complexity trap
I have a specific memory from 2015 that explains how engineers get trapped. I was watching a YouTube video where two neural network agents, trained via genetic algorithms, learned to snipe each other. That moment sparked an obsession. I wanted to build that.
For the next few years, I fell into a common delusion: I believed that for work to be valuable, it had to be complex.
I spent my time computing gradients by hand and dissecting frameworks, trying to understand the machinery of AI from scratch. I was learning for the sake of learning, telling myself that "someday" I would be ready to contribute. But in reality, I was just a spectator. I was watching the world progress while I fiddled with toy problems, paralyzed by the gap between my code and the state of the art.
I was using complexity as an excuse to avoid being useful.
Looking back at 2016, the opportunity cost is obvious. While I was trying to replicate DeepMind in a dorm room, the world was being rebuilt by "boring" technologies. React was just taking off; Swift was redefining mobile apps. These tools were available, they were stable, and they had massive leverage.
I ignored them. I thought they were gimmicks. They didn't feel "novel" or "worthy" compared to the grand problem of General Intelligence. I refused to do the simple thing because I was too busy failing at the hard thing.
Now, as we enter the era of agentic computing, I see the same uncertainty and the same trap.
The instinct for a curious engineer is to try and build the agent—to compete on architecture, to understand the neurons, to own the complexity.
This is a mistake.
The way to leverage the uncertain is not to build the tool, but to use it before anyone else figures out how. The winners of the last decade weren't the people who wrote the React; they were the people who used React to build billion-dollar products.
We are in a similar window now. The agents are here. They are expensive, they hallucinate, and they are imperfect. But they are the lever.
Don't chase the novelty of how they work. Chase the utility of what they can do. Ask: How can I delegate the painful, boring parts of my life to this system? How can I use this unfinished tool to do finished work?
When you try to predict the future or build the "perfect" system, you stop living in the present. You get paralyzed. But if you look at how tech is actually adopted, the advantage goes to the pragmatist, not the theorist.
It might be hard to work with these systems today, but nothing beats the investment of using a tool that improves itself over time. Leverage the system as it matures, so that you mature with it.
Learn to think in terms of delegation. Be critical of the part of your brain that craves complexity.
Focus on the goal, and let the rest pass you by.
If you enjoy the content, feel free to subscribe to my blog via email or RSS feed.