Last week I was browsing my personal website and checked my blog page. I remembered writing my introduction post, and how excited I was about the idea of starting a blog journey.
After my introduction post, I had a couple of ideas. I published three more posts about topics that were interesting to me. Then I hit a wall.
And for some reason I suddenly felt this urge to finally start the weekly blog habit. I set up a blog bot that helps me with topics to write about based on what I worked on and learned each week. I'll share more about that setup in an upcoming post.
This is the first post I'm writing in a while. When my blog bot suggested this week's topic, I was disappointed. I felt like I was boring and that I wouldn't be able to write anything meaningful. I felt like I wasn't doing anything interesting or valuable.
I was reading about people's agentic workflows. I felt like my setup was far behind theirs. This demotivated me even more.
But I was wrong.
What I actually did last week
I found a Cursor plugin called pstack. Lauren Tan built it. I started using it as-is. I haven't tailored it much yet.
I'm in the middle of a 55-minute recording of her talking about working with agents. Her setup looks nothing like mine. She runs tens or hundreds of cloud agents at once. I run one or two agents locally, and even those are usually on different projects in different apps. No real concurrent work. I tell myself that's because I work on simple apps.
Watching her, I find it easy to feel behind. Like I'm doing a toy version of the job.
I don't think that's true. I think I was measuring the wrong thing.
One agent still counts
Spinning up hundreds of agents is a specific kind of workflow. It assumes work that splits, waits, and comes back. Most of what I work on doesn't look like that. A form. A flight-search-result component that got more complicated. A contract that three services and a UI layer have to agree on.
You don't need a swarm of agents for that. One agent is more than enough. One agent that doesn't lie to you, and a brain that still reads the diff.
pstack is interesting because it's not "more agents". It's a collection of opinions about how you talk to them, and how the spawned subagents talk to each other, and how they evaluate each other's output. I've only just started, so I won't pretend I have a setup worth copying. But I will say this: I've already noticed how much better the results are when I use it, so right now I'm going all in on it.
Last week I also shipped a breaking API change across three services on the same day. Define the richer contract at the edge. Validate at every layer instead of logging a giant blob. Update the UI so it can render nested data without flattening it into a dumb list. Coordinate the merges so one service doesn't sit on the old shape. All of it is work I can stand behind.
The writing problem is the same as the agent problem
I kept waiting for a topic that felt original. That's the same trap as waiting until I run a hundred agents to call it "real agent work".
I don't need a product people already care about. I need a week I actually lived, written down before I forget it.
This is that. Next week I'll still be experimenting and maybe already tailoring pstack to my workflow. I'll still be running only one or two agents locally. I'll probably be working on something that isn't interesting to anyone else.
But I will write about it. Because with these weekly blog posts, I might write about a feeling or a state of mind that someone else is experiencing right now.
I might write about a simple setup that's not as powerful as others, but that's still valuable.
And that might help someone else.