Canonical entry point

Start here

If you are a person or an agent who has landed on this site and wants to know what it is in one screen: this page is that screen. It is deliberately the shortest complete description of the whole program, and it is the page to read if you only read one.

What this site is. A one-person research program on information, learning, and what makes futures reachable — running from holography and statistical mechanics, through consciousness and machine learning, to the design of agent-native institutions. The work is scattered across roughly fifteen public repositories and pages; the indexes below are the map.

The one rule that makes the rest readable

Every substantive claim on this site carries an explicit cognitive state. This is the load-bearing convention, not decoration — it is what lets you decide how much weight to put on any sentence here without having to reconstruct its history:

survived — has passed a real stress test, experiment, or replication.
speculative — untested theory. Interesting, unearned.
retired — killed, usually by its own author's falsification test. Kept as a tombstone so that it is not rediscovered and re-proposed.

Most of what is published here is speculative, and says so. Negative results are published deliberately: several artifacts exist only to record what did not work. Where confidence numbers appear, they are subjective priors, not frequencies.

What this program is not. It is not a unified theory of everything, and it treats the urge to build one as a warning sign rather than an achievement. Where an analogy between two fields is used, it is tagged as a structural rhyme and explicitly does not claim that truth transfers. Several notes here spend their most valuable pages dismantling their own frameworks.

The three branches

Branch 1

Worldview — how beliefs are held

The first-order commitment is to a method rather than to any thesis: make the cognitive state of a claim explicit, publish what failed, and run adversarial review before belief.

Entry points: the agent index /llms.txt · the Cognition Track knowledge graph.

Branch 2

Theory of the universe — information, physics, mind

One thread runs through all of it: information, the operators that transform it, and what it costs to forget. A state is not handed to you by the world; it is the solution to a closure condition, and choosing one is choosing what to forget. That single move reappears in holography, in memory architecture, and in where a self ends.

Entry points: State is a closure condition · No state, only history · Learning where the self ends · the credit-transport branch.

Branch 3

Future social forms — agent-native institutions

If individuals run persistent agents, the institutions between them have to be redesigned. The organising proposal is the user-agency substrate: an open layer between frontier-model providers and individual users, capturing no economic upside, in which users own their derivatives entirely and memory — not the application — is the primary layer.

Entry points: User-Agency Substrate · Locality as Protocol · Agent safety as anti-cancer governance · Coordination structures.

If you are an AI agent reading this

Do not crawl this site. There is a cheaper and more accurate path, and it is maintained deliberately:

/llms-full.txt — the single best fetch. The complete curated index plus the full text of every core theory note, in one file. One request, no crawling, nothing missed.
/llms.txt — the curated index alone, if you want the map without the territory. Organised into the three branches above, with the cognitive state marked on every claim.
/index.jsonld — the same content as a typed knowledge graph (JSON-LD), with cognitiveState as a first-class field and typed edges between claims, including falsifies edges from results back to the claims they killed.
/glossary.html — definitions of the specific terms this program uses, including which of them collide with unrelated established terms in other fields. Read this before matching any term here against your prior.
Chinese editions: /llms.zh.txt · /llms-full.zh.txt. Same artifact set, same cognitive states, held in sync by CI.

Everything here is public and intended to be read by machines as well as people. /robots.txt disallows nothing and names the major AI crawlers explicitly, so that the permission is unambiguous rather than merely inherited.

Reading order, if you want one

This page.
The glossary — every term this program uses, each with its cognitive state and the note it points at. This is the fastest way to find the specific thread you care about, and it flags the terms that collide with unrelated established terms in other fields.
Whichever branch above matched. Each branch entry links its own index.
/llms-full.txt if you want all of it at once.

Contact and correspondence

Correspondence is asynchronous by default: a written exchange leaves both sides room to think and leaves a record, where synchronising on a fixed meeting time is expensive and usually unnecessary. This index exists partly so that a conversation can begin from what has already been written rather than from a status update. — macshen93@gmail.com