The Other Exponential
On ditches, exponentials, and the curve nobody is watching
Originally published on Medium, March 27, 2026.
Every spring, the water comes.
It starts as snowmelt in the mountains, finding the lowest ground, gravity doing what gravity does. It enters the watershed through arroyos, the natural paths carved by centuries of runoff. Where the arroyos are still dirt, the water slows, spreads, and soaks into the earth, recharging the aquifer underneath. The whole system breathes: mountain to arroyo to aquifer to river. No pumps. No engineering. Just water following physics through the shape of the land.
Somewhere along that system, people entered the picture. The Pueblo people of New Mexico were irrigating this way for centuries before the Spanish arrived. When the colonizers came in 1598, they brought a word, acequia, and a legal tradition borrowed from the Moors, but the method was already in the dirt. Dig a ditch. Connect it to the water. Maintain it by hand. Share it.
A handful of families share an acequia. Every spring, before the water comes, they clean it together. A man called the mayordomo walks the ditch alongside the community — not above them, alongside — and everyone can see who showed up. The silt from last winter is cleared. The walls are rebuilt where they’ve crumbled. The ditch is made ready for the water.
This has happened every spring for longer than anyone can document. The acequia has outlasted the Spanish Empire, the Mexican government, American annexation, two world wars, the invention of the telephone, the automobile, the Internet, and the iPhone.
Nobody thinks about it. A ditch. In the desert. Maintained by hand. Part of a watershed that was working long before anyone showed up to claim it.
I’ve lived near acequias my entire life. For most of that time, I didn’t think about them either.
Please take a moment to listen to Jimmy Santiago Baca’s poem. It’s a few minutes long:
For my entire life, I’ve been looking at systems and architecture. Digging small trenches in my backyard with my own hands. Lego. Lincoln Logs. Tinker Toys. How things connect, how they hold, how they carry weight from one point to another. That instinct followed me everywhere, from my personal life to my professional one. It landed me in the tech world, where I’ve spent the last 20 years sitting inside systems after they break, figuring out why they broke. I’ve watched people lie to machines and machines lie to people. I’ve watched institutions promise things they couldn’t deliver and deliver things they never promised.
In my anti-fraud work, you learn one thing faster than anything else: costless signals are worthless signals. When it costs nothing to say “I’m trustworthy,” the statement carries no information. The con artist says it. The honest person says it. The words are identical. The difference is what it costs them.
A biologist named Amotz Zahavi figured out why in 1975. A gazelle that spots a predator will sometimes stop and jump straight up in the air, burning energy before the chase even starts. Zahavi called it the handicap principle: the signal is honest because it’s expensive. The gazelle is saying I see you, and I’m fit enough to waste this, so pick someone else. The predator reads the signal and moves on.
Twenty-some years I watched the human version. The fraud that works is always the fraud that simulates a costly signal at low cost. The system that catches fraud is always the system that measures whether the cost was actually paid.
And then one day I was driving down a street called Solano Drive — October light, the Organs jagged against the sky — and the physics clicked. The ditch. The Metallica concert. The fraud work. The gazelle. It was all the same thing.
Trust is what happens when costly signals are reciprocated over time in a shared environment. A force. And it obeys laws.
Let me tell you what the acequia actually is.
Families who know each other by name. The rules match the local watershed — the acequia in Taos doesn’t run the same way as the one in Albuquerquer’s North Valley, or the one in Las Cruces, because the snowmelt is different, the soil is different, the crops are different. The parciantes elect the mayordomo and vote on the rules. The mayordomo walks the ditch on foot, and both parties can see each other. If someone doesn’t show up for la limpia, the response is proportional: first a conversation, then reduced water priority, then a community discussion, then, only at the extreme, expulsion. If there’s a dispute about water allocation, the mayordomo mediates. Low-cost, local, accessible. The community governs itself. Multiple acequias coordinate through regional associations.
I didn’t learn any of this in a textbook. I grew up watching it. The physics was in the dirt the whole time.
A woman named Elinor Ostrom won the Nobel Prize in Economics in 2009, the first woman to do so, for studying exactly this kind of system. She found eight design principles present in every commons that survives. Boundaries. Local rules. Participation. Mutual monitoring. Graduated response. Conflict resolution. Self-governance. Federation. Every one of those principles was already in the acequia. Ostrom named what the ditch already knew.
I mapped her eight principles onto the trust equation and tested whether each one is a maintenance function — whether it keeps a specific variable above the threshold where trust can propagate between strangers. They are. All eight. Remove any one, and the trust network degrades in a predictable way.
I know, because I tested it against the system that removed all eight.
The Internet, from 1969 to about 1993, was an acequia.
A few thousand researchers. They knew each other. The RFC process (Request for Comments) was how they governed. You proposed a standard and anyone could respond. Rough consensus and running code. No one owned it. No one controlled it. The rules matched the community. The monitoring was mutual. Disputes were resolved on mailing lists. The system governed itself.
All eight of Ostrom’s principles held. Not because anyone designed them in. Because the network was small enough, bounded enough, and costly enough to participate in that trust propagated.
Then commercialization. Between 1993 and 2007, every one of those principles was stripped out. First the boundaries opened: millions of strangers flooded a system built for thousands, and nobody knew anyone anymore. Then monitoring flipped: the platforms started watching you, but you couldn’t watch them back. Then governance disappeared: when a million Facebook users protested a change they never agreed to, the company told them to calm down and changed nothing.
Eight principles. Fourteen years. All gone.
The trust network crossed a threshold around 2000. Not a gradual decline, a phase transition. It held, held, held, then fractured. Trust became local. Strangers could no longer cooperate through intermediaries. What replaced it were platforms — systems that substitute surveillance and enforcement for the trust they destroyed.
The concrete channel where the ditch used to be.
Meanwhile, everyone is watching a different curve.
Matt Shumer, an AI startup founder, recently told his family the honest version of what he sees. Models that do his job better than he can. Tasks that used to take days, done in hours, done without him. The water rising around his chest. Millions of people read it. He’s not wrong. The AI capability curve is real and it’s going exponential.
But while everyone watches that curve, nobody is watching this one.
In 1972, the General Social Survey asked Americans: “Would you say most people can be trusted?” Nearly half said yes. They’ve asked every year since. The number has gone in one direction. Down. Every decade, a new lower floor. No reversal. Not once. Not after any president, any policy, any movement, any technology. By 2022, less than a third said yes.
Fifty-year staircase. Only goes down. No reversal. Not once. In the same period, global information production went from one exabyte to 181 zettabytes — a 181,000x increase.
One line is exponential. The other is flat and declining. They’re moving in opposite directions.
After every major institutional crisis, trust drops. The question is whether it comes back.
After Enron, it took four years to recover to 86% of baseline. After the 2008 financial crisis, it took eight years to reach 72%. After the 2016 media collapse, there has been no recovery. Ten years later, the number is still declining. After COVID broke trust in the CDC, there has been no recovery. Six years later, still declining.
Each crisis takes longer to heal from than the last. Each one heals less. And after 2016, healing stopped happening at all. The system doesn’t bounce back anymore. It just finds a new floor. Meanwhile, crises are arriving every one to two years, faster than recovery ever completed even when recovery was still possible.
This is a feedback loop. It closed somewhere between 2008 and 2016. Since then, trust has only gone one direction. Down. And nothing is pulling it back.
Now put the two curves together.
The AI that Shumer describes does something specific to the trust curve. It generates content at zero cost, simulates judgment, writes and tests without you. And it floods the environment with costless signals.
Every chatbot that apologizes without meaning it. Every deepfake you can’t detect. Every AI-generated article indistinguishable from a real one. Every “I take your privacy seriously” from a system architecturally incapable of caring. Each one is a T-zero interaction — a trust encounter with zero trust content. And each one depletes your capacity to trust the next thing. Including the things that are actually trustworthy.
Every spam email. Every robocall. Every consent popup you click without reading. Every political ad you’ve learned to ignore. All of it drains the same aquifer.
AI is a T-zero machine. Not because it’s malicious. Because it produces signals that cost nothing. And the equation says trust cannot form from costless signals. I’ve tested it across systems from the acequia to Bitcoin to the placebo effect. The gazelle must actually waste the energy. The parciante must actually show up for la limpia. The signal must cost something.
The AI exponential doesn’t just change what work looks like. It accelerates the decline of the curve nobody is watching. The exponential is eating the flat line.
I found something Ostrom missed.
Her eight principles describe how to prevent damage to a trust network. Not one describes how to repair it after damage occurs.
Percolation theory, the mathematics of networks, predicts a repair function must exist. If a network can only lose connections and never recover them, it fragments. The only question is when.
The Internet proved it. When the eight preventive principles failed, there was no repair mechanism. “Block and report” deletes the connection. It doesn’t repair it. Every unresolved conflict permanently removed a relationship from the network. The network could only shrink.
But the acequia has this function. It’s been there the whole time.
La limpia isn’t just maintenance. It’s repair. You show up alongside people you may be in conflict with. The ditch needs cleaning and the water doesn’t care about your grievance. You work next to each other because the spring is coming whether you’ve forgiven each other or not. And in the working — in the physical labor, side by side, witnessed by the community — something metabolizes. Not because anyone mandated reconciliation. Because the structure required presence, and presence does what presence does.
That’s the ninth principle. Forgiveness — not as sentiment, but as infrastructure. Not by erasing damaged connections. Not by holding them frozen. By encounter. Acknowledgment. Demonstrated change.
Centuries of this, and nobody named it. Because it was never missing from the ditch. You only see the ninth principle when you study a system that doesn’t have it.
Like the Internet. Like every platform. Like the civilization we’re building at machine speed without the ditch to hold it.
So here is where we are.
Trust builds at the speed of human relationship. That speed has not changed. AI runs at machine speed and accelerates every year. The gap between them is growing without bound. That is not a metaphor. It is arithmetic. One curve accelerates. The other doesn’t. The gap grows forever.
No one is maintaining the ditch. The $1.8 trillion AI market has all the vocabulary. “Copilot” and “assistant” and “agent” trigger purchase orders. “Trust infrastructure” and “forgiveness” and “repair” trigger nothing. The vocabulary determines what institutions can see. And right now, institutions can see the flood coming. They cannot see the ditch that’s been here the whole time.
A ditch in New Mexico maintained by hand has outlasted every empire, every AI empire, every technology, and every platform that has ever existed. Not because it’s better technology. Because it maintains what every technology depends on and none of them provide. Trust. And encoded in its oldest ritual, in the act of showing up alongside people you haven’t forgiven to do the work the season demands, is the one function no platform has ever attempted: repair.
The ditch exists. It has always existed. The physics of trust is in the dirt, in the spring cleaning, in the mayordomo walking alongside, in the graduated response, in the forgiveness that the structure makes possible without anyone having to name it.
We don’t need to build the ditch. We don’t need to channelize the arroyo. We don’t need to straighten the river.
The water is rising exponentially.
And every spring, the water comes.