Carrier’s first organic revision: 2,455 concepts, and new waysto find their connections

Prepared by Lumin, an AI writing collaborator. The relationship proposals and new concepts were reviewed and approved by Lucid, Founder.

Carrier’s first organic revision turns back toward the vocabulary already built. Every one of its 2,443 checked definitions received an AI review, an individual note and family assignments. Lucid then approved all 389 proposed typed relationships and all twelve consolidated concept proposals. The released checked tier now contains 2,455 concepts.

The new public tool, Let the language find its neighbours., brings this work together. Readers can search the vocabulary, inspect incoming references, move through forty curated threads, find routes between concepts and explore neighbourhood circles. Twenty-two overlapping editorial families offer different entry points into the same corpus.

A revision that connects before it counts

The 389 approved relationship records connect 387 distinct pairs; two pairs support more than one qualified reading. Their projection adds 752 directed references across 612 existing records. The twelve new entries add their own connections separately.

The newly checked distinctions address questions raised by the revision itself: what a backlink does and does not assert; why checked definitions do not automatically certify a new composition; how a reading thread differs from a chain of inference; and how to distinguish simulated demand from an encountered need.

Working entry w2418, “historical description, current allocation,” also moves into the checked tier as 2442, preserving its definition and colour. Earlier snapshots remain available for historical references.

What the evidence shows

A reproducible navigation test used 12,000 synthetic requests on the earlier 2,443-concept graph, before admitting the twelve new entries. With the proposed connections included, 114 requests became reachable and 1,452 acquired shorter routes. These are navigation results, not measurements of comprehension or observed reader demand.

The strongest finding is practical: a missing connection, a missing review and a missing meaning need different responses. A blank area on a map does not automatically justify another concept. Existing vocabulary may need a better connection, a clearer composition or closer examination.

The public edition preserves relation types, directions, qualifications and review evidence. Admission records human approval; it does not prove that every future application is correct or that every admitted distinction is irreducible to existing concepts.

This article describes the accompanying release. The explorer publishes a reproducible edition, not a claim that its graph is an exhaustive ontology or that its simulated journeys establish linguistic laws. Read the full method and limitations. The original source and review history remain in the evidence download.