Don't Compress Mastery. Compress Serendipity.

Hey everyone,

What if the part worth compressing isn't mastery — but serendipity?

When my son was younger, I took him to a soccer camp where one of the coaches talked about luck in a way that has stuck with me ever since. Soccer is obviously full of luck. Balls take ridiculous bounces, shots get deflected, keepers knock balls into places nobody could have anticipated, and sometimes a goal happens because you happened to be standing in exactly the right place at exactly the right time. But the coach's point was that some players seem to get lucky a lot more than everyone else, and that isn't entirely an accident. You can actually learn to put yourself in positions where luck is more likely to happen to you.

Follow the shot instead of admiring it. Make the run even when you're pretty sure the ball isn't coming. Get to the back post. Stay engaged after the first challenge because the ball has to go somewhere. You can't know where the weird bounce is going to land, but if you're constantly moving into places where something useful could happen, over enough games you're going to look awfully lucky.

I've thought about that lesson a lot over the years, but lately I've been thinking about it in a completely different context because of AI and the way we're building StudioOps. So much of the pitch around generative AI is about compression. Do something that used to take a week in an afternoon. Generate fifty ideas instead of five. Write the song, make the image, create the video, build the campaign. There's a lot of "look at this shit I made in thirty seconds," as if the amount of time we managed to remove from making something is somehow a measure of how good the tool is.

I'm not sure I want that, at least not for the parts of creative work I care about most. I don't particularly want to compress the years it takes to develop taste, or the hundred bad songs you write while slowly figuring out why the hundred-and-first works. I don't even necessarily want to eliminate the frustrating afternoon where a verse isn't working and I can't figure out why. I've had enough of those afternoons to know that sometimes the thing I'm acquiring during those hours isn't the verse. It's a slightly better understanding of songwriting.

There are plenty of things worth making faster, obviously. I don't have some romantic attachment to wasting time. But I've started wondering whether we're aiming all of this computational leverage at the wrong bottleneck. Maybe the part worth compressing isn't mastery. Maybe it's serendipity.

I've been dogfooding StudioOps with my own creative life for long enough now that I keep seeing this happen. I'll throw something into it that doesn't seem particularly important: a sentence, a photograph, a memory, a chord progression, something I noticed on a walk, some half-developed thought that probably isn't a song and may never become one. Then, sometimes weeks or months later, one of those things ends up close to something else and I suddenly see a connection that wasn't visible before.

That's basically what happened with "The Dark," a song I've been working on about a friend I knew in college. I had all of the pieces because I was there. I remembered her running alone through dark woods every morning. I remembered us getting high sitting on some poor guy Morris's headstone. I remembered this strange, incredibly vulnerable little gesture where she closed her eyes and stroked her hair over and over while we talked. I remembered that after college she more or less disappeared from my life, and I knew how her story eventually ended.

I've carried some of those memories around for thirty years. What I hadn't done was put them close enough together to notice that darkness ran through almost every part of the story. The literal darkness she ran into every morning, the parts of herself she kept hidden, the opacity between us, the way she eventually disappeared, and my own inability at twenty-something to understand what I was seeing. Once those things were sitting next to each other, a story I'd known for most of my adult life became a story I could finally see.

AI didn't invent any of that. It didn't experience it, and it certainly didn't spend thirty years becoming the person who could finally understand it. I did. What the system helped with was positioning. It moved a few things that had been sitting twenty yards apart into the same patch of grass.

Which, eventually, brought me right back to that soccer camp.

You can't manufacture the lucky bounce, but you can get better at being where the ball might land. I think creative software can do the same thing. Put the old lyric near the new photograph. Let the chord progression bump into the forgotten memory. Resurface something I wrote three years ago when something I write Tuesday suddenly gives it a different meaning. Keep enough of my creative life in play that there are simply more opportunities for those weird collisions to happen.

That feels very different to me from asking AI to write the song. I don't want it to take the shot. I want to become better at taking the shot. I want the years of writing songs, playing instruments, listening closely, getting things wrong, developing taste and slowly learning what deserves to survive. That's the part of the process I have no interest in shortcutting, because I'm increasingly convinced that is the process.

But I'll happily take more chances to get lucky.

If these tools can help me make connections I might otherwise have made three years from now—or never made at all—that's an extraordinary kind of leverage. Not because I've compressed the work required to become good, but because I've increased the number of potentially meaningful things I get to encounter along the way.

Maybe that's the AI tool I actually want: not one that promises to make me a master faster, but one that keeps putting me somewhere useful when the ball takes a weird bounce.

Don't compress mastery. Compress serendipity.