Extol Essays

What We Build On

Ken Leiserson — Sun, 13 Sep 2026 00:00:00 GMT

When someone you trust vouches for a person they've worked beside, something valuable has happened: testimony has traveled. The decisions that matter most, from who gets hired to who gets trusted, still turn on testimony, and almost none of it leaves a record the next stranger can check. The Attestation Notary protocol is a set of rules for writing that testimony down. Extol authored it, maintains its reference implementation, and operates the first production network built on it.

This essay is the protocol's family tree. Nothing in it is a clean-sheet invention, and every part descends from work someone else published, tested, and in several cases watched fail. We have written elsewhere about the graveyard of reputation systems and the laws it taught us.[^1] What follows is the other half of the inheritance: the traditions that got something right, what each one solved, and the one layer none of them built.

The notary: Solana Attestation Service

Start at the bottom of the stack, with the least human question: where does a signed record physically live? A signed claim needs to be durably recorded somewhere a verifier can check it, without trusting the recorder. This problem is solved. The protocol binds its notarization layer to the Solana Attestation Service (SAS), and it publishes the criteria that made SAS the right choice.

The chain is the protocol's notary, not its religion. One criterion outranks the rest: the substrate must refuse to be a directory. Nothing in how records are stored may let a stranger assemble everything Alice ever signed, or everyone who ever signed something about Bob. The SAS binding enforces that refusal, and it is the bar any future replacement must clear. A companion implementation, backed by nothing more than a Postgres database, shows the signing layer working with no chain at all, and documents what that choice gives up.

  ┌───────────────────────────────┐
  │                               │
  │            Extol              │
  │                               │
  ╞═══════════════════════════════╡
  │  ATTESTATION NOTARY PROTOCOL  │
  │                               │
  │  disclosure:                  │ ← Elinor Ostrom
  │    shown, never pulled        │
  │                               │
  │  the constraint:              │ ← soulbound tokens
  │    standing is not for sale   │
  │                               │
  │  the envelope:                │ ← W3C VCs, Open Badges
  │    the signed claim           │ 
  │                               │
  │  the notary:                  │ ← Solana Attestation
  │    the public record          │   Service
  └───────────────────────────────┘

The stack, bottom to top: each layer, and the tradition that supplied it.

The envelope: W3C Verifiable Credentials and Open Badges

One layer up sits the data format. The fields a portable, cryptographically signed claim about a person must carry. This problem was solved by a decade of serious work. The W3C Verifiable Credentials standard established the full model: signed claims about a subject, expressed in a portable envelope, verifiable by any conforming implementation.[^2] Open Badges built the peer-attested credential pattern at real scale in education, with over a thousand issuing organizations and millions of badges minted.[^3]

The protocol inherits the signed-claim shape from both. It departs in one structural place: Verifiable Credentials assume issuer authority hierarchies. A school issues a diploma, a government issues a license. The Attestation Notary protocol treats every signer as equal and lets weight derive from graph structure at read time, not from institutional privilege at write time. What a signature is worth depends on who signed, what they staked, and what the graph shows about them, decided when the record is read, not when it was written.

The deeper inheritance from these two projects is a warning: millions of badges were minted, and employers did not check them. The full post-mortem is the first law of our field guide. What matters for the lineage is what that failure revealed: the envelope existed, and what did not exist was a protocol governing who may fill it, at what cost to themselves, and under whose control it is shown.

The constraint: soulbound tokens

In 2022, Ohlhaver, Weyl, and Buterin published "Decentralized Society: Finding Web3's Soul," and ruled something out.[^4] Their observation was structural: any reputation primitive tradable in a fungible market will, over time, come to represent what someone was willing to pay for the appearance of standing rather than what the holder actually did. Their proposal, soulbound tokens, identity-bound and permanently non-transferable, moved the fix from instinct to design principle.

The protocol honors the constraint by not over-promising it. Any implementer can wrap a record in a token and trade the token, and no rule about bytes can stop them. A protocol that promised non-transferable standing would be making a promise it cannot keep. So the constraint lives where it can be kept, in the commitments adopters publish.

Extol's guardrails answer it directly: standing is non-transferable, any conversion of standing into transferable value happens in the open, and a firewall stops records from ever being wrapped into tokens. Other implementers may commit differently, and the protocol does not choose for them.

The protocol is deliberately silent on a further set of questions implementers will face immediately: voting weight, delegation, how standing converts to influence. That silence is not an oversight; it is a boundary. The protocol's job is to make signatures verifiable, not to prescribe how downstream systems read the graph those signatures build. Posner and Weyl's Radical Markets provides a hint of where to find such prescriptions.[^5]

The reasoning tradition: Ostrom

Elinor Ostrom never wrote a line about credentials, and her work governs more of the protocol's design than any document that did. Governing the Commons demonstrated that communities can steward shared resources durably, without privatization and without a central authority, when the rules meet certain design conditions: defined boundaries, locally fitted rules, community monitoring, graduated sanctions.[^6] Trust is such a resource; we make that argument in full elsewhere. A community's willingness to vouch is depletable: it can be over-harvested by costless endorsement, polluted by paid attestation, and enclosed by any platform that custodies it.

Ostrom's principles are why the protocol's disclosure rule reads the way it does: records move only with the person they concern, shown by them, never pulled in bulk, never searched by strangers. Any protocol that permits pulling records about a person without them converts that person's reputation from a commons they co-govern into an extractable resource they cannot defend. Ostrom's work is what tells us this failure is structural, not incidental. Where the earlier projects in this lineage gave us mechanisms, she gave us the criteria for judging them.

The closest relative: PCARE

One predecessor enters the family tree differently. Every ancestor so far shaped the protocol. Puja Ohlhaver's 2025 community-currency model, PCARE, shaped what Extol built on top of Attestation Notary: the token layer, which lives in one implementer's products rather than in the protocol every implementer shares.[^7] We credit it in the most detail anyway, because the work shapes important parts of what Extol builds, just not the open half.

PCARE offers a currency design for the economics of power: each community issues its own currency, and members choose between permanently locking it for influence or keeping it liquid to exchange outward. Community currencies are an old idea; two of PCARE's mechanisms for them are not, and Extol inherits both openly. The locked/liquid split is hers. And the subsidiarity condition is hers: local stake, meaning voice and standing, must weigh more heavily inside a community than global money does, or outside capital can simply buy local outcomes. Extol builds that requirement into how its communities draw their boundaries.

The two models answer different questions. PCARE prices influence: the weight a member's locked currency carries in a community's decisions. The protocol governs the record: who did what, witnessed by whom. Extol runs them side by side and never lets one pay for the other. Attesting earns no tokens, because the moment a signature makes money for its signer, the signal dies.

Influence weighted per person also invites the oldest attack in the genre: one person splitting into several accounts to be counted more than once. The durable defense has always been in-person, high-context community, where one body cannot show up as two, and witnessed attestation makes that defense native to the protocol. A trail crew that watched you swing a tool all season cannot be persuaded to sign for a person who does not exist. Extol's deployments are field data on whether that defense holds. We expect to keep reporting what we find.

The composition

Lay the ancestry end to end and what remains is not a missing idea but a missing integration. SAS notarizes. Verifiable Credentials and Open Badges define the envelope. Soulbound tokens supply the non-transferability constraint. PCARE structures influence within and between communities. Ostrom tells us how to judge the whole arrangement. Every piece exists somewhere in this lineage. No one had composed them into a single specification: testimony a person earns from witnesses, keeps for life, and shows on their own terms.

What the protocol adds is not novel cryptography and not a novel idea. It is the requirement that the pieces come as a set, because each one fails without the others. A signature without stake is cheap talk. A record anyone can pull is not reputation, it is a dossier. And a substrate that acts as a directory turns even a well-designed record into surveillance infrastructure. Each ancestor solved its piece. The protocol's work was choosing them, requiring all of them, and publishing the arrangement as a protocol rather than a product so that anyone can build on it.

Extol holds three roles today, author, reference maintainer, and first production network, because that is how specifications begin, not how they should end. If the protocol does its job, other implementations, networks, and maintainers will emerge alongside it. The specification is on GitHub now. Review happens where specifications live: in issues, in pull requests, and in the implementations that decide to build against it. That is how this family grows, one review and one implementation at a time.

[^1]: The laws in that essay are not coincidences. They are the predictable behavior of feedback systems: reward a measure and the measure corrupts; make a signal free and it inflates until nobody reads it. The tradition that teaches this way of reading is system dynamics; see Jay W. Forrester, "Counterintuitive Behavior of Social Systems" (1971).

[^2]: W3C Verifiable Credentials Data Model, a W3C Recommendation since 2019. The full cryptographic stack for portable, verifiable claims has been an open standard for years; its adoption history, not its design, is the lesson.

[^3]: Open Badges, originated at Mozilla in 2011, now stewarded by 1EdTech (formerly IMS Global). Mozilla's own retrospective identified the stalemate precisely: learners wanted proof employers would check badges, and employers wanted proof educators would issue them.

[^4]: Puja Ohlhaver, E. Glen Weyl, and Vitalik Buterin, "Decentralized Society: Finding Web3's Soul" (SSRN, 2022).

[^5]: Eric A. Posner and E. Glen Weyl, Radical Markets (Princeton University Press, 2018), and the quadratic voting literature that followed.

[^6]: Elinor Ostrom, Governing the Commons (Cambridge University Press, 1990); Nobel Memorial Prize in Economic Sciences, 2009.

[^7]: Puja Ohlhaver, "Community Currencies" (SSRN, 2025). Idena's sublinear identity staking is a precursor worth crediting alongside it.

Ken Leiserson builds tools for trust between people who work together. He can be reached at extol.work or on X @kenserson.

Reputation Is a Commons

Ken Leiserson — Sun, 02 Aug 2026 00:00:00 GMT

High above the Vispa valley in the Swiss canton of Valais sits Törbel, a village of sun-browned wooden chalets and about six hundred people. Its parish records run back to 1224. In 1483, its families signed a charter establishing an association to govern the things no family could own alone: the alpine meadows, the forests, the irrigation channels that carried glacier water to the fields. The charter's most famous rule fits in a sentence: no citizen may send more cows to the summer alp than he can feed through the winter.

Think about what that rule is doing. It ties a person's claim on the common pasture to work the whole village can verify, because every neighbor knows how much hay you cut; they watched you cut it. It makes overreach visible, cheap to catch, and embarrassing. And it has kept the alp green for five hundred and forty years.

When the anthropologist Robert Netting arrived in 1970 to study the village, colleagues asked why an Africanist was going to Switzerland. His answer became Balancing on an Alp (1981), a portrait of a community that had managed a fragile shared resource, without a landlord and without a state, for half a millennium. And Netting's work caught the attention of a political scientist named Elinor Ostrom, who was assembling something unusual: an empirical catalog of commons that had not ended in tragedy.

The third option

The standard story was set in 1968, when Garrett Hardin published "The Tragedy of the Commons." Every herdsman on a shared pasture, Hardin argued, captures the full gain of adding one more animal while the cost of overgrazing is spread across everyone. So each adds another, and another, and the pasture dies. The failure he described is real, and no one watching an unmanaged commons today needs convincing. But his prescription hardened into policy orthodoxy for a generation: a commons must be either privatized or nationalized, because the people who use it cannot be trusted to govern it. Törbel's meadow has never been privatized, has never been nationalized, and has never collapsed.

Ostrom's response was to go look. Törbel's meadows. The mountain commons of Hirano and Nagaike in Japan. The huerta irrigation communities of Valencia, whose water court has convened at the cathedral's Apostles' Gate every Thursday at noon for roughly a thousand years, and is often called the oldest functioning court on earth. Fisheries, forests, groundwater basins. Again and again she found the third option Hardin's model said couldn't exist: communities governing shared resources themselves, with no owner and no ministry, for centuries at a stretch. Governing the Commons (1990) distilled what the survivors had in common, and in 2009 the work won the Nobel Memorial Prize in economics.

Her famous distillation is eight design principles. Compressed brutally: boundaries: it is clear who is a member of the community and what the resource is; fit: rules match local conditions rather than a template; collective choice: the people affected by the rules can change them; monitoring: done by community members accountable to the community, not by outsiders; graduated sanctions: a first violation draws a warning, not exile; conflict resolution: cheap, fast, and local; the right to organize: outside authorities let the arrangement stand; and nesting: small units federate into larger ones rather than dissolving into them.

Note what the principles are made of. Not incentives, exactly, and not technology at all. They are answers to four questions any shared resource forces: who is inside, who watches, what watching costs, and what happens when someone takes too much. Answer them well and you have the third option. Leave them unanswered and you have Hardin's pasture.

The pasture made of trust

Here is the claim this essay exists to make: a community's reputation is a common-pool resource, and it behaves exactly the way Ostrom said such resources behave.

A common-pool resource has two defining properties. It is hard to fence: many people can draw on it. And it is subtractable: what one person takes is genuinely gone, at least for a while. A community's collective credibility fits both. When any member vouches, endorses, certifies, or signs, they are drawing on a shared stock of believability that the whole community built. And every hollow claim grazes it down. The pasture doesn't die when one cow overgrazes; it dies when overgrazing becomes free.

Once you see reputation this way, the last fifteen years of internet reputation systems reorganize themselves into a set of commons failures so clean they could be teaching cases.

LinkedIn endorsements are the overgrazed meadow. In 2012 LinkedIn shipped the exact concept, peers attesting each other's skills, atop the largest professional network on earth. But endorsing cost one click and nothing else, and the algorithm actively prompted it. Reciprocity spam followed within months: strangers endorsing strangers for skills no one had witnessed. Recruiters learned to ignore the feature within about a year, and it has been decorative ever since. Free grazing; dead pasture. The written recommendation, costly and specific and signed in full sentences, still carries modest weight on the very same platform, which tells you the failure was never the idea. It was the price.

Grade inflation is the same failure inside the university. An A is a claim drawn on the institution's collective credibility. When professors stopped paying any reputational cost for granting them, the A's multiplied and the transcript quietly died as a signal. Meanwhile the admissions gate, which stayed ruthlessly guarded, kept the brand alive. One commons inside the institution was governed; the other was grazed to mud.

And Klout was the enclosure movement. From 2008 to 2018, Klout computed a 0-to-100 influence score for hundreds of millions of people, most of whom never asked. It extracted reputation from every context at once and fused it into a single portable number, and it earned gaming, mockery, and a resentment deep enough to become its obituary. The sharpest post-mortem said, in essence, that reputation is the residue of action and belongs to the context where the action happened; strip it out of that context and it means nothing. Ostrom's first principle, the boundary, said the same thing decades earlier from the other direction. A commons is governable because it has an inside.

Different platforms, different decades, one diagnosis. These systems didn't fail for lack of users, capital, or cryptographic sophistication. They failed because nobody governed the pasture.

The natural experiment

Then, starting around 2016, the internet ran the largest commons-governance experiment since Hardin's essay. Thousands of them, in public, at speed, with real money.

Decentralized autonomous organizations deserve a more honest telling than they usually get. Strip away the coin-press cynicism that attached to the era and the ambition was almost precisely Ostrom's question: can a community govern a shared resource (a treasury, a protocol, a collective project) without a corporation above it or a state behind it? Tens of thousands of people joined in good faith. Real treasuries, in aggregate worth billions, were placed under genuinely collective control. Some jurisdictions even created new legal forms to let these communities hold property and sign contracts; a real institutional innovation, and worth keeping.

But the first generation built its governance from the design vocabulary it had inherited, which came from corporate finance: shares and votes. And measured against the eight principles, the standard design was not a partial violation. It was a near-perfect inversion.

Influence was linearly purchasable. One token, one vote, means the price of a governance outcome is quoted on an exchange. This is not a bug that crept in; it is the first-order property of the design.

There was no boundary. Anyone holding the token was a "member," anonymously, instantly, with no distinction between a founder of five years and an arbitrageur of five minutes. Ostrom's first principle, the one Törbel wrote down in 1483, was absent by construction. No inside means no context, nothing at stake, and no way to tell a neighbor from an opportunist.

Voting was costless, so almost nobody voted. Studies of the era's governance found turnout on most proposals running in the low single digits; even flagship protocols struggled to bring a tenth of their tokens to a vote. The vacuum filled with delegation to whoever asked: governance capture without governance activity. Monitoring by accountable community members, Ostrom's fourth principle, requires monitors who exist.

And there was no wall between reputation and speculation. When the governance instrument is also a tradable asset, every vote is also a position. A community cannot deliberate about its long-term good in an instrument its members are simultaneously day-trading.

The failure modes were structural, not cultural, which is why they arrived on schedule regardless of how idealistic any particular community was.

The cleanest demonstration took thirteen seconds.

On April 17, 2022, a lending protocol called Beanstalk was governed by token vote, with an emergency mechanism allowing a two-thirds supermajority to execute a proposal instantly. An attacker had submitted two proposals the day before: one, prominently, a $250,000 donation to Ukraine relief; the other containing the code that mattered. Then he borrowed roughly a billion dollars through a flash loan, a DeFi instrument in which the loan is taken and repaid inside a single transaction. For the length of that transaction, about one Ethereum block, roughly thirteen seconds, he held a supermajority of Beanstalk's governance. He voted, passed his own proposal, transferred $182 million out of the protocol, repaid the loan, and kept about $80 million, before most members knew a vote had occurred.

Resist the urge to file this under crime. The contract executed exactly as written. The vote was legitimate under the constitution the community had adopted. Beanstalk's members were not fools, and their loss deserves sympathy, not a smirk; many were volunteers who believed in the project and were left holding a collapsed token. What failed them was a governance design under which a thirteen-second member, with no name, no history, and no work witnessed by anyone, could hold more standing than everyone who had built the thing combined. Hardin's herdsman needed a season to ruin the pasture. Composable finance got it down to one block.

Ostrom's principles, violated wholesale, at speed, in public.

What the survivors would build

The temptation after Beanstalk is to conclude that online communities can't govern shared value at all: Hardin's conclusion, freshly reheated. That reading gets the evidence exactly backwards. Törbel is five hundred years of counterexample; the huertas are a thousand. The first generation of internet commons didn't disprove Ostrom. It replicated, at unprecedented cost and documentation quality, her account of what happens when the principles are ignored. Which means the design brief for a second generation is not mysterious. It is sitting in a book from 1990, and each violation has an answer.

Earned influence. The answer to purchasable influence is not to banish money. Ostrom never pretended markets away, and neither should we; her communities traded constantly (hay, water turns, labor) while keeping governance anchored in membership and contribution. The design translation: money can participate, but it cannot dominate. Recent mechanism-design work has made this precise. Puja Ohlhaver's research on community currencies, itself built consciously on Ostrom, shows how influence can grow sublinearly with committed stake: the tenth unit buys far less voice than the first, so concentrated capital flattens into diminishing returns. Add commitments that are permanent and non-transferable, stakes a member locks for good as a lasting bond to one community, and you get something capital cannot retroactively acquire: time. A patient contributor's years of standing become a governance position a wealthy newcomer can approach only slowly, on the community's terms. Purchased influence can exist in such a system; it simply arrives bounded, dampened, and subordinate to the earned kind.

The boundary. The boundary is the collateral. A community with a real membrane, legible membership, doors the community itself controls, recovers everything Beanstalk lacked: context, stakes, and the ability to tell a member from a visitor. And standing earned inside a boundary should stay scoped to it, because that is where the witnesses are. This is Klout's lesson read forward: portability of reputation, where it ever comes, will arrive the way it always has, through federations of communities that trust each other's rules (Ostrom's nested enterprises), not through a universal number.

Witnessed weight. Make voting weight something that had to be earned by doing witnessed work, and the electorate changes: the votes are cast by the people who did the work, each with standing to lose. This is also where Ostrom's monitors reappear in modern dress. Peer attestation, members vouching for specific work they saw, is monitoring by accountable community members almost word for word. It is worth being precise about what makes the monitors accountable, because there is no automated punishment waiting behind them, and there shouldn't be. The accountability is the membrane's own economics. Every signature is visible inside the boundary, so a hollow one is a public act among the people best positioned to know it was hollow. And everything a signer holds, standing and stake alike, is denominated in the community's credibility: debase the commons and you have debased your own account in it. A protocol that slashed automatically would be more dramatic, and less Ostrom. Her fifth principle assigns sanctions to the community, graduated and fitted to the offense, precisely because the village knows things no formula does. Törbel ran this exact design: your claim on the alp was tied to hay your neighbors watched you cut, and the punishment for cheating was theirs to choose.

The firewall. Make it absolute. The moment attesting to a neighbor's work earns the attester anything financial, testimony becomes advertising and the commons begins to consume itself. So the rule is easy to state. A community's currency can be traded, priced, even speculated on. Its testimony cannot, because attesting earns nothing, ever, and a signature is the one thing money cannot place. Markets get the currency. The commons keeps the witness.

None of this is speculative institutional design. It is amonng the oldest institutional design we have records of, restated for communities whose pasture is trust.

The instruments have caught up. Records can now outlive any platform, and verification can be performed by anyone. For the first time, a Törbel-shaped institution can hold together among people who do not share a valley.

The evidence, too, has caught up. The first generation of internet reputation systems and on-chain governance experiments failed in public, block by block. The second generation's designs will be tested the same way: in participation rates, in collusion attempts, in what real communities actually do with real boundaries.

Ostrom would have insisted on nothing less. Her whole career was a rebuke to armchair conclusions about what communities cannot do.

Thirteen seconds and five hundred years are both now data points in the same ledger. The village above the Vispa valley never heard of a flash loan, but it would have understood the attack instantly: a stranger arrived with a vast herd, grazed the alp bare in one afternoon, and left before dark. And it would have understood the remedy too, because it wrote the remedy down in 1483. Know who is in the village. Tie every claim to work the neighbors witnessed. Make the watching cost something and the overreach cost more. Let the village make its own rules, and let villages federate without dissolving.

Reputation is a commons, and Ostrom told us how commons survive.

Eight Laws from the Graveyard

Ken Leiserson — Sat, 01 Aug 2026 00:00:00 GMT

Every reputation system fails. Most fail the same ways. This is a field guide to the failures.

Ten specimens below, each a serious attempt, each with people who cared. Some got funding, some got users, some got press. All of them tried to make reputation portable, corroborable, or automated. None of them survived the way their designers hoped. What killed each is instructive: the failures cluster, and the clusters name the laws that any successor has to obey.

Read the specimens first. The laws follow, each paired with a survivor. That rare system, sometimes centuries old, that obeys the law and lives.

Ten specimens

Open Badges (2011–ongoing, hollow)

An open standard for digital credentials, developed at Mozilla in 2011 with MacArthur Foundation backing, later stewarded by IMS Global. A "badge" is a portable digital credential: JSON metadata about who earned what from whom, cryptographically signed by the issuer. By 2013, over 1,450 organizations were issuing badges; millions were minted. By the mid-2010s the standard was technically excellent and functionally unused. The problem was that nobody checked. Employers did not query badge repositories before hiring. Universities did not verify inbound badges before granting credit. A credential nobody looks at is a credential nobody has. Worse, adoption skewed toward populations without conventional credentials, and the badges came to signal the absence of the credential the reader was actually looking for. The standard survives today as plumbing inside enterprise learning systems, absorbed by the incumbent it was built to disrupt. Never a rival to it.

Died from: Law 1 (checking never emerges on its own) and Law 4 (adverse selection stigmatizes alternative credentials).

W3C Verifiable Credentials (2019–ongoing, hollow)

A W3C recommendation for cryptographically signed claims about a subject, expressed in a portable data format. Verifiable Credentials solved the technical problems Open Badges did not: better cryptography, better privacy, better composability. Adopted in specific verticals (educational transcripts, government identity documents) where a specific institution demands the credential. Not adopted as general reputation infrastructure, and for the same reason as Open Badges: outside those verticals, nobody checks. A well-designed credential is not a solution to the checking problem; it is a prerequisite for one.

Died from: Law 1 (checking never emerges on its own).

LinkedIn Endorsements (2012–ongoing)

One-click endorsements of colleagues' skills, launched September 2012. Users click a button, endorsement appears on the recipient's profile. Frictionless and viral: two hundred million endorsements in the first month, a billion within six. And within about a year, recruiters had learned to ignore them entirely, recognizing them as reciprocity spam and algorithm-prompted strangers endorsing skills they'd never witnessed. The written recommendation, costly and specific, retains modest value to this day; the one-click endorsement retains none. The word for a sign-off that costs nothing is noise.

Died from: Law 2 (attestation value is proportional to attestation cost).

POAP (2019–2026)

"Proof of Attendance Protocol": a system for issuing on-chain badges to event attendees, born at the ETHDenver hackathon in 2019. Show up, scan a QR code, receive a POAP that travels in your wallet. Portable, verifiable, permanent, with millions minted. The design solved attendance-recording cleanly. It also demonstrated that attendance is not attestation. A POAP proves you were physically present; it does not prove that you contributed, or that anyone would vouch for you. Collectors accumulated POAPs by the hundreds. Gatekeepers do not read them, because "was in the room" is a very small unit of information about a person, and one available essentially for free. Cryptographic verifiability adds nothing when the underlying claim is costless. The wind-down announced in August 2026 confirmed a second failure mode: the badges live on-chain, but the meaning-and-discovery layer dies with the platform.

Died from: Law 2 (attestation value is proportional to attestation cost; showing up costs almost nothing).

Klout (2008–2018)

A single 0–100 score of a person's online influence, computed from social media activity — for hundreds of millions of people, most of whom never asked. For a moment it had real traction: conference organizers checked Klout scores when booking speakers, and the number showed up in hiring conversations. Then people noticed the score was gameable, opaque, unrequested, and imposed from outside any community the person belonged to. Public revolt. Lithium Technologies acquired the company for roughly $200 million in 2014 and shut it down in May 2018. The deep failure was structural: Klout tried to make reputation portable by extracting a score from every context and averaging it. Reputation resists that extraction. A number abstracted from the settings where it was earned is not portable reputation; it is a fiction dressed up as one.

Died from: Law 3 (reputation resists leaving its context; portability requires federation, not extraction).

Braintrust (2018–ongoing, hollow)

A talent network for freelance knowledge workers, founded in 2018, out of stealth in 2020, structured as a decentralized cooperative and monetized with a native token launched in September 2021. Talent kept full contract value; clients and members earned tokens for hiring and for referring peers. The mechanism that hollowed the community was subtle: paying members for the act of referring converted referrals from social signal into economic activity. Referrers optimized for volume rather than fit. "I refer you because I know your work" became indistinguishable from "I refer you because I earn tokens if you sign." The token peaked within weeks of launch and has since lost effectively all of its value. Meanwhile the community discovered it had been running on economic activity that no longer paid enough to sustain the behavior.

Died from: Law 5 (paying for prosocial acts tends to destroy them; the referral was the prosocial act).

Stack Overflow reputation (2008–ongoing, decaying)

Question-and-answer platform for programmers. Users earned reputation by answering questions upvoted by peers, and for a decade that reputation was a genuinely useful signal. Rigorous public participation, visible to peers, audited by voting. Note why it worked: reputation gated real internal resources such as editing rights, moderation tools, closing votes. But it lived entirely inside Stack Overflow. The platform tried to externalize it and failed: Stack Overflow Jobs was discontinued in March 2022, because employers never checked SO reputation at scale despite fifteen years of perfect data. Then came the harder ending: as language models trained on the site's own answers reduced question volume and veteran contributors drifted away, the reputations stayed on profile pages and stopped meaning anything outside them. Users who spent a decade accumulating rank discovered the rank was a lease from a platform that no longer prioritized it. The coda is sharper still: the platform went on to sign data-licensing deals with major AI labs, selling access to the very answers its contributors had written for free with no record of answerers receiving any cut. The standing was never convertible, even at the moment the work itself was finally monetized. By someone else.

Died from: Law 6 (platform-custodied standing dies with the platform), with a supporting diagnosis of Law 1 (external checking never formed, even with perfect data).

SourceCred (2018–~2022)

An open-source contribution measurement system that used graph algorithms on GitHub, Discord, and Discourse to compute per-person "cred" scores, then distributed treasury allocations proportionally. Well-designed, mathematically rigorous, thoughtfully governed by the projects that adopted it. Two failure modes appeared quickly. First, the moment cred paid, contributors optimized for what the algorithm counted rather than what the community needed: comment frequency skyrocketed, and it was clearly farming. Second, communities fractured over the scoring function; every parameter change felt like reallocating money because it was reallocating money. Several major DAOs adopted, then abandoned. The project wound down around 2022.

Died from: Law 7 (measurement politicizes; peers collude around whichever parameters favor them).

Coordinape (2021–ongoing)

Peer-attribution mechanism for crypto-based decentralized autonomous organizations (DAOs). Each member gets tokens to allocate to other members each epoch; allocations become the basis for treasury distribution. Elegant, small-group, low-friction. Fails when the group is either too small (collusion is trivial; three people can agree to allocate to each other) or too large (voters don't know each other well enough for the allocation to reflect real observation). The sweet spot is roughly ten to thirty people who all work together intensely which is a limited market, and one that many DAOs outgrow within a year.

Died from: Law 7 (peers collude; the allocation game reveals coalitions faster than it reveals contribution).

On Deck (2019–2022 in its original form)

Cohort-based programs marketed as elite communities of ambitious founders and operators. Selectivity was the value proposition: get accepted and you're in a room with peers who matter. Growth pressure led to more programs, larger cohorts, weaker selection: over ten thousand founders and professionals at peak. The acceptance rate climbed to accommodate the business model's need for revenue, and "On Deck alum" carried less signal each year precisely because there were so many of them. In 2022 the company went through two rounds of layoffs months apart, shuttered and refunded programs, and retrenched to a fraction of its ambition. On Deck arrived at the selectivity trade-off the way every credential does, and made the choice most business models require: accept more, dilute the signal, extract revenue from the volume.

Died from: Law 8 (credential value is proportional to selectivity; growth pressure destroys the credential it was distributing).

The Eight Laws

Each of the ten specimens is diagnosable by the same eight laws. The laws are not novel; the specimens surface them repeatedly. Each comes with its survivor: proof the law can be obeyed.

Law 1: Checking never emerges on its own

You cannot build a credential system by publishing well-formed credentials and hoping the demand side arrives. Open Badges and Verifiable Credentials are the paradigm cases: technically sound infrastructures that solved every problem except the one they most needed to solve, which is that nobody looks. Every project in this graveyard solved supply. Not one solved demand. Checking emerges only when the population that signs is also the population that reads: the same person attests on Saturday and sits on the grant panel on Tuesday. There is no other documented mechanism. Any system that treats issuance and verification as separate populations will discover that the verification population never forms.

The survivor: the volunteer coordinator who signs records and later reads them from the other side of the table. She checks records because she makes records; she knows what they cost. Demand isn't a market you build. It's a habit that travels inside individuals who inhabit both sides of the record.

Law 2: Attestation value is proportional to attestation cost

The signature that survives is the one where the signer has skin in the game. LinkedIn endorsements and POAPs sit at the same costless end of the spectrum, one on a social graph and one on a blockchain: the one-click endorsement is nearly free, "I showed up" is nearly free, and both produce records that carry the corresponding weight. The substrate changed; the physics didn't. The cost need not be money; it can be reputation, time, or the observable act of putting one's own standing next to someone else's. But it cannot be zero. Costless testimony is a different thing that borrows the vocabulary of real testimony.

The survivors: the term sheet. An attestation collateralized with actual money. The reference letter: costly, specific, still read. The guild mark. Collateralized with collective reputation and enforced by the members who share it. Note the kind of cost that works: staked reputation on a specific witnessed claim. Not solely effort. Effort just kills completion. The signer must have something to lose if the claim is false.

Law 3: Reputation resists leaving its context

Klout tried to build portability by extracting a score from every context and averaging it. The extraction destroyed the very thing it was trying to make portable. Reputation earned in one community means what it means because the people reading it (the gatekeeper, the grant panelist, the receiving local) belong to the community that watched the work happen. Strip the context away and the signal collapses into a number everyone can game and nobody can trust. Every universal score built so far has been rejected by the people it purported to represent.

The survivor: the IBEW traveler system. A journeyman electrician can walk into a distant local, sign the books, and work. Not because his reputation was converted into a universal score, but because the receiving local trusts the issuing local's training, through a shared institution in which both locals have standing staked. That's the ingredient Klout never had: a vouching party with skin in the game. Nationalization solved the same problem to a more limited degree: state licensing makes competence portable within a jurisdiction by enforcing minimum standards. But a license ports only the floor, and only for work the state chooses to license. The traveler card ports the whole reputation. Portability was never a cryptography problem. It's a federation-of-trust problem, and only guild-shaped institutions have ever fully solved it.

Law 4: Adverse selection stigmatizes alternative credentials

Open Badges was rejected in most markets not because the technology failed but because the population that adopted it, on average, was the population that could not or did not go through the standard credential. A credential adopted disproportionately by people avoiding the standard credential gets read as evidence of failure to meet the standard. This is the trap every alternative credential falls into.

The survivor (really an inversion): Facebook launched Harvard-only. A credential's status is set by its least needy holders. Seed a new form of recognition with people who conspicuously don't need it, people with impeccable establishment credentials who choose the new one anyway, and the stigma reverses; the new credential borrows their standing before it can lend any of its own.

Law 5: Paying for prosocial acts tends to destroy them

This is the Titmuss lesson (1970), refined by fifty years of motivation-crowding research that followed: paying for a prosocial act tends to reduce it, because the act stops signaling generosity and starts signaling economic activity. Titmuss's specific claims about the 1960s blood supply have been substantially refined (paid plasma donation now supplies most of the world's therapies because plasma can be industrially processed), but the underlying finding has replicated across domains, and the mechanism turns on what the payment is contingent on. Braintrust rediscovered it for referrals. The moment reward attaches to the act of endorsing, referring, or vouching, members convert into optimizers of the reward function; the behavior that used to signal care now signals farming, and the signal is dead within a quarter.

The fine print of the crowding-out literature is a design specification, and it turns on what the payment is contingent on. Payment corrupts a prosocial act when it is offered as a bounty for the act, to the person performing it. Titmuss's donor, Braintrust's referrer.

The survivor: the honor. The earned title (master, fellow, laureate) the standing that gates access and voice but cannot be sold. And the law's "tends to" is precise, because the destruction is reversible: a Swedish field experiment (Mellström and Johannesson, 2008) that reran Titmuss found cash payment cut blood donation among women nearly in half, but that the effect disappeared entirely when donors could redirect the payment to charity. Same money, prosocial frame, signal restored. The requirement was never that money be invisible, it's that the system be inspectable: signing pays nothing, standing can't be transferred, and anyone can verify both.

Law 6: Platform-custodied standing dies with the platform

Stack Overflow proves this in slow motion. Reputation held only inside a platform is on lease from the platform's continued attention; when the platform's priorities changed and its traffic collapsed, twenty years of earned standing stayed on profile pages and stopped meaning anything outside them. The successor pattern separates the notarization of a record from the custody of its display: the record's authenticity should not depend on any single platform's continued existence, only on the durability of the notarization substrate.

The survivor: the county recorder of deeds. Land titles outlive the clerks, the software, and sometimes the governments that recorded them, because the record's permanence was engineered as a separate concern from any institution's fortunes. Community members' records should be able to outlive any company, including the one that hosts them.

Law 7: Measurement politicizes; peers collude

Systems that compute a score from an algorithm are politically fragile: every parameter is a fight, every recalculation is a reallocation, and every value chosen is contestable. SourceCred experienced this repeatedly. Systems that let peers allocate directly face the opposite failure: coalitions form and allocate to each other regardless of contribution, as Coordinape found in every group large enough for coalitions. The stable point sits between these failure modes: human attestation of specific witnessed work, attributed to specific witnesses. That structure is durable because the graph of who signed what is not a fight; it is a fact, subject to interpretation but not to negotiation. The pressure toward favoritism never disappears on that ridge, which is why the signing graph must be visible and governed from the start — not bolted on after the first scandal.

The survivor: Modern court systems have needed to establish contribution and fault for centuries, and they have never once done it by computing a score or by polling the room. They call a witness, a named person who saw a specific thing, and make the testimony costly: under oath, on the record, punishable if false, cross-examined in front of everyone. Not a measurement, not a preference; a fact with a name attached, governed visibly from the first word.

Law 8: Credential value is proportional to selectivity; growth wants volume

The credential earns its worth by excluding. Growth strategies that dilute the excluding set destroy the credential they were trying to distribute — On Deck's story in one sentence.

The survivors: Y Combinator, which stays ruthlessly selective despite every incentive to expand. Harvard, which hasn't meaningfully grown its class in decades despite a fifty-billion-dollar endowment, because the rejection rate is the product. Michelin, whose anonymous inspectors stake the Guide's name on every star. Michelin also passes Law 3: a star means the same thing in Tokyo as in Lyon because a single institution with staked standing binds every context it appears in. The law's reach extends to anyone who builds infrastructure for recognition: the infrastructure should never try to be the credential. The selectivity burden belongs to each community that issues one, and infrastructure that monetizes member volume will produce credentials whose value is inversely proportional to its own success.

Photo by Michelle Li on Unsplash

What this rules in

Read backward, the eight laws are ruthless. Laws 1 and 3 alone kill the grand version of every credentialing dream, because external verification demand has never spontaneously formed and reputation has never survived leaving home. But the laws don't prohibit everything.

Overlaid, the eight laws describe the shape of a successor. Checking flows through the same population that signs. Each signature risks the signer's own standing, and never earns them anything. Reputation remains rooted in where it was earned; portability works through federated recognition between communities, not by extraction into a universal score. Records outlast the platforms that host them. Seeding uses people who don't need the credential, so its early adoption resists stigma. The infrastructure records specific witnessed acts rather than computing scores from parameters, and never tries to become the credential itself. The credential belongs to each community, on the community's own terms.

None of the ten specimens got all eight right. Several got most of them wrong. The successor need not be more clever than any of them. Only more disciplined about which failures the graveyard has already documented, and structurally incapable of repeating the ones for which the record is now clear.

Every one of these projects believed it was the exception. The laws don't care what you believe. They only care what the design costs the signer, who owns the record, and whether anyone was ever going to check.

Ken Leiserson builds tools for trust between people who work together. He can be reached at extol.work or on X @kenserson.

The Last Signal

Ken Leiserson — Thu, 30 Jul 2026 00:00:00 GMT

In 2021 a well-crafted cover letter still meant something. It was evidence of care, of effort, of a candidate who could compose their own thoughts. In 2026 the same letter arrives, just as well-crafted, from a candidate whose recommenders drafted one paragraph and delegated the rest, whose portfolio is polished and may or may not be their own work. The letter, the degree, the portfolio, the recommendations: all still arrive. What no longer arrives with them is a signal.

This is not a story about AI. It is a story about verification.

The economy of expensive signals

Every institution in the modern economy runs on signals that used to be expensive to produce. A cover letter cost time. A published paper meant a peer review committee stayed up late. A licensed contractor cleared a board's exam, in a room, with a proctor watching. A five-star review came from a real person who had actually eaten dinner. A GitHub commit meant somebody had thought carefully enough about a problem to type a solution and put their name on it. Each signal cost something, and that cost was the reason to trust it. When a signal becomes cheap to produce, it stops being a signal. The word for a signal that costs nothing is noise.

The list of signals that have collapsed in the last three years is longer than most people notice. Undergraduate essays. Coding samples for junior developer positions. Book reviews. Product reviews. Testimonials on real estate agents. The friendly outreach message that used to signal a real person had read your work. AI did not destroy any of these directly. What the AI destroyed was the cost of producing a plausible version of them, which is the same thing.

The vacuum is not new

The printing press did this once already. Before 1450 holding a book meant truth and status. Books were rare, expensive, laboriously reproduced by hand, and their content was implicitly trusted because the friction of production filtered out the frivolous. After 1550 the same book was one of ten thousand. The signal of "I have written this down" collapsed. The response was not to un-invent the press. The response was a century of confused publishers, forgeries, and pamphlet wars, and then verification. The institutions invented to fill the gap were new, but the mechanism inside them was not. The chartered societies and learned academies that emerged to vouch for a new discipline called science were guilds in everything but name: admission by the testimony of members who knew your work, standing that could not be printed, only conferred. And when the first scientific journals wanted their papers believed, they had practitioners read them first and put the society's name behind them. We call that peer review now. It is witnesses signing.

Verification is what filled the vacuum the press created. It took a while. The Royal Society was chartered two centuries after Gutenberg. Many of the institutions we now think of as timeless were built or remade in that gap, one by one, as answers to a specific question: given that anyone can now produce a plausible book, how do we know who to listen to?

We are in that vacuum again. Not because writing has become cheap; writing became cheap generations ago. We are in it because everything that used to be evidence of care, effort, or authorship can now be produced without any of those things, in seconds, by a machine or by a person with the machine's help. The paper, the code, the letter, the design, the review. All still arrive. None of them any longer prove that anyone in particular did the work.

What survives

There is one signal this doesn't touch, and it is the oldest one: someone who was there, staking their name on what they saw.

A colleague who reviewed your code and can tell you what surprised them about the design. A collaborator who worked on the album and can name what you contributed on which track. A student who was in your lab for three years and can describe the specific patience you showed them the week they wanted to quit. This kind of testimony can be lied about, but it cannot be automated, because the thing being spent is not words. It is a reputation that took years to build, staked by a person who has something to lose if the sign-off proves false. Their reputation lives with the mark they leave on yours. That is the friction the printing press could not eliminate, and the friction the current generation of tools cannot eliminate either.

This is not a new kind of trust. It is the oldest kind. Guilds ran on it. Bar associations run on it now. Every real reference offered voluntarily, on a candidate someone actually worked with, is one instance of it. Every editor who has ever put their name on a piece by a young writer, every senior engineer who has vouched for a junior in the promotion committee, every board member who has signed off on a treasurer's audit, is participating in an institution built on witnessed testimony.

What is new is that we have never systematically recorded it. The closest the digital era has come is the LinkedIn endorsement: a one-click vouch, offered in seconds, often for a skill the endorser never saw used. It failed exactly the way this essay would predict; a sign-off that costs nothing is noise. The best we have managed on paper, for the last three hundred years, is the letter of recommendation, which is impossible to check and hostage to the writer's willingness to spend an hour on someone else's career.

Recording it

The problem is recording, not trust. Trust between people who work together has always existed. What we've never done is record it in a form that survives, owned by the person who did the work, and shown when they choose to show it. Never pulled from a database behind their back by an employer or a recruiter or a search engine.

The vacuum the press created took a century to fill. This one will not take a century, because we already have the infrastructure to record durable sign-offs in real time, at costs that used to be impossible. The signature of a colleague, the vouch of a mentor, the certification of a professional body, the peer reviews in a journal. Each of these can now be recorded in a form that outlasts any single company and belongs to the person named in it. What is required is the collective will to record what actually matters, to protect it in ways that make it worth doing, and to keep it out of the hands of the aggregators that would immediately turn it back into another cheap signal.

The unverified signals are gone. What's left is the signature of someone who was there.

Ken Leiserson builds tools for trust between people who work together. He can be reached at extol.work or on X @kenserson.

The Guild Never Died

Ken Leiserson — Thu, 30 Jul 2026 00:00:00 GMT

The word "guild" survives mostly as an insult. Protectionist, exclusionary, medieval. A closed shop of masters keeping the trade small and the price high. The story ends with the guilds swept away by industrialization, replaced by open markets and modern professions, and good riddance.

The story is wrong in an interesting way. The guilds were not swept away. Their functions were absorbed, one at a time, by four institutions that took over the work of vouching for competent people. Each institution made the guild's job easier by doing part of it. Each is now decaying. And the collapse of unverified signals in the age of cheap production is arriving at exactly the moment those four hosts struggle to carry the weight.

This essay is about what guilds actually did, who took over the work, and what happens when the successors fail.

What a guild was

A guild, at its functional peak in the fifteenth century, did five things. Ignore the ceremonial hats and the trade-secret propaganda and you find a working institution running five rings around a craft:

Mutual aid. When a member fell sick, was injured, or died, the guild's treasury paid the widow. It buried the dead. It fed the apprentices during hungry years. It ran the medieval equivalent of insurance for members who had none.

Quality certification. A guild-mark on a piece of work meant a specific human had signed off on it. That mark could be verified: the master who made the mark was named, findable, and answerable if the work failed. Buyers trusted the mark because a specific person's standing was staked on it.

Market access. In most towns you could not sell your work without membership. This is the ring that later generations remember most, usually with contempt. But it was inseparable from the certification: the guild controlled the market because the market wanted quality guarantees, and the guild was the only institution capable of providing them.

Self-governance. The guild wrote its own rules, chose its own officers, disciplined its own members. When a master cheated a customer, the guild court decided the case, not the town. This is where "governance" happened in a life before parliaments, and it is why a guild could speak to the town with one voice when it needed to.

Training pipeline. An apprentice bound himself to a master for seven years. He worked, watched, took a modest allowance, and eventually, if the master vouched for him, became a journeyman. Then, if he demonstrated a masterpiece and enough masters signed off on it, he became a master himself. His trade was learned inside a community of practice that observed him for a decade before certifying him.

Five rings. One institution. What made the guild work was that all five happened together, in the same community, with the same members holding standing in each.

The four hosts

Guilds did not die. Their five rings were absorbed, unevenly, by four institutions over three centuries. The order below is functional, not chronological; the hand-offs overlapped, ran in parallel, and were never clean.

Nationalization took Ring 2 and shrank it. Licensing boards, first for medicine, then law, then engineering, plumbing, electrical work, cosmetology, real estate, over a hundred trades in the United States, replaced the guild-mark with the state-issued license. But the license does a smaller job than the mark did. A guild-mark staked a named person's standing on a specific piece of finished work. A license certifies a minimum competence threshold: this person is safe to practice at all. The floor is real and valuable. It is also all the state can administer, because a bureaucracy can examine a candidate once but cannot witness a career.

Industrialization took Ring 5. The factory taught its own workers. The apprenticeship model collapsed because the assembly line did not need masters; it needed workers competent enough to do one repeated task. Where trades survived, the guild's training pipeline was replaced by the vocational school, the technical college, or in-house corporate training. The community of practice fragmented into instructors, students, and the personnel departments who signed the paperwork.

Unionization took all five. This is the most important host to understand, because it is the only one that attempted a full transplant. A twentieth-century union ran a mutual-aid fund, certified skill through the union card and the apprenticeship credential, controlled market access through the hiring hall, governed itself through elections and locals, and trained the next generation at scale. For a generation, the union was a guild wearing modern clothing. Its problem, as we will see, is not that the transplant failed. It is that the transplant covers a shrinking sliver of the people who work.

Universitization took Rings 2, 3, and 5, and fused them into a single credential. The four-year degree became certification (the school signs off on you), market access (the degree is the ticket to the profession's entry-level positions), and training pipeline (four years of apprenticeship, though we do not use that word). This is the host that swallowed the largest fraction of the guild's functions in the twentieth century. It is also the host under the most visible strain right now.

Four hosts. Each took a piece of what a guild once did. None took all of it. And each is now, for reasons the next four sections name, failing at what it took.

Nationalization is retreating

Occupational licensing never stopped growing. Roughly one in twenty American workers needed a state license in the 1950s; today it is roughly one in four. By headcount, the regime is at its historical maximum. And yet the fraction of real skilled work the license covers is shrinking every year, because two things outran it.

First, work fragmented. A licensed general contractor still exists, but the work that used to require the license (a homeowner hiring someone to fix a shower) increasingly gets done by an unlicensed handyman found on an app, or a semi-professional who watched a hundred YouTube videos and can do the job well enough. The license still exists. The market that used to require it has grown around it.

Second, the trades themselves proliferated faster than the licensing regime could follow. Every state has a licensing category for cosmetology and none has one for the person editing a podcast for a living. Every state licenses electricians and none licenses the person building a dashboard for a small business. The categories the state chose to certify are the ones that existed in the 1970s. The fastest-growing categories of skilled work (anything digital, anything cross-functional, anything hybrid) the state does not touch.

The result is a paradox: the licensing boards are bigger than ever and still perform their function well inside their scope. Their scope, as a share of the real economy of skill, gets smaller every year. And even inside the scope, the license certifies only the floor. Whether this particular electrician is exceptional or merely minimum-viable is recorded nowhere; it lives in foremen's heads and reference phone calls.

Industrialization is over

The corporate ladder, the internal training pipeline that inherited Ring 5 from the guilds, was dismantled in a generation. The industrial firm that trained its own workers, promoted from within, and offered a career-long relationship in exchange for loyalty is gone. What replaced it is a labor market that hires from outside, holds workers for three to five years, and expects them to arrive already trained by someone else.

This did not stop people from learning at work. It stopped anyone from underwriting the learning. A firm that expects to hold a worker for three years has no reason to invest in a seven-year development arc, and even less reason to certify what the worker learned in a form the next employer could trust: every credential it issues makes its own people easier to poach. So the learning still happens, often more of it than four years of lectures ever produced. It just happens incidentally, unevenly, and off the record. Every graduate is now expected to arrive with skills a corporate pipeline used to build deliberately, at their own expense, on their own time, with no institution vouching for whether they actually acquired them.

The community of practice that used to observe an apprentice for seven years does not exist for most modern workers. The people who do exist (the mentor, the senior colleague, the manager who took an interest) can only record what they observed in forms that do not last, cannot be checked, and cannot be carried: a reference call that evaporates when it ends, a recommendation paragraph that is unverifiable by design, a one-click endorsement that recruiters learned years ago to ignore. And whatever record does survive is siloed where it was made. The performance review stays in the old employer's HR system. The recommendation lives on a platform that owns it. The manager who saw the work moves on, and the worker starts the next job as a stranger. The observation is real. The record is informal, perishable, and stuck.

Unionization is narrow

Here is what almost every account of union decline gets wrong: where the full guild transplant survives, it works.

Look at a building-trades union like the IBEW. Run the five rings against it. Mutual aid: health funds, pension funds, strike funds, death benefits. Certification: the union card and a joint apprenticeship credential widely regarded as the gold-standard training path in the trade. Market access: the hiring hall, where the union itself allocates jobs to members. That is the guildhall, functioning, today. Self-governance: elected officers, local bylaws, member votes. Training pipeline: apprenticeship programs funded by employer contributions, training the next generation at scale. Five for five. The traveler system even lets a journeyman from one local sign the books at another local across the country and work under its jurisdiction: standing that transfers, on paper, across borders. The medieval guild would recognize every piece of this.

The failure is not function. It is coverage, and the reasons for the collapse in coverage are structural, not sinister. The full ring stack is expensive to run, and someone has to pay for it: employers, through wages and benefit contributions they spent half a century fighting to escape, and members, through dues that only make sense against a long career in one trade. The model was built for the fixed shop and the stable industrial firm, and it moved at the speed of the contract cycle while the economy around it sped up, fragmented, and moved offshore or online. American private-sector union density peaked in 1954 at around thirty-five percent. It is now about six percent. The full ring stack survives in government work and the building trades and almost nowhere else. And the fastest-growing parts of the workforce (gig work, freelance work, creative work, care work) are structurally out of reach: no fixed employer to bargain with, no shop to organize, no hall to run.

So the union is neither a success story nor a cautionary tale. It is evidence about the model: the five rings still work where the conditions that fund them still hold, and those conditions now cover a tenth of the people who work. The vocabulary of solidarity spread everywhere. The institutional machinery that gave the vocabulary meaning serves fewer workers every decade, and nothing about the machinery as built can be stretched to cover the rest.

Universitization is fraying

The university is the most interesting failure because it is the largest. It absorbed three of the five rings (certification, market access, and training) into a single instrument, the degree, and rode that instrument for a century. The instrument is now losing on three fronts simultaneously.

Its knowledge monopoly is gone. YouTube, Wikipedia, MOOCs, and the current generation of AI tutors have collapsed the marginal cost of learning almost any subject to zero. The reason to attend a university has not been "access to knowledge that is otherwise unavailable" for at least a decade. Everyone in higher education knows this. The rhetoric of "the college experience" is what filled the gap.

Its screening credibility is eroding. When forty percent of the adult population holds a bachelor's degree, and grade inflation has severed the connection between a good transcript and demonstrated capability, the degree stops functioning as a scarce signal. Employers know this. Two decades of employers requiring degrees for jobs that did not need them was a rational response to a signal in inflation; the more recent turn of major employers dropping degree requirements is the honest admission that the signal is depleted.

Its community function was never institutionalized at all. The networks formed in four years are real, and they remain one of the strongest reasons to attend: the roommate who refers you into a company, the classmate who becomes a co-founder, the weak ties that surface opportunities for decades. But all of that value persists as private folklore. The university treats its alumni as a donor base, contacted twice a year with a giving appeal. It keeps no record of who actually worked alongside whom, on what, and how it went, and it offers no mechanism for one former classmate to vouch for another's work five years later in a form anyone else can check. The most durable thing the university produces is the one thing it never learned to record.

The degree remains. What the degree used to certify (capability, standing among peers, membership in a working community) now travels by word of mouth or not at all. The university is left competing on the college experience alone, and that experience costs six figures.

The vacuum arrives

Four hosts, each carrying a piece of the guild's work, each now hollowed. Meanwhile the signal collapse the previous essay named, the AI-driven cheapening of every signal that used to require effort, arrives in the same decade.

This is not a coincidence, but it is not a conspiracy either. The hosts were weakening for structural reasons that predate AI: labor market restructuring, credential inflation, union decline, licensing scope shrinkage. AI is not the cause. AI is the accelerator that makes the underlying weakness impossible to paper over.

A licensing board can still certify a plumber. It cannot certify a data analyst, a video editor, a community organizer, a grant writer, a technical program manager, or the tens of millions of other modern workers whose actual competence is invisible to any state-issued instrument.

A degree can still open a door. It cannot describe what its holder actually did during four years, who worked with them, what problems they solved, or whether they can be trusted with a real one.

A union can still run the full stack for the six percent it reaches. It cannot reach the rest.

A firm can still teach. People learn more on the job than they ever learned in a lecture hall. But the learning is incidental to the work, the record of it sits in systems the employer owns, and no firm has any reason to certify a worker for the benefit of the next one.

The five rings of the guild are unattended for most modern work. What replaced them cannot do the work at scale for the categories of work most people now do.

What comes next

The previous essay noted that before Gutenberg, a book verified truth and status, and after the press it signaled nothing, because production had become cheap. That was a story about reputation. The story that matters for this essay is about skill, and it is more precise.

Before Gutenberg's printing press, you could not learn goldsmithing or glassblowing without physical proximity to a master who chose to teach you. The techniques existed nowhere else. Knowledge scarcity was the guild's deepest moat. Then the press made the trade manual publishable. Agricola's De re metallica laid out mining and metallurgy in exhaustive detail for anyone with the money to buy it, and books like it did the same for craft after craft. The moment the technique was on sale, the guild-mark's implicit claim ("only we know how to do this properly") began to collapse. What could not collapse was the other half of the mark: not "we alone know how," but "this specific person did this specific work, and I watched."

The response, over about a century, was the invention of new institutions built on that surviving half: chartered societies, learned academies, peer-reviewed journals. Every one of them was a body of practitioners who read each other's work, staked their reputations on their evaluations, and left a record of who had signed off on what. The Royal Society admitted new fellows through the testimony of existing fellows. Its journal was built on the premise that a claim was trustworthy if credentialed practitioners had read it and put their reputation behind it. Peer review is not a modern innovation. It is what the guilds always did, restaged for a new domain where the old domain's rules did not fit.

We are in the same position now. YouTube, TikTok, and the LLM tutor have collapsed the moat around knowledge; AI has collapsed the cost of producing any plausible artifact (a portfolio, a paper, a review, a letter, a resume) to nearly zero. The signals that used to indicate care, effort, or competence are noise. The response, if history rhymes, is not to un-invent the press. It is to build institutions on the half of the mark that survives cheap production: the witness who was actually there, signing.

Two lessons from the aftermath

The history after the guilds' formal end teaches two things, and they point in opposite directions. Both matter for anyone building what comes next.

The first lesson is from 1814, when Parliament repealed the Statute of Artificers and guild membership no longer gated practicing a trade in Britain. Most histories treat this as the guilds' extinction. It was not. The guilds whose internal value was real (whose masters actually vouched for competence, whose training actually produced skilled tradespeople, whose reputation among customers was earned) kept operating for decades, and some persist today as livery companies and professional bodies. What died in 1814 was the external gate. What survived was the internal function. The lesson: build so that membership is worth the effort on the internal merits alone, with no state license, corporate mandate, or degree requirement propping up demand. An institution that survives only because it holds a monopoly is one policy change from irrelevance. Externally mandated adoption is upside, never a dependency.

The second lesson is from what filled the gap. In Victorian Britain, working-class mutual-aid societies (the Foresters, the Oddfellows, the Rechabites, dozens more) grew to enroll about one in three adult British men by 1900. They ran the funerals, the sickness insurance, the widow support. They were serious institutions with real scale, and they almost entirely vanished within two generations, because when the state arrived (National Insurance in 1911, then the modern welfare state) it absorbed everything about them that could be standardized: the fixed benefit, the uniform payout, the national scheme. What a bureaucracy can do at scale, a bureaucracy will eventually do at scale. What it could not absorb, and what no state will ever administer, is the part that depends on specific human judgment about specific work.

Put the two lessons together and they mark out the durable ground. It is the ring that needs no external gate to be valuable and resists standardization by any central actor: the peer signing off on the peer, the specific human staking standing on a specific piece of work. That ring, the guild-mark ring, requires a witness who was actually there. It always has.

What comes next

The four hosts are hollowing. The vacuum from the previous essay is arriving. The historical pattern is clear: when a certifying institution no longer covers most of the trade it certifies, its certification stops carrying signal and new institutions emerge to do the job.

Two things are different this time in ways that matter.

The first is that we now have the technical ability to record durable, portable sign-offs at costs that used to be impossible. A witness signing for a colleague no longer needs the guild court to convene, the scribe to record, the seal to be applied. A cryptographic signature over a specific claim, committed to a public record, is the digital-native form of the mark a master used to press into a piece of finished silverwork. The mechanism is old. The medium is new, and it is fast enough, cheap enough, and durable enough to run at the scale of the modern workforce.

The second is that no state is going to build this, no corporation is going to build this, and no union is going to build this at the scale required. The institutions that took over from the guilds are, by definition, the institutions that are decaying.

The successor will look more like the friendly societies at their inception than like a licensing board: voluntary, federated, member-serving, resistant to standardization. It will succeed to the extent that its members find it worth participating in on the internal merits, with or without any external gate.

The five rings are not going to be rebuilt as a single institution. The mutual aid ring will remain mostly an employment benefit, with some moving to community treasuries and cooperative funds. The training pipeline will add a mix of AI tutors, peer and mentorship networks, and specialized bootcamps. The market access ring may go to the marketplaces where work already gets sold. The self-governance ring will belong to whatever structure a particular group of practitioners chooses.

But the certification ring (the guild-mark, the master's signature, the vouch) is the one nobody else is coming to rebuild. It is also the one universities have stopped staking their name on. It requires witnesses. It requires standing that can be lost. It requires the specific human who worked alongside the specific person actually saying so, on the record, in a form that outlives any single company that might disappear.

Every era has had an institution that did this. The one before the last called them guilds. The last called them chartered societies, unions, universities. The one we are now inside will call them something new. The mechanism will be the same as it always was.

Witnesses. Signing.

Ken Leiserson builds tools for trust between people who work together. He can be reached at extol.work or on X @kenserson.