The Credibility Commons

February 24, 2026

What replaces the firm when verification gets cheap

In 1991, Paul Glover started printing money in Ithaca, New York.

Not counterfeits. A local currency called Ithaca Hours. One Hour equaled one hour of labor — regardless of what the labor was. The carpenter and the babysitter traded at parity. About two thousand people used it. Millions of dollars equivalent changed hands. Glover himself paid most of his expenses in Hours.

But Glover wasn't just the founder. He was the protocol. He rode his bicycle around town tracking where Hours piled up. He matched businesses with complementary needs. He kept the currency circulating. He evangelized new participants one conversation at a time.

When he moved away, the system collapsed.

His own diagnosis: "Every local currency needs at least one full-time networker to promote, facilitate, and troubleshoot."

One person, on a bicycle, being the coordination layer for an entire alternative economy. That's what coordination looked like when it had to be human.

Coordination All the Way Down

The previous posts in this series traced a pattern: every enclosure is a solution to a coordination problem.

  1. Sacred enclosure — priests controlling texts — worked because literacy was scarce and coordinating access was expensive
  2. Copyright — worked because enforcing claims required coordinated legal infrastructure
  3. Infrastructure enclosure — works because building and running datacenters requires massive capital coordination

Each enclosure broke when coordination got cheaper. The printing press. The internet. And now...

AI is an intelligence revolution. What comes with this revolution is a coordination-cost collapse — the costs of searching, verifying, and monitoring that made firms necessary in the first place.

Why Firms Exist

In 1937, Ronald Coase asked a question that won him a Nobel Prize: if markets are so efficient at coordinating production, why do firms exist at all? Why don't we all just contract with each other directly?

His answer: transaction costs. Writing contracts is expensive. Specifying every task in advance is expensive. Monitoring compliance is expensive. It's cheaper to hire employees and tell them what to do than to negotiate every interaction on the open market.

But why are transaction costs high?

Because verifying credibility is expensive. Can this contractor do what they claim? Will they deliver on time? Are they worth what they're charging? Every transaction requires answering these questions, and the answers are hard to get. George Akerlof won a Nobel Prize for demonstrating how information asymmetry breaks markets — buyers can't tell good from bad, so they assume the worst and pay accordingly.

The existence of entire industries built around verification — background checks, credit bureaus, professional licensing, auditing firms — is evidence enough. If credibility were cheap to verify, these industries wouldn't exist.

The firm bundles credibility. In the 1970s, the line was "nobody ever got fired for buying IBM." In the 2000s, it was Cisco. The pattern holds: delegating to the institution is cheaper than assessing every individual contractor. The institution vouches for its members. You delegate your credibility assessment to the firm, and the firm does the verification internally.

AI attacks this directly. An agent that can verify claims, assess track records, monitor delivery, and compare pricing is eliminating the overhead that made credibility-bundling necessary.

Finding Dave

When I need someone to fix my furnace, I have options — and they degrade in order:

  1. Personal knowledge — I know Dave. He's fixed my furnace three times. I've seen his work. I trust him because I've watched him over years.

  2. Trusted referral — My neighbor used someone last winter. My brother-in-law knows a guy. The referral carries weight because I trust the person making it.

  3. Institutional reputation — I call Acme Furnace Co. I don't know anyone there, but they're a company with a reputation to protect. The institution vouches.

  4. Platform heuristics — I search online. 4.8 stars. Maybe the reviews are real. Maybe they're gamed.

This is a credibility degradation curve. Each step is a proxy with more noise, further from someone actually knowing whether Dave is good at fixing furnaces and won't cut corners or gouge me.

This is also why fly-by-night contractors change their company name every few years. The entity is disposable. The person behind it isn't — but without a way to track Dave across companies, you'd never know.

The AI-mediated version isn't step five. It's a return to step two, at scale.

My agent reviews Dave's work history across two hundred jobs — or across every job Dave's ever done. It verifies his certifications against the issuing bodies. It checks complaint records, assesses his response patterns, compares his pricing to market rates. My agent knows Dave the way my neighbor does — someone who's lived next door for a decade and talked to every one of his customers.

Village-level credibility assessment without the village. The agent does the knowing. The anonymity of the city dissolves — do bad work, and it follows you the way it would have in a village of 150.

The firm's value proposition was knowledge asymmetry: "You don't know if this contractor is good. We do." Acme Furnace Co monetized the gap between what they knew about their people and what you could learn on your own.

When your agent can verify a contractor as thoroughly as Acme Furnace Co's internal HR, the asymmetry collapses. The firm doesn't disappear — maybe it's still convenient, maybe it adds other value — but you're not dependent on it. It's an option, not the gatekeeper.

Before Currency: The Village Economy

In a village of 150 people, nobody needed currency. Everyone knew who contributed what. Reputation was the economy. You knew who pulled their weight and who didn't. The blacksmith knew the baker knew the carpenter knew the midwife. Marcel Mauss called this gift economy — exchange built on reciprocity and memory, not transaction.

Money emerged because reputation doesn't scale. David Graeber argues currency is fundamentally about trust between strangers — a way to extend credit beyond the village where everyone knows you. You cannot track 10,000 contributions in memory, nor can strangers trust claims of being "good for it." Currency abstracts contribution into a transferable token, letting strangers exchange value without knowing each other's histories.

Currency also enables specialization. In a village where everyone knows everyone, the blacksmith can only sell to 150 people — and must remain a generalist to meet their varied needs. But currency lets the market extend beyond reputation. Adam Smith saw this clearly: "The division of labor is limited by the extent of the market." The blacksmith can sell to a stranger passing through, accumulate currency, and buy from specialists he'll never meet. Currency doesn't just abstract trust — it scales the market, and the scaled market enables specialization.

Currency is a solution to scale-induced anonymity. So is privacy. In the village, everyone knew you were a farmer, a drunkard, a debtor. The city offered anonymity — strangers who didn't know your business. Privacy as we understand it is an artifact of urban scale, not a natural right. Both dissolve when credibility assessment becomes cheap again.

Where Credibility Lives

But here's the question that determines whether any of this is liberation or just a new enclosure:

Where does the credibility data live?

It's always lived somewhere. The progression mirrors the credibility curve:

  1. Personal knowledge — data in your own memory (village scale)
  2. Trusted referral — data in your network's memory (village scale) — this is where the agent operates
  3. Institutional reputation — data in the firm's records (city scale) — collapses into layer 2
  4. Platform heuristics — data in the platform's database (global scale) — collapses into layer 2
  5. A distributed commons — the open question (global scale)

Layers 3 and 4 collapse into layer 2. It knows Dave the way your neighbor does — but it can know everyone that way. Firms and platforms don't disappear; they just stop being the only path. You can still use Acme Furnace Co. You can still check Google Reviews. But you're not forced to because they're the only ones who know.

The question is: what feeds the agent?

If the agent queries centralized platforms, you're back to layer 4. Google becomes the gatekeeper again — same structure, different interface. If every agent collects independently, there's massive duplication, and Dave has no portable reputation. He rebuilds credibility from scratch with every new client's agent.

The interesting option is a commons — credibility governed by protocol, not owned by any platform. Maybe Dave publishes his own track record. My agent verifies against his published history, your agent does the same, and somewhere in this there's a system that doesn't route through Google.

The problem: who runs it? Who pays for it? Or maybe — does it need to be "run" at all?

The Recursion Problem

This is where we hit the same wall that killed Ithaca Hours.

Glover got a three-year grant — in dollars — to work full-time on Hours. The system meant to replace dollar dependency was itself dependent on dollars to fund its coordinator.

Every commons faces this recursion problem. Linux is free, but Linus Torvalds is paid by the Linux Foundation. The Foundation is funded by Microsoft, Intel, Google. The gift economy floats on top of the money economy. It doesn't replace it. It's subsidized by it.

Can a commons sustain its own coordination layer without falling back to the system it's trying to transcend?

I don't know yet. But I think I know what the answer has to look like: protocol governance, credibility as the primitive rather than bundled in institutions. Maybe infrastructure that doesn't require shared infrastructure at all — where the data lives where it originates, with the people who earned it.

Something like Elinor Ostrom's design principles for commons management — clear boundaries, collective choice, monitoring by participants — but without needing a foundation, a grant, or a person on a bicycle.

I'm still working through what that actually looks like.

This is part of an ongoing series exploring what happens when knowledge stops being property. Previously: Knowledge Was Always Free. Copyright Is Dead. Copyleft Is Too. The Third Enclosure.

James Henry is a senior engineer tracing what happens when credibility assessment gets cheap. He works with LLMs liberally — including in the writing of this post — because he believes the collaboration is the point.