Protect the Conditions That Let Ideas Appear
My best ideas rarely begin at the computer. Understory started during an afternoon at the hair salon. I came across the Japanese word komorebi, the light that filters through leaves, and wondered whether I could make something that helped people feel such words instead of merely reading their definitions.
Smoke Story began with an ordinary conversation. A Yale classmate who works in remote sensing reminded me how much I had enjoyed that data at WRI. When I asked myself which image from recent years had stayed with me, the California wildfires came back immediately. The product direction followed.
When I stay in problem-solving mode for too long, those flashes become rare. Ideas come more easily when I leave some time unclaimed, pay attention to things outside AI, and get enough rest.
Why I Make Something Becomes Part of the Work
None of the projects began as part of a plan, yet they gradually fell into three families: climate intelligence, workplace and personal tools, and web art and learning.
True Zero and Smoke Story grew from years of climate work. NoThanks came from observing workplace communication. Echo and Understory gave me new ways to preserve a feeling.
After reading Project Hail Mary, I wanted to hold on to its cold, spacious quiet, a solitude that did not feel lonely. In the past I might have written a paragraph. This time I made Echo, a night sky that breathes with music.
The most personal part of these projects is not the feature list. It is the curiosity, experience, and judgment that made me begin.
Nobody Had a Clear Answer, and That Helped
I once went to a Yale alumni gathering hoping to find an answer. Instead, I discovered that people who seemed steady were also working things out. The uncertainty was not a failure of expertise. The question simply did not have a settled answer yet.
At the time, I felt pulled in several directions. Hardware or software? Start making money now or keep experimenting? How should the business work? The more outside advice I consumed, the harder it became to hear my own thinking.
When I reduced the input, my ideas began to return.
Self-Knowledge Still Needs Action
Many people can now build quickly with AI. What makes someone seek you out is often your taste, judgment, and way of seeing the problem.
Knowing your strengths does not turn them into work on its own. I did not discover AI coaching through a market analysis. A friend asked for help, so I taught one session. More people came, and I realized the work felt natural because it drew on abilities I already had.
Most of my projects began without a clear market signal. I followed the impulse, made the thing, and watched what happened next.
Two months in, there was still something I did not know every day. The questions were becoming more specific, though. I had moved from wondering what AI could do at all to wrestling with one concrete problem at a time.