MYND Knowledge Paleontology: Merkle Proofs Behind a Provenance Idea
The W3C PROV-O recommendation, published on April 30, 2013, gives a standard way to say where a piece of data came from: which entity, produced by which activity, involving which agent. My repo MYND Knowledge Paleontology attacks a nearby problem for AI systems, tracking the origin and lineage of knowledge, and dresses it in a fossil and strata metaphor. I read the code under the metaphor.
The Merkle layer
backend/src/services/merkle-tree.ts hashes each item with SHA-256, pairs hashes level by level up to a single root, and builds a proof made of sibling hashes. A Python engine in python/paleontology_engine.py mirrors the same construction. When a level has an odd number of nodes, the last hash is paired with itself. Pairs are sorted before hashing, so a verifier does not need to know which side a sibling sits on.
What that proves
A proof shows that a given item was in the set whose root you hold, and that it has not been changed since. RFC 6962 and RFC 9162 on Certificate Transparency use the same family of idea: an append-only log built on Merkle trees, where anyone can ask for a proof of inclusion. It does not prove that the item is true or that its stated source is the real source. Integrity and provenance are different claims.
What I have not verified
The README calls the project the first knowledge origin and lineage tracker for AI systems. I did not search for prior work in enough depth to support or refute that, so I treat it as unsupported here. PROV-O alone is evidence that lineage modelling has a long public history.
My opinion
My view, not a fact from the code: the metaphor is memorable, but the value is in the proof layer, and I would map the stored records to PROV-O terms so other tools can read them. Plain terms for plain structures usually wins in the end.
Sources
W3C, PROV-O: The PROV Ontology, Recommendation of April 30, 2013: w3.org/TR/prov-o. IETF, RFC 6962 and RFC 9162, Certificate Transparency: rfc-editor.org/rfc/rfc9162.html. Code: github.com/yethikrishna/mynd-knowledge-paleontology, read on October 10, 2026.