If I was born in 1464, I'd die on a ship searching for Atlantis.

Instead I type "create next.js site with shadcn" three times a day. Sending scouts into the fog of war. Every experiment pushes back the darkness. The more you see, the more you realize is there.

Every new tool, every weird output, opens up territories I couldn't see before. The map keeps getting bigger. The tools are ahead of what I can even imagine.

I always had the urge to explore — voice, crypto, VR. Each world taught me new ideas and expanded my imagination. Crypto democratizing finance made me think about democratizing software creation before vibecoding did.

But I learn most from clashing ideas with reality. Feel it on screen. Going from idea to manifestation usually took months, sometimes years. My imagination quickly caught up to my execution limits and the industry's general constraints.

That just changed. For the first time, the tools are ahead of what I can even think up. And the gap between idea and execution collapses at the same time.

Now building software feels like a jam session. I start something, see what happens, then build on whatever comes back. Maybe it's the speed, but this whole coding/design thing for the first time in 10 years feels like painting to me. There's improvisation. Serendipity. The thrill of not knowing exactly where you're headed.

Keith Jarrett's "Köln Concert" — one of jazz's greatest albums — happened because the piano was broken. Stuck keys, faulty pedals. Jarrett almost canceled. Instead, he worked with the constraints and created something impossible to plan. That's what building with AI feels like right now.

The weird stuff a model does, how the tools work, all the random outputs — any of it might become our next app. Inform the next interface. Every little detail can break open a whole new world no one has found before.

Recently, I turned the Eames "Powers of Ten" video into a scrollable website. Los often has movies playing on mute while working — this time it was Harmony Korine's "Baby Invasion" I glanced over, saw a chat within the movie commenting on the movie. Stole the idea and made an AI-generated absurdist Twitch chat on top of the website.

Now AI Twitch chat is part of our toolkit — we're building multiple projects around it. AI twitch chat criticizing my work seems more generative than a detailed one paragraph answer from ChatGPT. These moments happen constantly now. They make me look closer at the world. I hear musicians and painters talk about this all the time, but for the first time it feels true to my work.

"There's always a new note to be found." - Thelonious Monk

When everyone can build anything, thinking deeply about what to build becomes the advantage. But my thinking is limited by my imagination. Columbus wanted to find India and stumbled on America. You only discover what’s really there by moving.

Finding the balance here seems key. Thinking is what gives you a compass. Building is what draws the map. One without the other is guessing?

I’ll try a tool like ComfyUI for five minutes, think it’s interesting, and move on. Then a real project comes along that needs it, and I get to dive deep. Vision turns into reality fast, and that reality suggests the next vision. Every prototype discovers whats possible.

Sometimes the lesson is that an idea isn’t worth it. I thought building apps while walking would be sick — until I tried it. Now I know.

I made a curator that learned my taste in media. It figured me out way quicker than expected. So we built Curate Curator. That experiment turned into dozens of conversations and thoughts about new AI interfaces.

Think to set the course. But importantly ship to discover.

The conditions for this kind of discovery have never been better. Experimenting has never been cheaper — tokens are VC-subsidized, prototyping is almost instant, everything runs serverless. We can explore faster, safer, and cheaper than ever before.

None of the above is advice. It's what makes this moment special to me and what I deeply enjoy doing.