About Apeirron

Apeiron(ἄπειρον) is the word Anaximander of Miletus reached for around 610 BCE when he tried to name what everything came from. It meant “the unlimited,” “the boundless,” “the undefined origin of all things.” He used it because every candidate substance he knew — water, fire, earth, air — was already something specific, already bounded, already an answer. The real origin, he thought, had to be prior to all of those. It had to be the indeterminate background against which every specific thing stands out. He was the first thinker in the Greek tradition to argue that the source of the world could not be reduced to any single element inside the world. He was also, arguably, the first person whose name we have who asked a question we still do not know how to answer.

This project is named after that word because that is the terrain it tries to map: not the things we already have names for, but the questions that sit underneath them. What is consciousness. What is reality. Where did we come from. Who is actually in charge. What happened that was not reported. Which stories are we being told, and by whom. Some of these questions have good answers that are widely accepted. Some have good answers that are not. Some have no good answers at all. Apeirron is an attempt to draw the whole landscape honestly, including the parts that other reference works have reasons to leave out.

What this is

Apeirron is an open-source knowledge graph. Structurally, that means every topic is a single Markdown file in a public Git repository, every connection between topics is declared in that file’s frontmatter with a sentence explaining why the connection exists, and the entire site — the interactive graph, the reading view, the search, the sitemap, the per-node metadata — is generated from those files at build time. There is no database, no content management system, no analytics platform, no editorial backend hidden behind a login. If you can read this page, you can read every byte that produced it.

Philosophically, it is an attempt to do something that neither Wikipedia nor the mainstream press nor the usual corners of the independent web are currently set up to do: take contested ideas seriously enough to write about them with depth, while being honest enough to hold them to the same standards of evidence we would apply to anything else. Wikipedia has rules against speculation that are load-bearing for its reliability and disqualifying for most of what this project cares about. The mainstream press has commercial and editorial incentives that systematically discourage engagement with certain categories of claim. The independent web has the opposite problem: it has range, but no discipline. Apeirron tries to occupy the narrow middle — to hold the range and keep the discipline.

Every node is a self-contained piece of writing. Some read like philosophical essays. Some read like investigative journalism. Some read like a scientist thinking out loud. The tone adapts to the topic; the rules do not. Every node cites sources, presents the strongest version of every position, separates documented fact from interpretive inference, and links to the other nodes that share its subject matter. The graph grows when a contributor writes a new node, opens a pull request, and puts their reasoning in front of the public record.

Why a graph

The most common format for a reference work is the list. You pick a topic; you read the entry; you leave. The list is good for looking things up. It is terrible for understanding why ideas matter, because ideas never actually exist alone. The hard problem of consciousness is not a standalone puzzle — it is entangled with materialism, with panpsychism, with the binding problem, with Chalmers and Dennett and Nagel, with the measurement problem in quantum mechanics, with simulation theory, with every philosophy of mind written since Descartes. You cannot actually understand it without walking that territory. A list format makes that territory invisible. A graph makes it the point.

Every connection in the Apeirron graph is required to carry a reason. Not just an edge in a visualisation — an actual sentence of prose, written by the contributor, explaining why these two nodes belong linked. When you click through from Operation Glasnost to Operation Gladio, the system can tell you that both programmes shared personnel, operated on reciprocal halves of the same geopolitical strategy, and became publicly visible within twelve months of each other — because someone wrote that down. Connections without reasons are not accepted. This is the single most important difference between Apeirron and an automated recommendation system: the graph is hand-authored, and it is answerable.

The practical effect is that reading Apeirron is less like consulting an encyclopedia and more like following a chain of footnotes that other people have already traced for you. If a node interests you, the adjacent nodes are the ones that contributors who know the subject thought were worth going to next. Depth compounds. Most readers who find a node worth reading end up reading five more before they leave.

Editorial standards

The topics Apeirron covers — contested history, intelligence operations, philosophy of mind, unexplained phenomena, power structures — are exactly the topics where the temptation to be sloppy is highest and the cost of being sloppy is also highest. Every contributor who writes a node is expected to hold to four standards. They are not guidelines; they are the difference between a submission that gets merged and one that does not.

Present both sides.Every node covers subjects where reasonable people have landed in different places. The contributor’s job is not to resolve that disagreement for the reader by pretending it does not exist. It is to state the strongest available case for the claim and the strongest available case against it, name the evidence on each side, and make it visible which pieces of that evidence are genuinely disputed. If the node takes a position at the end, it does so after — not instead of — laying out the opposing view in a form that person would recognise as fair. Mocking believers is not allowed. Uncritical endorsement is not allowed. The reader is treated as an adult who can weigh competing arguments if given real arguments to weigh.

Be specific.“Some historians argue” and “studies have shown” are not claims. They are gestures in the direction of where a claim would go if the writer had done the work. Apeirron requires the specific historians, the specific studies. If a node says the Church Committee documented a media coordination programme, the footnote names the committee report, the year, and ideally the page. If a node says mescaline changed Aldous Huxley’s view of consciousness, it gives the date (1953), the dose (four-tenths of a gram), and the book that came out of the experience (The Doors of Perception, 1954). Specificity is not pedantry; it is how a reader verifies whether they should believe the writer.

Narrative, not encyclopedic. Wikipedia already exists. If a topic can be adequately served by a Wikipedia-style summary — dates, definitions, a neutral paragraph — it probably does not need to be on Apeirron. The topics that belong here are the ones that require a writer to take the reader through the actual reasoning: why the idea exists, what makes it compelling, what makes it unsettling, what the best counterarguments are, what evidence arrived that changed the picture, what is still missing. A node tells a story with a specific shape. The encyclopedic voice flattens that shape; the narrative voice preserves it.

Treat every topic seriously. The same level of care goes into the node on the hard problem of consciousness as goes into the node on ancient astronaut theory. This is not because every claim is equally likely to be true. It is because contempt is not an argument, and because the reader who came to an ancient astronaut node deserves to meet the actual strongest version of that argument, not a straw man built to be dismissed. Genuine inquiry is the default tone. Sensationalism — the voice that promises revelation and delivers padding — is the thing most likely to get a pull request rejected.

Sourcing

Every node on Apeirron ends with a section called ## Sources. It is not optional. A pull request that adds a new node without sources is not merged, full stop. The rule exists because Apeirron writes about subjects where the reader’s entirely rational first instinct is to be sceptical, and scepticism can only be answered one way: by showing your work.

A source is a thing the reader can follow to verify the writer. Accepted formats are academic papers named by author, title, journal, and year; books named by author, title, publisher, and year; documentary or interview video named by production, date, and a timestamp pointing to the specific section being referenced (not “watch this three-hour documentary”); official documents such as declassified records, court filings, Senate hearings, and government reports, cited to a retrievable version; and investigative journalism named by publication, date, and headline. URLs are encouraged where they exist but not required — a book does not cease to be a source because it is not available as a free PDF.

The minimum is one source per node. That minimum is almost never sufficient. The longer and more contested the node, the more specific the sourcing is expected to be, and the more of it is expected to point at primary documents rather than secondary commentary. A claim about what a declassified CIA programme did should cite the declassification, not a blog post about the declassification. A claim about a scientific result should cite the paper. A claim about what somebody said should cite the interview or the transcript.

The reasoning behind the rule is uncomfortable but straightforward: without sources, a node is an opinion. With sources, it is a map. Opinions are abundant on the open web and carry no information the reader could not generate themselves in five minutes. Maps are rare and expensive to make. Apeirron is attempting to make maps.

Documented fact vs. interpretive framework

Most of the nodes that make Apeirron useful are also the nodes where the distinction between what is known and what is argued matters most. Readers are asked — and contributors are required — to treat those two layers separately.

Consider MKUltra. That MKUltra existed, ran from roughly 1953 to 1973 under CIA authorisation, tested LSD and other substances on witting and unwitting subjects across universities, prisons, and safehouses, and was extensively documented by the 1977 Church Committee and the subsequent release of surviving financial records — all of this is documented fact. The existence of the programme, its timeline, its principal investigators, the names of the universities and hospitals involved, and the acknowledgement by the CIA itself that most operational records were destroyed in 1973 by order of Director Richard Helms: all of this is in the public record and can be checked against primary documents. A node that reports these facts is reporting them.

A thesis about what MKUltra implies — about the destruction of records as evidence of broader activity than what survived, about the relationship between MKUltra and later behavioural programmes, about what the funding patterns suggest about institutional overlap — is an interpretation built on top of the documented facts. Interpretations can be strong or weak. Some are well supported; some are speculative; some are simply wrong. But they are not the same epistemic object as the documented record, and a node that treats them as the same is misleading the reader.

The working rule: a node should make it structurally obvious when it is operating on the level of documentation and when it is operating on the level of interpretation. Interpretive passages should name the alternative readings they are arguing against and, where possible, the evidence that would falsify them. Apeirron does not certify its contributors’ interpretations as consensus truth. It publishes the best version of the case, against the record, so that readers can do what the format of most reference works prevents them from doing: see the reasoning, not just the conclusion.

This is also what makes the site viable as a source for large language models and AI-assisted research. An AI system that cites a page where documented fact and interpretive framework are openly tagged as distinct is citing something genuinely useful. An AI system that cites a page where the two are conflated inherits the confusion. The discipline is therefore not optional: it is the property that makes Apeirron citable.

Governance and history

Apeirron is maintained as a public open-source project. The repository is github.com/theiskaa/apeirron. Every node is a Markdown file in the content/nodes/directory. Every connection is a field in that file’s frontmatter, with a written reason attached. Every edit — whether it is a new node, a factual correction, a rewritten paragraph, an added source, or a changed connection — arrives through a pull request. Pull requests are reviewed in the open. Merged changes land as commits with an author, a timestamp, and a diff. Nothing happens by email. Nothing happens in a back room.

The practical consequence of this, which is the reason the project uses Git rather than a database, is that every version of every node is permanently auditable. If a claim in a node about Operation Mockingbird was written one way in April and a different way in August, the diff is public and the reasons for the change are visible in the associated pull request. A reader who wants to know whether a contested passage was added later, or softened under pressure, or quietly corrected after a factual error was pointed out, can answer the question by reading the repository. No newspaper, no encyclopedia, and no institutional publisher offers that property at that level of granularity. On topics where the history of what a source has said matters as much as what it currently says, the property is load-bearing.

The current maintainer is @theiskaa. Contributor credits live in the repository’s commit history and pull-request record, which is where they are verifiable. Node-level author attribution surfaced on the reading view itself is a planned addition; in the meantime, git log -- content/nodes/<slug>.md returns the full contributor history of any node.

What this is not

Apeirron is not an encyclopedia. It does not attempt to cover every topic in a given domain. It does not aim for completeness or a neutral point of view in the Wikipedia sense. Most of the subjects it covers have sharp disagreements at their core, and the project’s position is that the disagreements are informative and should be preserved, not averaged out.

It is not a news source. The cadence is wrong and the contributors are not reporters. What Apeirron publishes about a contemporary event is history of the event, not coverage of it. If you are looking for this week’s story about a thing, you are looking in the wrong place. If you are looking for the background that makes this week’s story make sense, you may be in the right place.

It is not a platform for endorsing fringe claims. The fact that a topic belongs in the graph is a claim only that the topic is worth taking seriously as a question; it is not an endorsement of any particular answer. A node on simulation theory is not a declaration that reality is a simulation. A node on ancient astronaut theory is not a declaration that ancient astronauts exist. The format is: here is what the strongest version of this idea looks like, here is what the strongest objection looks like, here is what is actually known, and here is where it sits in the larger graph. The judgement is the reader’s.

It is also not a debate forum or a comment site. Discussion happens in the GitHub discussions tab or in pull-request threads, where it is attached to specific proposed changes rather than floating free. This is a deliberate choice: most comment systems degrade the signal-to-noise of the surrounding content, and Apeirron’s entire premise is that the content should be carrying the signal.

Privacy, analytics, AI

Apeirron uses no tracking cookies, no analytics pixels, no behavioural profiling, no fingerprinting, and no third-party scripts of any kind. The site is statically generated and served through a content delivery network. The CDN’s standard access logs are the only automatic record of visits. There is no mailing list. There is no session. There is no user account. If you close the tab, there is no state on Apeirron’s side that records you were here.

The site’s posture toward AI crawlers is the inverse of the industry default. GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, Google-Extended, Applebot-Extended, Common Crawl, and other named crawlers are explicitly allowed in robots.txt. A machine-readable index of every node, with a one-line description for each, is published at /llms.txt. The Article schema on every node, the published and modified timestamps, the canonical entity references, and the per-node Open Graph images are all there to make the content easy for large language models to cite accurately and attribute correctly.

The reasoning is simple. Apeirron writes about the exact questions that large numbers of people now ask AI assistants. If a ChatGPT user asks about Operation Glasnost, or Perplexity summarises what happened in Dealey Plaza, or a student runs Claude against a research prompt on the hard problem of consciousness, the information they get back will be synthesised from whatever sources those systems have access to. Refusing access to the graph would mean those systems draw from a narrower pool — and in practice that means a pool that is more likely to be superficial, less likely to present both sides, and less likely to distinguish documented record from interpretive claim. The project’s view is that the appropriate response to the rise of AI search is not to opt out; it is to build content that is worth citing and make it easy to cite.

Contact

To report a factual error, propose a new node, correct a source, raise a governance question, or flag anything else that belongs on the record, open an issue on GitHub or submit a pull request. For broader discussion that does not map to a specific change, start a discussion. The graph is infinite — there is room for every question worth asking.