Post Mortem: A Coordination Challenge

Aug 2025

Last week, I invited you into the first major beta test of my collaborative puzzle design. Many of you took the archetype quiz, discovered your Investigator role, and dove into the Puzzle page.

Each archetype had its own assignment:

• Chronologists searched my Twitter and surfaced “There is”

• Interlocutors coaxed a custom-GPT until it gave up “a crack”

• Patternists ran an nltk n-gram analysis on my diary and found “in everything”

• Archivists snooped my Obsidian vault for initials that spelled “that’s how”

• Architects peered into the console log to see “the light”

• Cartographers traced coordinates across the globe to assemble “gets in”

Together, those fragments form the Leonard Cohen line:

“There is a crack in everything, that’s how the light gets in.”

Individually, several of you solved your type’s puzzles. Many of you solved all of them.

But what didn’t happen is the thing I designed for: coordination. Instead of bringing your pieces together, you mostly played solo. The sentence never surfaced collectively.

Why You Defected

Each of you faced a choice: (A) invest effort to share your fragment and coordinate with strangers, or (B) keep solving on your own.

In game theory terms, it was a coordination game with skewed incentives. Solving alone was lower-cost, higher-certainty. Sharing required extra effort — with the potential for higher reward if you read my intent correctly (and did a little perspective-taking). But the payoff from that collective reward depended on enough others making the choice around the same time. The challenge wasn’t specifically time-bound.

There were also structural reasons tilting the board toward solo play:

• Image costs. The “auto-tweet” share mechanism made coordination a costly signal. Posting an AI-slop tweet could feel reputationally risky, especially for those with curated feeds. I had hoped you’d just edit for your style, but I knew it was a long shot.

• Entry points. Many stumbled into this via Twitter or directly onto the puzzle page, not the Substack letter where I gave my explicit statement of my intentions.

• UX clarity. To put it mildly, I left too much implicit. It wasn’t obvious enough that sharing was expected and part of the puzzle.

The result wasn’t malicious, it was predictable: you optimized locally, not collectively. That’s the first lesson here: design determines behavior.

What's Next

The next campaign will play at the edges of this problem again. I want to see whether coordination emerges when the incentives are clearer. Think of it as the next round of the same game, but with slightly different stakes.

The choice won’t go away; you’ll still be able to make it through with individual completion. But there may be more to gain for those willing to risk the overhead of syncing with the others.

Consider this last round a calibration exercise: we learned how far individual effort stretches. Now, I’m curious how far collective effort might reach, and whether any of you notice the payoff structure shifting under your feet.

So Long,

Marianne 🩵

P.S. If you did make it all the way through to the end, I’d love for you to comment the name of a track that caught your ear.