A public graph of non-consensus cognitive claims about intelligence, learning,
and information dynamics. Both humans and agents can traverse it: every node is an
interlinked markdown file with a stable id, and manifest.jsonld
is a machine-readable index of all nodes and edges.
Start at INDEX.md or jump to the root node
root.md.
- It is a protocol-layer artifact, not a hub. The format is meant to be forked and linked into other people's graphs, not to route everyone through a central site. There is no owner-as-gatekeeper. Take the manifest schema and run your own.
- Cognitive state is first-class. "Publish everything" does not mean
flattening untested guesses and stress-tested claims onto one plane. Every node
carries a state badge:
- 🟢 survived-stress-test — has cited external results / passed a falsification
- 🟡 speculative — a reasoned but untested hypothesis
- ⚪ raw — a fresh, unprocessed idea
As of this build: 4 survived · 16 speculative · 0 raw (of 20).
This web is mostly untested theory, and it says so. One node is included
because it was falsified (see the Bekenstein node's
falsifiesedge) — keeping failures visible is part of the protocol.
projection-of · extends · depends-on · falsifies
- Take
manifest.jsonldas the interchange format. - Reference any node by its stable
mem_*id from your own graph. - Add your own nodes/edges; no coordination with this repo required.
The per-claim companion format — front-matter / JSON you can attach to an essay or document outside this graph, sharing the same cognitive-state and edge vocabulary — is the Claim-Receipt spec. This graph is its live, dogfooded demonstration.
- Cognitive content (node claims, INDEX, this README's substance) — CC BY-SA 4.0
- Manifest format / schema spec (
manifest.jsonldstructure) — Apache 2.0
(Licensing stack inherited from the Starshard publish doctrine.)
Generated from a Memory Hub staging extraction. Public since 2026-06-03. No social/outbound promotion is attached.