Provenance, Not Enclosure
March 3, 2026
Your Get Out of Jail Free Card
My wife and I talked this morning about how she'd protect her work. Not the files--the thinking. The analysis, the synthesis, the judgment calls that take years of domain expertise to develop. "What stops someone from just taking it?"
I said something glib about Pythagoras. We still call it his theorem 2,500 years later. Attribution persists. Create good work, and the provenance follows.
She wasn't convinced. And she was right not to be--because I was wrong.
A clay tablet the size of a phone sits in Columbia University's rare book collection. Plimpton 322, dated to roughly 1800 BCE, contains systematic Pythagorean triples--including numbers like 119, 120, and 169--generated by a method so sophisticated it couldn't have been stumbled on by trial and error. That's 1,300 years before Pythagoras was born. The Indian Baudhayana Sulba Sutra stated the same relationship around 800 BCE: "The rope which is stretched across the diagonal of a square produces an area double the size of the original square."
Pythagoras left no writings. Neither Plato nor Aristotle credited him with any mathematical discovery. The Stanford Encyclopedia of Philosophy puts it bluntly: "there is not a jot of evidence for a proof by Pythagoras." The attribution stuck because he ran a secretive community that assigned all members' work to the master, and later hagiographers amplified the myth for centuries. Word-of-mouth provenance, running on cult loyalty and institutional prestige, got it wrong for two and a half millennia.
So my glib answer proved the opposite of what I intended. Attribution doesn't just persist. It persists wrong--when the provenance infrastructure is bad.
The real answer to my wife's question isn't copyright. Copyright Is Dead. Copyleft Is Too. already showed why: you can't enclose copies anymore, and the legal frameworks built to try are collapsing. But that was never an argument that creators shouldn't reap the rewards of their thought work. It was an argument that enclosure--controlling copies--was the wrong mechanism.
What creators actually need is provenance. Not fences around knowledge. Receipts for who made it.
Landlords and Jailors
The problem isn't platforms. I use Gmail. I publish on Substack. I share links on social media. These are useful services, and I'm not here to tell you to leave them.
The problem is when we can't tell our landlords from our jailors.
If your email address is @gmail.com, Google is your identity. Leave Google, lose the address, lose every account tied to it. If your blog lives at yourname.substack.com, Substack is your URL. Leave, and the address goes dark.
But if your email is @yourdomain.com and Gmail just handles the mail--you can change an MX record and move to a different provider in an afternoon. If your site is at yourdomain.ca and the VPS is just where it's hosted--you can point the DNS elsewhere and everything follows.
The distinction isn't renting versus owning. Everyone rents something. The server this post is served from is someone else's metal in someone else's datacenter. The distinction is what you rent. Rent the compute, rent the hosting, rent the inference. But own the name. Own the keys. Own the content. Own the layer that makes you you.
This series has traced the same pattern through every enclosure. Priests controlled access to texts--"we'll hold the knowledge for you." Publishers controlled copies--"we'll distribute for you." Platforms control data and identity--"we'll host you." Each one offered a real service. Each one made leaving expensive enough that the service became capture.
Convenience becomes dependency becomes enclosure.
Here's what it looks like to use the service without accepting the capture.
What I Actually Built
You're reading this at waypoint.henrynet.ca--or, if you followed the link from Substack, you're welcome to. Five posts about sovereign infrastructure, published on someone else's platform. The irony was always there, I felt it keenly as soon as these thoughts started crystallizing.
My waypoint site is a static site built with Claude Code. The entire build process is a 257-line Python script: it reads markdown files, converts them to HTML, and generates an RSS feed. No framework. No CMS. No database. The content lives as plain text in a git repository--the same repository I write in.
My cryptographic identity lives at did:web:james.henrynet.ca--a W3C Decentralized Identifier anchored to my domain, signed with an Ed25519 key. When a machine--or eventually an agent--needs to verify that I am who I say I am, it doesn't ask a platform. It checks my published key.
Substack still exists. It's syndication now, not source. Gmail still handles my email, but I own my MX records (the DNS required telling mail where to deliver). Gmail is my mail handler, not my address.
What did this cost? Time--less than you'd think, because Claude Code built the code to do it. A rented VPS. And the tradeoffs: I gave up Substack's discovery mechanism, their email infrastructure, their payment processing, their network effects. Those are real costs. I'm naming them.
But nothing was lost. Substack still gets a copy of the posts. The audience doesn't have to move. The source moved. The distribution didn't.
And yes--this site was built with Claude, on Anthropic's inference. The VPS is rented cloud compute. I used a rented tool to build infrastructure on a rented server to write about sovereignty. That's not hypocrisy. That's the point. I moved from a jailor to a landlord. The next move is owning the building. But this move matters.
When to Rent
A few years ago my wife and I took a guided paddle trip down the Grand River in Ontario. Our guide knew the river, the local flora and fauna, every eddy and landing. At the end of the trip he called a taxi company to shuttle us and the boats back to the put-in.
I asked him why. Every other outfitter on that stretch of river owns vans and trailers. His answer stuck with me: "I'm in the business of moving boats by river, not by land. If I'm managing a fleet of vehicles, that's effort I'm not putting into my business."
There's a cost to ownership that doesn't show up on the balance sheet. That question comes up constantly now: should I build this myself or rent it? Local inference will change the math--I'm sure of that--but right now it's cheaper for me to subscribe to Anthropic for inference than to build local capacity. That's a renting decision I'm making with my eyes open.
The point isn't to own everything. It's to know what you're renting, and to make sure the things that define you--your name, your keys, your work--aren't someone else's to revoke. Rent the shuttle. Own the river.
Receipts, Not Fences
The Pythagoras story isn't just a footnote. It's the connection between sovereignty and the economics of knowledge work.
The Credibility Commons argued that credibility accrues through verifiable track records. But track records need infrastructure. Without it, you get Pythagoras--credit assigned by prestige and proximity, wrong for millennia, because nobody could check. The Babylonians had clay tablets. The failure wasn't the record--it was that checking it required physical access, specialist knowledge, and institutional authority. A DID document is a clay tablet any machine can verify in milliseconds, without asking anyone's permission.
The DID document, the signed content, the timestamped git history--these aren't just protection against deplatforming. They're provenance infrastructure. They answer the question my wife asked: "What stops someone from taking credit for my work?" Not a copyright notice. A cryptographic signature, a verifiable timestamp, a published record on infrastructure you control.
Copyright tried to protect credit by controlling copies. It's the wrong tool--in a digital world you can't control copies anymore, and the attempt produces the enclosures this series has been tracing. Provenance protects credit by making the record available and verifiable. The difference is the direction: enclosure says "you can't use this without my permission." Provenance says "use it freely--and the record shows where it came from."
Academia already knows this. The currency isn't the paper--it's the citation. Your h-index, your Erdos number, the web of attribution that connects your work to everyone else's. The value isn't in locking the paper away. It's in being the person others cite. Credibility comes from the verifiable record that you did the work, and that the work mattered enough for others to build on it.
The Map
None of what I built is novel. Every component was made by other people, under open standards, and made available for anyone to use. Telling you who built them is the model working.
Your domain. DNS (Paul Mockapetris, 1983, IETF open standard) is the oldest decentralized identity system we have. Your domain is your namespace. It follows you between hosts, between providers, between platforms. This is the single most important thing you can own. Everything else builds on it.
Your content as portable files. Markdown (John Gruber, 2004, BSD license) in a git repository (Linus Torvalds, 2005, GPLv2). Plain text, readable by any tool, portable to any host. No proprietary format. If your host disappears tomorrow, you clone the repo somewhere else. The content was never trapped--it was deployed, not captured.
Your cryptographic identity. Decentralized Identifiers (W3C Recommendation, July 2022, permissive license). A JSON file at a well-known path on your domain. Signed with Ed25519 (Daniel J. Bernstein et al., 2011, public domain). This is how machines verify that you are you--not because a platform vouches for you, but because you hold the key. Open standards, not products.
Your feed. RSS (Dave Winer, 2002, Creative Commons). The original decentralized content distribution. No algorithm decides whether your readers see your work. Subscribe means subscribe. Yes, browsers killed native RSS support--Google Reader shut down in 2013, Firefox dropped its reader in 2018. But RSS never died. It's still the backbone of podcast distribution, and dedicated readers like Feedly and NetNewsWire serve millions. The infrastructure is there. The browsers just stopped showing it to you.
Your syndication strategy. You don't leave platforms. You use them. Post to Substack, cross-post to Medium, share on social. But the source is yours. Every platform is a distribution channel, not a home.
This entire stack is assembled from open work, built by people who made it available. I didn't invent any of it. I used it, and I'm telling you who made it. That's provenance without enclosure.
How to get There from Here
This is enthusiast-level today. I know that. Most people aren't going to register a domain, set up a DID, and run a static site generator. The convenience gap from the last post is real, and I'd be dishonest to pretend otherwise.
Someone will offer to "host your sovereign identity for you." Most people will accept, and in ten years that host will be the new jailor. That's the centralization tendency--it's physics, not a moral failure. Discovery without platform distribution is a genuine cost; RSS doesn't have an algorithm, and that's both the feature and the limitation. My post distribution still lives on Substack. I haven't solved that one yet, and I'm naming it.
But every component above is portable. Domains transfer. Git repos clone. DID documents are a JSON file. Nothing locks you in. You can start with just the domain and layer the rest on over time.
And the tools to build this are closer than you think. If you have access to Claude Code or a similar agentic coding tool, the specification is the skill. Here's the prompt that gets you started:
"Generate an Ed25519 keypair. Create a did:web DID document for my domain [yourdomain.com] with the public key, authentication, and assertionMethod. Output the JSON for .well-known/did.json and tell me where to put it."
That's it. One sentence, and the agent builds your cryptographic identity. The 257 lines of Python that build this site started the same way--a specification, not code. The gap between "enthusiast-level" and "accessible" is the same gap that inference is closing everywhere else.
The gap between "the platform captures me" and "I can leave in an afternoon" is the whole game. You don't have to close it all at once. You just have to start closing it.
The Direction
This series traced enclosures: sacred, legal, infrastructure. It traced what replaces them: credibility, sovereign compute. Each one offered a real service that became a real prison.
The answer was never "stop using services." It was: know the difference between a landlord and a jailor. Own the layers that matter. Rent the rest. Attribute freely, because provenance is what replaced enclosure and it only works if you practice it.
Sovereignty isn't a destination. It's a direction. And the first step is owning your name.
This is part of an ongoing series exploring what happens when knowledge stops being property. Previously: Knowledge Was Always Free. Copyright Is Dead. Copyleft Is Too. The Third Enclosure. The Credibility Commons. Inference Comes Home.
James Henry is a senior engineer who wrote five posts about sovereign infrastructure on someone else's platform, then moved. He works with LLMs liberally--including in the writing of this post--because the collaboration is the point. You're reading this at its source.