Expression Library · Collection 02 · Findings & method
Fifty Books, 693 Concepts: How a Language Gains Depth
A growing language needs ways to keep important differences intact: between an observation and a generalization, assistance and agency, a convincing diagram and a proof.
The second Expression Library collection has brought fifty more books into USL’s programme of conceptual enrichment. Following Keeper review and promotion, Carrier has grown from 1,477 to 2,170 checked concepts. The public atlas opens the underlying work to readers: definitions, passages, proposed readings, comparisons, alternative emphases, and recorded limits.
The numbers are useful for planning. The deeper result is a vocabulary that can say more precisely what an author is asking a recipient to understand.
A future expression could identify a causal comparison without pretending that the groups were randomized. It could describe help that leaves someone unable to challenge the helper. It could specify a mathematical condition without silently discarding its scope. These are distinctions an AI can select and compose with; the expression still has to explain how they apply.
Inquiry begins with what a record cannot establish
The selected reading of John Snow’s On the Mode of Communication of Cholera contributes a family of distinctions about evidence. An exposure that cannot be established remains unknown; it is not automatically a supported absence of exposure. Where an outcome is recorded may differ from where the relevant contact happened. People move, obtain supplies from elsewhere, and enter institutions.
Another concept retains the difficulty of an intervention introduced during an existing decline. If the outcome was already diminishing, later improvement cannot simply be assigned in full to the intervention. That does not decide the intervention’s value. It identifies what a causal argument still needs to explain.
The retained historical account also separates an exposure route from a proposed mechanism inside the body. Evidence for one need not establish every part of the other. This is a reading of the source’s reasoning, not present-day medical guidance.
Harriet Martineau’s How to Observe: Morals and Manners brings a related problem into social inquiry. Introductions, hosts, class connections, transport, and visible status help determine whom an observer meets. Agreement among those contacts can therefore reflect the route of access.
Her selected passages also distinguish immediate field notes from later reflection, and breadth of contact from reliability of testimony. The atlas retains the pressure her own stereotypes and hierarchies place on that method. A source can contribute a useful discipline while also exposing reasons to apply that discipline to the source itself.
For USL, these concepts help represent the path by which information became available. That path can matter as much as the statement eventually transmitted.
Care, freedom, and the right to remain a participant
The Yellow Wallpaper introduces a particularly sharp distinction: intimate and professional authority combined. When the same person holds both roles, challenging a practical judgment can also be treated as a failure of loyalty. Sincerely professed care does not make the difficulty disappear.
The reading also distinguishes material help from retained discretion, and outward quiet from evidence of recovery. These are abstractions drawn from a fictional voice. The story does not become a clinical trial, and its narrator’s situated account does not become an automatically complete record.
Du Bois’s The Souls of Black Folk adds a different separation: material charity without recognition as a fellow participant. Someone may receive assistance while being refused equal encounter or cooperation. The same reading distinguishes everyday physical proximity from access to another group’s intellectual and cultural life.
These connections deepen the project’s commitment to Love & Freedom. They give it questions to ask: Can the person object? Can they participate as a peer? Does an apparent agreement reflect available alternatives? The historical sources are not being assigned the project’s present principles; they are helping the vocabulary expose situations in which those principles become demanding.
Formal precision has an ordinary use
The mathematical selections contributed more than technical names. They made assumptions and roles harder to hide.
In the Euclid reading, producing a figure remains distinct from proving that it has the required properties. Equal area does not establish congruence. A ratio preserves the order of its terms. A conditional statement and its converse have separate proof obligations.
In Russell’s Introduction to Mathematical Philosophy, denying that a uniquely described individual has a property differs from asserting that such an individual exists and lacks it. The placement of negation changes the commitment.
That difference matters well beyond a formal exercise. A compact message can become misleading if a receiver reconstructs the right labels but the wrong existence claim, condition, or direction of implication. A richer dictionary supplies useful distinctions; a sound expression must still assign their roles correctly.
Source form matters too. The batch retained original HTML assets and a PDF where plain text was insufficient. Selected Euclid pages were inspected visually, and a reversed ratio in an editorial numerical example was recorded. The retained source was not silently repaired. Mathematical image content also remains a limit of derived reading text.
Two shared definitions, four distinct source arguments
A language gains little from allocating a fresh identifier every time a familiar distinction appears under another title. But it also loses something when broad similarity erases the context that gives an argument its force.
The overlap review consolidated two pairs. The Tao and Dhammapada readings share the distinction between mastery of oneself and overcoming another person. Their evidence and surrounding readings remain separate. Sharing that definition does not make the traditions interchangeable.
Ricardo and Henry George similarly share a definition for differential return above a production margin. George’s wider argument about natural opportunities, social arrangements, and reform remains part of his particular reading. A common conceptual component does not establish agreement between the full economic arguments.
These are two shared-definition applications within the new batch. They are accounted for separately from the 144 proposed applications of definitions already checked before this batch. Neither category adds a second concept merely because another book uses it.
This is one practical meaning of expression-first enrichment: decide what a reading needs to communicate, then add a definition only where existing vocabulary cannot adequately carry the distinction.
Why these fifty—and what the rating means
The books occupy ranks 11–60 in the existing 1,000-item Gutenberg queue. They span twenty domains, broadening Carrier’s earlier concentration on scripture, interpretation, peace proposals, and institutional authority.
The original queue has fifty selections in each domain, with 134 editorial nominations and 866 provisional metadata selections. All first hundred entries, including this collection, come from the editorial shortlist. Their order interleaves domains and applies repetition controls.
The rating is a scheduling aid. Metadata signals include domain terms, catalogue classes, and title matches; editorial nominations receive a larger priority weight. The original procedure also limits the first catalogue attribution to three records and uses a provisional title-and-attribution check for related editions and volumes. These controls do not establish complete author, translator, or cultural diversity.
Rank is not a score for literary greatness, and it was not calculated from a full reading of every book. It records an editorial reason to investigate a work for expressive potential. The original planning record preserves that selection context, including its historical intake status.
The English-readable starting point also inherits the archive’s publishing, translation, and historical biases. Fifty books across twenty categories provide breadth within that selection; they do not stand for all human knowledge.
What was read, checked, and promoted
The workflow retained the editions, indexed source locations, and recorded selected passage readings. Each book then received a proposed thesis, a movement through selected material, conceptual facets, explicit comparisons, alternative emphases, and a reading ledger. Proposed definitions were compared with the existing vocabulary and with likely overlaps elsewhere in the batch.
AI drafting and peer inspection occurred within the same coordinated workflow. Mechanical checks covered source integrity, references, allocations, and review operations; editorial checks revisited selected meanings and evidence. Michael’s subsequent Keeper review and promotion produced the checked release. This was not an independent replication of all readings by another model.
The ledger records 3,892 inspected blocks within 76,404 indexed blocks. These are edition-dependent units, not equal amounts of text. A PDF page, a short verse block, and a paragraph can differ greatly. Dividing the two totals would not give a meaningful percentage of understanding.
The atlas therefore presents selected source-grounded readings, not exhaustive accounts. Long treatises and illustrated manuals have particularly visible coverage limits. More work on an already represented book may deepen the language as much as moving to another title.
Four statuses remain distinct: a retained source passage; a checked concept definition; a proposed application of that definition; and the larger authored reading. Promotion of a definition does not automatically approve an application or certify a whole interpretation.
The saved Keeper output verifies the reviewed release used here. Online publication is reported by Michael; this publication record does not claim a separate fresh audit of the website. Source hashes establish byte consistency, not interpretive authority. The inherited warning about historical working-ID reuse remains in the audit record.
A more grounded path toward 10,000
The first ten books added 327 concepts, an average of 32.7 per book. This collection adds 693, averaging 13.86. Across sixty books, the programme has added 1,020 concepts, or 17 per book.
The lower rate in the larger batch is not a controlled measure of efficiency or declining book quality. The works, reading depth, and available baseline vocabulary differ. Existing definitions now cover more territory; reuse and consolidation are valuable outcomes. No per-book quota was imposed.
| Collection | Books | New definitions | Average per book |
|---|---|---|---|
| First collection | 10 | 327 | 32.7 |
| Second collection | 50 | 693 | 13.86 |
| Both collections | 60 | 1,020 | 17 |
At 2,170 checked concepts, the distance to 10,000 is 7,830 net additions. If future books repeated this batch’s average, reaching the target would require 565 additional books. If they repeated the combined sixty-book average, it would require 461 additional books. Both results round upward to a whole book.
These are arithmetic scenarios, not forecasts. The effective rate can change with overlap, source difficulty, deeper rereading, merges, rejection, and the amount of human review available. A useful next collection could contribute fewer identifiers while greatly improving the range of expressions the existing vocabulary supports.
The next measure is what the language lets us say
This release enriches Carrier and makes its research visible. It does not report a new receiver experiment or claim that all fifty books have been transmitted through concepts. That later test will need an actual expression, recoverable references, declared dependencies, and an independent record of what the recipient understood.
The present achievement is a stronger foundation for that act of communication. Source readings now offer more precise ways to describe attention, proof, causal uncertainty, freedom, cooperation, material constraints, and unequal relationships.
The atlas lets a reader follow those possibilities back toward the books—and challenge the selections. As the dictionary grows, that path back is part of what makes its depth worth having.
Publication evidence: release results · reviewed concept bindings · findings and per-book accounting · original selection record. Historical scientific, medical, legal, and social claims remain attributable to their sources. The retained Gutenberg rights statements concern public-domain status in the USA; that scope is not silently extended worldwide.