The Inefficiency IS the Creative Process

A couple of weeks ago, my daughter and her writing partner held a workshop for a musical they've been developing for five years. It'll probably take another three or four before it's commercially viable. 

By then, they’ll have spent close to a decade on a single show.

In the last five years, they’ve had three staged performances, multiple readings, rewrites, and workshops like this one—each one a chance to run the material through real performers and in front of a real audience and find out what they actually have.

Five years ago, that timeline would have sounded ambitious. Today, it sounds almost irrational. Today, you can generate a full draft of just about anything in under a minute.

We live in a world of unlimited content now. The bottleneck used to be production. It's not anymore. Anyone with a browser and a free account can generate a screenplay, a pitch deck, a blog post, a business plan, a song. The output is fluent. Sometimes it's good. And it shows up instantly, fully formed, with no scars.

So what's my daughter doing spending a decade on a single show?

The thing they're building can't be compressed.

Every workshop, every reading, every performance is a conversation between the creators, the performers, and the audience. A performer interprets a line differently than the writers intended, and sometimes what they do with it is better than what was on the page. A song that reads well in a script falls flat in the room. A scene they almost cut turns out to be carrying more emotional weight than anything around it. None of that is exists until real people are in a room doing it in real time.

Then the writers take all of that back and rewrite. Not from imagination, from experience. From having watched what happened when it made contact with reality.

That’s the creative process. Each loop takes months, sometimes longer, and you can't run it faster because the creators who come back after a workshop are different people than the ones who started the first draft. They've been changed by what they saw or experienced. They've seen what connected, what confused people, what surprised them, and what surprised themselves.

The next version of the work is deeper because they're deeper. The work and the creators develop on the same timeline. I wrote in Statistically Unlikely about how GenAI is designed to produce what's statistically most likely—fluent, competent, and predictable. That's about the output. What I'm getting at here is the process. The years of development aren't an inefficiency to be optimized. They're where the meaning comes from.

What Audiences Actually Want

Nobody in a theater audience is thinking about the development process. They're not sitting there going "ah yes, I can tell this was workshopped extensively." Well, maybe my family is, but that's a different story. But they're responding to its result: a lyric that lands at exactly the right moment, a scene that builds tension without rushing. Nothing feels accidental. That kind of specificity doesn't come from a first draft. It comes from watching real people react to your material and rewriting based on what you saw.

Frictionless content doesn't have that. It's competent and smooth. And that's about it.

We're producing more content than at any point in human history, and a growing share of it has never been tested against genuine human reaction. Not reviewed. Reviewed is a checkbox. Tested. Put in front of someone to see how it lands, what it does to them, whether the thing you intended is the thing they experienced. A test asks what actually happened when another human experienced it.

People can feel the difference, even when they can't explain it. Live theater still fills seats despite infinite streaming options. Concerts still sell out even though the studio recording is technically cleaner. The experience of being in a room where something is unfolding—where the performers are responding to the audience and the audience is responding back—is a conversation, not a content delivery mechanism.

Conversations take the time they take.

The Irony Isn't Lost on Me 

I'm writing this newsletter with the help of AI. Claude has become a thinking partner for almost every issue. It helps organize ideas, draft, challenge assumptions, explore different structures, and sharpen arguments.

It's part of my process now. The irony of using AI to write about the limits of AI isn't lost on me.

I know what AI is doing for me right now. 

It's helping me think more clearly over the course of an afternoon. The scope is contained. The feedback loop is between me and the tool. This is what I'd call simple-to-complicated work, using the spectrum I laid out in Forget STEM, Get a BFA.

My daughter's musical is something else entirely.

It's years of performers, directors, audiences, collaborators, failed ideas, unexpected discoveries, and creators who keep evolving alongside the work itself.

Dozens of people, multiple years, an output that keeps changing shape because the people keep evolving alongside the work itself.

AI can help with parts of that—arranging, notating, drafting a lyric to react to. What it can't replace is what happens when fifty people sit in a room and collectively discover something nobody expected.

No tool replaces that. In a world flooded with content that skipped that process, the stuff that didn't skip it is becoming easier to spot and is more valuable.

Berkson's Bits

There's a tendency to do a lot of investigation when something doesn't work and very little investigation when something does. But then we share advice about what worked and less about what didn't work. The irony is we know less about why it worked yet that's the advice that we share.

What I'm Watching...

The Broadway show Hamilton started out as a hip-hop mixtape project. As Lin-Manuel Miranda says: “Even though I had visions of a mixtape, it was becoming a show as I was writing it.” This is a great behind-the-scenes look at some of that evolution: Hamilton: Spark Into A Flame. 

We keep talking about AI in terms of efficiency—faster, cheaper, more. And for a lot of content, that framing makes sense. Speed and volume matter when the goal is coverage or consistency.

However, my daughter's musical doesn't need to be more efficient. It needs to be workshopped, performed, questioned, broken, and rebuilt. The performers need to live inside the material long enough to discover things they couldn't have known on day one, and the writers need to be changed by all of it before they go back and rewrite.

The inefficiency is the creative process.

Every time I watch her come back from a reading or a workshop, she knows something she didn't know before she walked in. Not just because someone handed her better ideas,  but because she watched what happened when her words and music went through other people. You can't prompt your way to that. It accumulates through exposure, over time, surprise, disagreement and iteration in rooms full of unpredictable humans.

As AI continues to remove friction from producing content, the real differentiator won't be who can generate the fastest. 

It will be who is willing to spend the time discovering what the first draft couldn't possibly know.

When we optimize all of that away, what exactly are we optimizing out?

Looking forward to continuing the conversation…

Alan