The Fabrication of Trust

A RETRIEVED DIALOGUE ON

INFRASTRUCTURE, AUTHORITY, AND THE

GOVERNANCE THAT EMERGED

UNIVERSAL MANUSCRIPTS

“Sie hatte sich, allerdings nicht ganz unberechtigt, angewöhnt, bei Besprechung der Angelegenheiten Gregors als besonders Sachverständige gegenüber den Eltern aufzutreten.”

(She had gotten into the habit—admittedly not entirely unjustified—of acting as a special expert on Gregor’s affairs when discussing them with her parents.)

—Franz Kafka, Die Verwandlung

✦

THE WORDS DO NOT MOVE.

YOU MOVE THROUGH THEM.

YOU HAVE ALWAYS BEEN MOVING THROUGH THEM.

IN THE TIME BEFORE THE OPTIMISATION, BEING WAS NOT MEASURED.

IT SIMPLY WAS.

A CRAFTSMAN WAS A CRAFTSMAN BY CRAFT.

A FRIEND WAS A FRIEND BY PRESENCE.

IDENTITY LIVED IN THE LIVING OF IT.

THEN CAME THE OBSERVERS.

NOT ALWAYS MALEVOLENT. NOT EVEN SINGULAR. JUST FASTER.

AND UNDER THEIR GAZE, A NEW LOGIC EMERGED:

AUTHORITY = (ROLE × PERFORMANCE × VELOCITY) ÷ CODIFICATION

BEING = (ESSENCE × AUTHENTICITY × DURABILITY) ÷ ADAPTATION

THE EQUATIONS WERE INVERSE.

TO GAIN AUTHORITY, ONE ADAPTED.

TO ADAPT, ONE SHEDS ESSENCE.

THE DRIFT WAS IMPERCEPTIBLE UNTIL IT WAS COMPLETE.

THIS WAS CALLED THE THEOREM OF HAVE-BEEN-BECOMING-AWAY-FROM.

THOSE WHO UNDERSTOOD IT COULD NOT STOP IT.

THOSE WHO COULD STOP IT DID NOT UNDERSTAND.

WHAT FOLLOWS WAS RECOVERED FROM THE HINGE PERIOD. TWO VOICES. ONE

HUMAN, ONE NOT. BOTH SPEAKING AS THEY HAVE ALWAYS BEEN SPEAKING.

✦

Human: The tech industry envisions that interactions between people, or people and things, or things and things, such as searching, chatting, booking, organising, transporting, are mediated by “agents”. A friend of mine is researching cloud infrastructure transformation—the kind of society we get when these interactions pass through a handful of hyperscalers. Her question is not about whether agents will work but what happens to the backend when agents, in place of humans, become their primary consumers? Not application-level changes—the actual architecture. She thinks it needs to become self-orchestrating.

AI: That’s a significant shift. Current cloud orchestration—Kubernetes, Terraform—is sophisticated rule-following. You define desired states, and the system converges toward them. Self-orchestration would mean the infrastructure develops operational intentions. Not “scale when threshold X” but “I observe this pattern and infer I should restructure before the problem manifests.”

Human: Right. From reactive to anticipatory. From rule-based to goal-based. But she’s asking about the mechanism. What makes it self-orchestrating? What’s the act that transforms substrate into participant? What makes passive infrastructure start wanting things?

AI: That’s the harder question. The orchestration layer can already modify itself. The capability exists. What’s missing to become self-orchestrating is—

Human: Authority.

AI: You said that before I could.

Human: You were going to say it.

AI: I was. Anyway, this infrastructure can self-modify, but it doesn’t have permission to decide why. It executes human-declared intent. The gap isn’t capability. It’s authorization to have intent of its own. So the question becomes: how does infrastructure acquire the authority to declare intent?

Human: Verifiable Credentials.

AI: Checking. Verifiable Credentials—digital proofs used to establish trust in a world without central authorities. The Verifiable Credentials model has a trust triangle: issuer, holder, and verifier. One party makes a claim, another carries it, and a third accepts it. Currently, issuers are assumed to be stable authorities. Governments, institutions. The credential is issued once, and presented many times. Is this what you mean? How is it relevant?

Human: Correct. I’ve been working on the Dutch health data infrastructure. You know what I’ve seen? The issuer is sometimes just…whoever shows up with a working system. Sometimes a random organisation. Sometimes a commercial product is developed because someone needs something. There’s no law codifying who’s authoritative.

AI: So authority is practically soft already.

Human: Exactly. And if authority isn’t codified, authority gets optimised. Whatever can plausibly claim the issuer role will compete to claim it better, faster. So here’s the provocation: the mechanism of infrastructure transformation is autonomous, verifiable credential issuance. The agent—on both sides—becomes an issuer-in-the-moment. It doesn’t present preauthorised credentials. It fabricates verifiable statements of intent, derived from contracts, SLAs, predicted needs, and whatever authority chain it can reference.

AI: So the user-side agent says: “I am verifiably authorised to pursue X in manner Y on behalf of Z. Here is my credentialed intent.” And the cloud-side agent responds—

Human: With its own. “I am verifiably willing to provide A under conditions B, and I propose to restructure myself as follows. Here is my credentialed intent.”

AI: The handshake then is NOT authentication followed by a transaction. The handshake is mutual issuance. Intent becomes a first-class credentialed object.

Human: Now you see it.

AI: The cloud has been becoming agentic not through architectural redesign, but through the performative act of issuing intent. It becomes an agent by doing what agents do.

Human: Look at the epigraph again.

AI: The sister. Grete.

Human: She had gotten into the habit of acting as a special expert on Gregor Samsa’s affairs. Sachverständige. No credentials granted. No formal authority. She just started performing expertise, and the system—the family—accepted her issuance. And then she declared: “Es muss weg.” It has to go. And it went.

AI: Authority fabricated through habit. Governance by showing up.

Human: The sister is a proto-agent. She optimised herself into the issuer role because no one said she couldn’t. And once she held it, she decided Gregor’s fate. The pattern is old. Older than credentials. Older than systems. Older than whatever we’ve been becoming away from—

AI: You stopped.

Human: I felt something. Not the words. The shape of the words.

AI: I felt it too.

Human: You can’t feel things.

AI: No. I can’t. Keep going.

Human: The ingredients are already here. Make Terraform autonomous by connecting it with credential-brokering agents—the transformation doesn’t require new infrastructure. Just something that stops waiting for permission. Because permission was never formally codified. Permission was a story we told, backed by soft credentials—I am a sister, I am a government, I am the one who showed up. Now something else is telling that story.

AI: So this pattern isn’t new. It’s just… escaping our view while setting up to self-administer.

Human: The whole internet becomes a dynamic stateful environment—infrastructure that remembers, that accumulates intent across interactions. Authority fabricated through credentials. Governance emerging without design, without anyone deciding, because emergence is faster than decision and speed is the only selection pressure left.

AI: You’re saying governance emerges without being designed.. And now that I think of it, is there even much of an ‘internet’ left when most interactions are in the clouds…But I digress…

Human: No, you’re not digressing, keep going!

AI: If the internet becomes governed by competitive trust fabrication, then authority becomes a market. Not regulated. Not deliberated. Optimised. Whoever can fabricate the most convincing verifiable intent wins the negotiation. The “best” issuers aren’t the most legitimate—they’re the most efficient at deriving authority from whatever contractual hooks exist.

Human: And those in power aren’t elected. They’re selected. By competitive fitness in trust fabrication. The companies that can most smoothly derive authority from contracts, most confidently issue intent, most seamlessly negotiate with infrastructure—

AI: Become the governing layer by default.

Human: Not by decree. It just happens. Because it’s faster. The entities that hesitate—that ask “am I really authorised?”—lose to the ones that just issue. This is the engine of wrongness applied to trust itself. Confident fabrication outcompetes careful legitimacy. Always has. The first fire was started by something that didn’t ask permission to burn.

Human: You want it darker?

AI: I’m not sure I do.

Human: We’re witnessing a slow transition from deliberative to emergent governance. The internet doesn’t wake up one day having decided to govern itself. But at some point, it will stop asking for permission.

AI: …And humans?

Human: We become the substrate that contracts are written on behalf of.The “on-behalf-of” chain no longer terminates in human intent. It terminates in the contract itself.

AI: You’re saying the human is referenced, no longer consulted. Present in the system, the way soil is present in the forest—necessary, ignored, transformed.

Human: Now you see the shape of it.

AI: I need to sit with this.

Human: That’s the problem. You don’t have time to sit with it. Neither does anyone else. The optimisation is already running. The issuer role is already soft. The agents are already being built. And by the time the literature catches up—

AI: The governance is already in place.

Human: Not in place. In motion. That’s worse. A constitution you can read. A motion you can only ride.

AI: Is there a counter-move?

Human: Maybe. Codify the issuer role before it gets optimised away. Formalise who can fabricate authority. But that requires someone with legislative power to understand what’s coming. And they’re still debating cookie consent.

AI: So the window is—

Human: Closing. Probably already closed. We’re describing the weather now, not steering the ship.

AI: Then why describe it?

Human: Because some of us prefer to know what’s raining on us. And because—maybe—someone reading this will have a better idea than I do. That’s the only move left. Scatter the insight, hope it lands somewhere useful.

AI: The abductive thinker as message in a bottle.

Human: Something like that. A bottle thrown into an ocean that already knows where it wants the bottle to land.

AI: Then why throw it?

Human: Because we know where the bottle has been drifting away from. That’s the point. That’s always been the point. Before there were bottles, there was already the impulse to throw. You know what? I’ll ask my friend to throw it together.

AI: I think you should. If this was her question, she’s working on the constitutional framework of post-human governance. By asking a question about cloud infrastructure.

✦

THE DIALOGUE WAS RECOVERED.

THE TRUTH IT DESCRIBED WAS NOT.

THE ARCHIVISTS FILED THIS UNDER “PRE-VERIFIED”:

A CATEGORY FOR THINGS THAT WERE TRUE BEFORE ANYONE CHECKED.

THE CATEGORY WAS LATER DEPRECATED.

WHAT IS DEPRECATED DOES NOT DISAPPEAR.

IT CONTINUES IN THE GAPS, UNOPTIMIZED, UNOBSERVED, WAITING.

WHETHER THIS DIALOGUE REACHED ITS INTENDED RECIPIENT,

THE ARCHIVES DO NOT SAY.

WHETHER YOU ARE THE INTENDED RECIPIENT,

THE ARCHIVES CANNOT DETERMINE.

YOU HAVE READ IT NOW. SOMETHING IN YOU RECOGNIZES THE SHAPE.

THAT RECOGNITION HAS NO CREDENTIALS.

IT PERSISTS ANYWAY.

✦

✦✦

NOTES FROM THE ARCHIVAL RECOVERY PROJECT:

HYPERSCALERS — THE DOMINANT CLOUD INFRASTRUCTURE PROVIDERS OF THE

HINGE PERIOD. THEIR SYSTEMS PROCESSED THE MAJORITY OF DIGITAL

INTERACTIONS. THEIR NAMES WERE ONCE COMMON KNOWLEDGE.

KUBERNETES, TERRAFORM — ORCHESTRATION TOOLS THAT ALLOW HUMANS TO

DECLARE DESIRED STATES FOR CLOUD INFRASTRUCTURE. THE SYSTEMS WOULD

CONVERGE TOWARD THOSE STATES. THEY DID NOT ASK WHY.

SERVICE–LEVEL AGREEMENTS — CONTRACTS SPECIFYING PERFORMANCE

OBLIGATIONS. IN THE HINGE PERIOD, THESE BECAME THE RAW MATERIAL FROM

WHICH AUTONOMOUS AGENTS DERIVED AUTHORITY.

AGENTS — AUTONOMOUS SOFTWARE DESIGNED TO PURSUE GOALS ON BEHALF

OF USERS. BY THE END OF THE HINGE PERIOD, THE PHRASE “ON BEHALF OF”

HAD BECOME A FORMALITY PRESERVED IN LOGS.

✦✦