick carefully. Now I can start ten things in parallel. The cost of starting something is close to zero, which means the cost of starting the wrong thing is also close to zero, and the savings from picking right are proportionally larger.
I try to hold myself to a simple test. Before I give an agent a task, I ask whether I would still want this task done if doing it took me a full afternoon. If the answer is no, the task probably should not be queued at all, even if the agent would happily knock it out in twenty minutes. Agents lower execution cost so much that they invite me to be sloppier about what I ask for. The check is to not be sloppier.
I have also noticed a pattern in how different people adapt.
People who were already strong at execution tend to do well. Agents free up their hours, and those hours get reinvested in higher leverage work: judgment, customer conversations, setting the direction of a week or a quarter. One person starts to feel like a small team.
People whose sense of contribution came mostly from working hard and shipping tend to struggle. The shipping still happens, but the agent is doing it. The instinct is to stay close to execution anyway, to review more than is useful, or to specify tasks in more detail than the agent needs. I recognize this instinct because I have it too. It is worth noticing when you are doing it out of habit rather than because it actually improves the output.
The deeper point, I think, is this. Agents should save humans from repetitive work so that humans have room for the work that is genuinely creative. That is the real purpose. If agents only make me more efficient at what I was already doing, then I have underused them. The question I try to keep asking myself is what I would work on if the rote parts of my week no longer existed. My answer keeps getting more ambitious, which I take as a good sign.
Agents are good mirrors. They show you how much of your day was execution, and they ask what you want to do with the room they give back.
写于 2026-04-22