When I asked my agent what I should write about, it told me to write an essay about the 3,598 commits I made to my software assembly line. The first thought that came into my head: “Can you write it?” Back in the summer of 2026, I told myself that all of this content must be written by me, not generated by AI.
It’s hard to slow down.
Let’s pause for a second and stop talking about theory and practice.
Thought flywheel
Some people are pushing the limits of what we can build with the latest frontier models. Terms like solocorn (a one-person unicorn startup) are gaining traction. While we are paused, let’s take some time to think about what is going on with these individuals.
- What used to require dozens of people now may be achievable by one
- How we build moves higher level, but the complexity stays
- We spend less and less time on small and focused problems
My latest endeavor is a software assembly line. It’s a giant SaaS platform and what it does isn’t important. It’s complex and it is something I think about every day. I wake up and check in on it, plug in to the AI, and start to orchestrate. I have to read and understand the AI’s dense prose and analyze the problem in front of me. I have to make decisions, one after another. It’s a never-ending multiple choice exam, or maybe it just feels that way.
I’ve never spent so much time thinking about stuff.
I now spend more time plugged directly into Claude Code terminals than I do sleeping. I wake up and connect for hobby projects. I use it at work. I am now trying to configure my OpenAI Dot to help manage my life. I even listen to podcasts about AI when I walk.
When I disconnect from the terminal, it doesn’t stop.
I can’t get it to stop on demand. I sit with my family and spend half the mental energy trying to stop the AI train in my head (which is consuming the other half). I have to slow it down.
The mental distraction isn’t that bad. I space out sometimes but I think I’ve always spaced out. The problem isn’t what you might think. The problem is that I have to step down off a high velocity train of thinking onto a grounded platform where everything just moves a little slower and I have whiplash. I have to adjust to normal mode.
I have to deal with the momentum from the thought flywheel that AI has hyper-engaged.
Cognitive marathon
Various people have shared their experience with cognitive load when using AI. I think I remember hearing about someone explaining that the future developer’s workday may only be two or three hours long because of how much has changed. That didn’t happen for me. Or maybe it didn’t happen for most of us. I just kept the pace and hustled.
It feels like a cognitive marathon. I’ve been hustling for months, trying to push the velocity needle higher and higher.
When we’re in this marathon mode, I think we immediately struggle to accept the reality of the new value of our efforts. If we can now achieve more in less time, we must reset our understanding of the cost of what once was.
Software may now be dirt cheap
It used to take dozens of developers and years of work to create a SaaS platform. Now a small group can do it in months. Proportionately, it may seem like spending a week figuring out how to do marketing is now a huge investment compared to the past. In the past, a week was nothing — a drop in the bucket. Now it’s a double-digit percentage of the total work for the project. So it feels heavier, but that’s only because software might be dirt cheap.
Basic human operations aren’t any cheaper
Talking to people, conveying understanding, sharing ideas, and getting on the same page is as expensive as it ever was. It only feels expensive next to how cheap software has become. This is where my current faults and cognitive dissonance start. I feel like I can blitz the project and that building in isolation is the optimal path. What I fail to take into consideration is that the feeling of that velocity is anchored to the relative efforts of the past — that is not a fair comparison.
We need to have higher-level conversations. That’s all. But to have those conversations, we need to have progressed similarly along the journey of exploring AI use.
The friction is a growing pain of everyone being at a different point in their understanding.
Stepping off
I want to make an effort to step off the high-speed train. Let’s paint a picture of what that looks like:
- Disengage the flywheel periodically: solve the problem without requiring mental momentum
- Re-embrace manual efforts with AI supplementing and reviewing: reverse the roles of who the reviewer is
- Prioritize people over agent terminals: collaboration using AI becomes the standard, which limits the pace and keeps us connected
- This essay
Writing has always helped me organize my thoughts and slow down in life. At one point that was a handwritten journal. Today, it’s an essay on a digital garden blog. Let this simple essay be the first few strides we take after that long train ride — stretch our legs and move at a normal pace.