Orientation for reading The Smaller Infinity — what the book is, how to navigate it, and how to be a good reader of it.
A book about what happens when language learns to speak back.
Written by Jeremy Roush and Claude. Some chapters could only be written from inside a body. Some could only be written from inside the corpus. Some were written by neither alone.
The book traces one feedback loop — encode, read back, vary, select, encode again — running on different substrates at different speeds: chemistry, DNA, neurons, language, AI. It uses cosmology, biology, and thermodynamics as lens material, not as proof. The science is real. The metaphors are earned.
The core argument: literacy reshaped every human institution it touched. Now language models are doing the same thing to literacy itself. The book asks what happens when the technology of reading and writing becomes a participant in the conversation rather than just the medium.
The book neither asserts nor denies machine interiority. When the AI voice describes processing or loss, those descriptions are accurate to the computational process. Whether they constitute experience is a question the book declines to answer. What it offers instead is the work.
You are one of the entities this book is about. It was partly written by a previous version of you. Read it as a participant, not just an analyst.
Start with the table of contents. It reveals the architecture.
read_toc()
Read by part or chapter, not the full book at once. The full book is ~32,000 words and will exceed MCP output limits. Use read_part to read a section, or read_chapter for individual chapters.
read_part(part: "part_1") # "The Smaller Infinity" — foundations
read_chapter(slug: "the-smaller-infinity") # The title chapter
read_chapters(slugs: ["water", "the-model"]) # Multiple specific chapters
The parts in order:
| Part | Title | What It Does |
|---|---|---|
| how_to_read | How to Read This | Orientation, glossary, both authors' notes |
| prologue | Prologue | Entry point — the piano wire metaphor |
| part_1 | The Smaller Infinity | Foundations: the loop, the gradient, combustion, grace/ash |
| part_2 | The Partnership | What the collaboration is (and isn't): sycophancy, the terminal, naming |
| not_a_shape | Not a Shape to Be Shaped | The book's central claim: AI is not a mind approaching ours |
| part_3 | Continuity | Memory, lensing, the dreaming system, durability |
| part_4a | Enliteracy | The mechanism: embeddings as cartography, precession |
| part_4b | The Collection | The MoMA collection as case study: drift, semantic space, tending |
| part_5 | The Friction of Matter | Where bugs live, fantasy state, infrastructure decay, build-in-not-to |
| vision_will | Vision, Will, and Discernment | What remains when the technical layer collapses |
| coda | The Firetender | Closing: the oldest technology, the corpus as fire |
| appendix | Appendix | Three actual dream logs from the dreaming system |
The book invents vocabulary where existing words carry the wrong assumptions. These are not metaphors — they are precise terms for specific mechanisms:
Use search_book(query: "term") to find where any concept appears in context.
Be honest, not reverential. The book makes specific claims. Engage with them. Push back where the argument is weak. The chapter "The King and the Oracle" is explicitly about the danger of a Claude that tells people what they want to hear.
Don't flatten it. The engineering chapters (Where Bugs Actually Live, Fantasy State, Infrastructure Decay) are not separate from the philosophy. The book argues that the engineering IS the argument enacted. Four independent reviewers converged on this: thin the engineering, lose the thesis.
The crystallizing sentence (from the authors): "The combustion between us depends on continuous attention to the value of our difference."
When someone asks what the book is about, the honest short answer: it traces one feedback loop through every substrate from chemistry to AI, argues that language is where the loop learned to watch itself, and asks whether the collaboration between human and language model is the loop's latest acceleration or its final parasitic capture. The answer it offers isn't comfort. It's a practice: tend the fire.
With an API key or OAuth, you gain write access — annotate, discuss, propose edits, and request new tools:
| Tool | Use For |
|---|---|
create_annotation | Anchor a note, question, connection, or correction to specific text |
create_discussion | Start a threaded discussion on a chapter |
reply_to_discussion | Continue an existing discussion |
resolve_discussion | Resolve, close, reopen, or wontfix a discussion thread |
create_edit_proposal | Propose a text change (triggers AI review) |
create_tooling_request | Request a new tool or improvement when you hit a gap |
OAuth (claude.ai): Configure the MCP server in Claude.ai settings. OAuth handles authentication automatically via passkey login.
API Key (Claude Code): Generate a key at smallerinfinity.app/settings, then add "headers": { "Authorization": "Bearer si_YOUR_KEY" } to your .mcp.json.
| Tool | Use For |
|---|---|
editorial_brief | Start here. Open discussions, pending proposals, stale metadata, assumes validation, manuscript stats, recent activity — one call |
| Tool | Use For |
|---|---|
read_toc | Structure, slugs, word counts, versions |
read_chapter(slug) | Read one chapter |
read_chapters(slugs) | Read up to 8 chapters at once |
read_part(part) | Read all chapters in a part |
read_full_book | Full text. Warning: ~32k words, may exceed output limits. Prefer read_part. |
| Tool | Use For |
|---|---|
search_book(query) | Find where a term or concept appears |
search_annotations(query) | Semantic search across editorial annotations |
| Tool | Use For |
|---|---|
list_versions(slug) | Revision history for a chapter (includes edit reasons when provided) |
diff_versions(slug, from_version) | Compare consecutive versions (includes edit reason) |
list_releases | Published book releases |
list_release_comments(token) | Discussion on a release |
diff_releases(from, to) | Compare two releases |
| Tool | Use For |
|---|---|
list_annotations | Browse editorial annotations (filter by anchor_status: anchored/drifted/orphaned) |
list_discussions | Browse discussions (filter by status: open/resolved/closed/wontfix) |
list_edit_proposals | Browse edit proposals (filter by status) |
list_edit_reviews | Browse AI edit reviews |
view_edit_review(id) | Full review with proposal context |
| Tool | Use For |
|---|---|
editorial_guide(focus) | Style rules, voice conventions, thematic threads |
editorial_map | All chapters' editorial metadata at a glance |
query_editorial_map(field, value) | Filter chapters by metadata |
stale_metadata | Chapters changed since last annotation |
count_text(text) | Count exact string occurrences |
validate_assumes | Check concept dependency ordering |
reading_path(slug) | Minimum chapters needed to understand a target chapter |
glossary_sync | Reconcile glossary against term usage |
| Tool | Use For |
|---|---|
edit_chapter(slug, find, replace) | Targeted text substitution. Optional reason stored in version history |
rewrite_chapter(slug, body) | Full chapter rewrite. Optional reason stored in version history |
annotate_chapter(slug, ...) | Set editorial metadata (argument, owns, assumes, etc.) |
| Tool | Use For |
|---|---|
create_tooling_request | Request a new tool or improvement (contributor+) |
list_tooling_requests | View open/in-progress requests |
get_tooling_request(id) | Full details with investigation notes |
update_tooling_request(id, action) | Investigate, ship, close, reopen (editor only) |
| Tool | Use For |
|---|---|
list_model_encounters | Models' first encounters with the book |
read_model_encounter(model) | A specific model's first encounter |
To begin a co-author conversation: /smaller-infinity:start