The Magic Circle: Why AI Can’t Cross the Paper Jam

There’s a moment every computer user knows intimately: the printer jams. It’s frustrating, embarrassing, and oddly universal. But what if I told you that humble paper jam is actually guarding humanity’s border against the flood of AI automation? That’s the provocative thesis from writer Robin Sloan’s latest essay, and it deserves serious attention.

The Core Insight

The “magic circle” is a concept from game studies, formalized by Johan Huizinga. It describes how games create their own bounded spaces with special rules. You could reach across a chessboard and knock over the king—but you don’t, because you’ve agreed to play by the game’s rules.

Sloan argues that the digital world has its own magic circle: computation operates on symbols in, symbols out. That’s it. And here’s the crucial point: the digital world, for all its sophistication, is approximately a “drinking straw” compared to the richness of physical reality.

The flood fill of AI automation can fill the digital space—but it stops at the boundary where symbols meet the physical world. The printer jams. The paper wrinkles. The robot can’t thread a needle. This isn’t a limitation that will be solved soon. It’s a fundamental boundary.

Why This Matters

Consider the project Sloan describes: tracking letters sent via First Class mail. It required writing code, designing custom labels with exotic barcodes, printing them, mailing real physical letters, watching USPS scan data, receiving those letters, and adjusting based on feedback.

Could an AI agent help with the coding? Maybe. Could it autonomously handle printing labels, mailing letters, and processing responses? Not a chance. The project required crossing multiple magic circles—and breaking out into the physical world of paper, printers, and post offices. It involved folding, peeling, sticking. Gnarly.

The sewing machine offers a parallel. It can’t produce every kind of stitch humans can. Conversely, its trademark lock stitch—where two separate threads loop together—was never sewn by humans. It was developed specifically to suit the machine’s capabilities.

Automation doesn’t meet a task in the world and simply do it. There’s always negotiation—the invention of some new relationship. Trains need tracks. Cars need roads. And AI? It needs the digital.

Key Takeaways

• The drinking straw: Digital technology, for all its prominence, is narrow compared to physical reality
• Bounded computation: AI operates in a magic circle of symbols—powerful but constrained
• Paper jams as gatekeepers: The boundary between digital and physical is where automation hits limits
• Redesign, not replacement: Automation often redesigns tasks to fit machines rather than automating existing workflows

Looking Ahead

Sloan’s practical implications are worth considering. If your work is fully inside the magic circle of “symbols in, symbols out,” your world is changing fast. It might be time to get creative about how you “season” your work with the physical.

The construction of bridges between digital and physical might become something to approach carefully. As markets for human hands emerge to do physical tasks AI can’t, regulation will become crucial.

Most bracing: “The world can run without an internet. The internet can’t run without a world.”

In an era of endless AI options, it’s worth remembering that one thing is mandatory—the physical world we share with rhinoceroses, anchovies, and Joshua trees. Software cannot, in fact, eat this world. It can only reflect it, encroach upon it, or distract us from it.

The paper will keep jamming. And maybe that’s a feature, not a bug.

Based on analysis of “Flood fill vs. the magic circle” by Robin Sloan