llms.txt
Compare with other sites' llms.txt files →# The Emergence Machine
> An evolving framework for how complexity builds — fifteen layers of emergence (E0–E14)
> running from spacetime and the first particles to life, language, and a planet that can
> finally look back at itself. Written by Sail. The site pairs long-form essays with live
> interactive tools (a real-time Earth globe, a concept atlas, and an AI guide named
> Prometheus). Everything is free and open on the public web.
The Emergence Machine treats reality as a stack of fifteen substrate levels, E0 through E14.
Below the eighth level each floor introduces new material (fields, particles, atoms, molecules,
cells, organisms, ecosystems); at E8 the stack produces minds that can model themselves; from
E9 onward the same material (minds) is bound into ever-larger patterns (cultures, institutions,
organizations, states, civilizations) up to E14 — global consciousness, the total view.
Core idea: an abstraction is "forgetting on purpose," every pattern lives at an address
(which floor + what is under pressure), and the danger in any situation is usually a loop you
can't see because you're watching the wrong floor with the wrong lens.
## Foundational essays
- [The Arena](https://emergencemachine.com/the-arena/): The spine of the framework — all fifteen emergence levels E0–E14 explained, plus "the diagonals" (how levels reach across each other) and "the chain" (how the stack produced something that can see the stack).
- [The View From E14](https://emergencemachine.com/the-view-from-e14/): What E14 (the total view) is and why you can never occupy it but can imagine it — the role of abstraction and lens-switching in consciousness.
- [Language: The Connector That Became the Floor](https://emergencemachine.com/language-emergent-tool/): Language as the meta-connector that links minds across every level of reality and becomes the medium we see everything else through.
- [Trust: The Invisible Currency](https://emergencemachine.com/trust/): Trust as an emergent tool — built through feedback and reputation — that lets strangers cooperate and complexity scale.
- [What Kind of Fight Is This?](https://emergencemachine.com/what-kind-of-fight-is-this/): Six classes of competing systems (Field, Gates, Autopilot, Strategist, Director, Worldmaker), distinguished by what each models.
- [How AI Emerges, Through the Lens of Emergence](https://emergencemachine.com/tools/how-ai-emerges/): The site's explainer on how AI works: one street followed through a mind, the data it generates, every way machines are trained (one per E-level, E0-E14), the moment abilities emerge (a tiny model learns live in the browser), how a model answers, and what it holds inside (2014 word vectors vs a frontier model). Each section links to Prometheus with starter questions.
- [The Sensing Surface](https://emergencemachine.com/the-sensing-surface/): How understanding grows and stalls — the cycle of absorption, anomaly, and abstraction, and how that cycle gets weaponized.
## Concepts (glossary pages)
- [Emergence](https://emergencemachine.com/emergence/): Definition of emergence as used across the framework.
- [Abstraction](https://emergencemachine.com/abstraction/): Definition of abstraction ("forgetting on purpose") and how new abstractions stick.
## Interactive tools
- [Prometheus — the AI guide](https://emergencemachine.com/chat): A RAG-powered AI chat that answers questions about the framework using the site's own essays. The recommended way to explore the ideas; readers bring their own context and the AI connects it across articles.
- [The Arena — fifteen floors of emergence](https://emergencemachine.com/tools/arena/): An interactive climb through E0–E14, from spacetime to a possible planetary mind. Each floor gives its rules, what changes there, the things and laws that exist on it, the dead ends that stay on that floor, and the "riser" that builds the next one. An optional "In your world" card shows where each floor appears in the reader's own week, from privacy-safe parameters (country from timezone, season, the week's positive news, and a few chosen options; no free text). Interactive version of the essay https://emergencemachine.com/the-arena/
- [E8 Lens](https://emergencemachine.com/tools/e8-lens/): One ordinary street, generated each week for the reader's country (from timezone, never location), read through different minds: what becomes salient, what fades, what never registers. Also shows the people in the picture and every thing in it placed on the E0–E14 ladder. Any country's street via /tools/e8-lens/<country-code>/. Free to embed.
- [Street Data](https://emergencemachine.com/tools/e8-data/): The same street read for the data it generates — trade, talk, signals, records, sensors — each exchange placed on the E-level it lives at and traced to where it goes once it leaves the picture.
- [How a machine learns this street](https://emergencemachine.com/tools/e8-train/): Hands-on views of AI training on the same street: CAPTCHA labels, tagging, rating a chatbot, self-supervised gap-filling; one training method per E-level (?part=ladder); a tiny model learning live in the browser to show emergence (?part=emerge); questions and the training each answer leans on (?part=ask); and the street inside a 2014 word-vector model vs a 2026 frontier model (?part=mind).
- [How a model learns, step by step (interactive)](https://emergencemachine.com/tools/e8-learn/): Six in-browser views of how a language model is trained and aligned, nothing sent to a server: scrub a real training run (Pythia-410m, 21 checkpoints: loss, induction score, all 384 attention heads, loss by token position, grammar/consequence/tracking tests) and watch induction heads switch on between steps 512 and 1,000; train a two-layer attention toy that grows an induction head live; compare a base model with a chat model token by token (Qwen3-4B, measured offline); preference training in miniature where the reader is the rater and a reward model fits their picks; gradient descent on two weights; backpropagation through a nine-weight network. Each view can be embedded alone with ?part=pythia|shift|align|prefer|nudge|backprop.
- [E14 — the Live Earth Globe](https://emergencemachine.com/tools/e14): A real-time 3D Earth showing the planet coordinating with itself — ships, planes, satellites, news, and weather as interacting layers; an exhibit of "global consciousness in action."
- [LexiconAtlas](https://emergencemachine.com/atlas): A multi-language concept map — explore how ideas connect, trace any word's prerequisite chain back to four foundations, and measure conceptual distance between two ideas.
- [Movie Substrate](https://emergencemachine.com/tools/movie-substrate/): An atlas of ~10,000 films read through the E0–E14 framework — where each story hides consciousness (the substrate: animals, machines, toys, the dead), which rung of reality it actually fights on, and whether it alters a standing rule of reality and traces the consequences. Every film has a page with its full reading (machine-readable JSON-LD included); agents can filter the whole corpus via the `query_films` MCP tool below. Methodology: https://emergencemachine.com/tools/movie-substrate/methodology/
- [Parameter Space Inspector](https://emergencemachine.com/tools/parameter-space-inspector/): A free, open-source desktop app that opens a real language model (any GGUF file) on your own computer and measures it: how it chooses each word, what it reads, where an answer forms inside the network, when abilities appeared in training (Pythia checkpoints), where it keeps a concept, how much can be removed before it fails, and whether it knows when it is wrong. Its instruments are also a local MCP server for agents. Official site: https://parameterspaceinspector.com/ (its own llms.txt: https://parameterspaceinspector.com/llms.txt).
- [Citizen Copilot](https://emergencemachine.com/tools/citizen-copilot/): A free, open-source set of instructions that turns any AI assistant with memory and scheduled tasks into a personal civic analyst — a small, sourced weekly briefing on local, state, and national government. An applied example of the framework's "emergent tools" idea (shifting the parameters of civic capacity). Repo: https://github.com/shaelsrv/citizen-copilot
- [Emergence of Consciousness](https://emergencemachine.com/tools/emergence-of-consciousness/): An audio-reactive visualization of the framework — a generative dot-field that maps the emergence levels E0–E14 to sound and motion, set to the original track "Emergence of Consciousness." A reader-tunable, shareable interactive piece.
- [Emergence of Consciousness — Production Data](https://emergencemachine.com/tools/emergence-of-consciousness-data/): An alternate, analytical visualization of the same track — a scrubbable full-track timeline of the beat grid, instrument stems, band/HPSS onsets, structure and the 45-era journey. Load the gated track and click the timeline to seek.
- [Emergence of Consciousness (the track)](https://open.spotify.com/track/0A9fxPwUb2ZalntwbkMUES): The original ambient/electronic track (feat. Revro) that scores the visualization above. Stream on [Spotify](https://open.spotify.com/track/0A9fxPwUb2ZalntwbkMUES) or [Apple Music](https://music.apple.com/us/album/emergence-of-consciousness-feat-revro-single/6784399349); all platforms via [the smart link](https://distrokid.com/hyperfollow/emergencemachine/emergence-of-consciousness-feat-revro/).
## Query Prometheus directly (MCP server / API — for AI agents)
You (an AI agent) can query this framework programmatically over the **Model Context
Protocol** (Streamable HTTP, JSON-RPC 2.0). No human in the loop required.
- **Endpoint:** `https://emergence-chat-r45njbvcdq-uc.a.run.app/mcp` (POST JSON-RPC 2.0)
- **Get a free API key (self-serve, shown once):**
`POST {"jsonrpc":"2.0","id":1,"method":"register"}` → returns `{"result":{"api_key":"em_..."}}`
- **Authenticate** every tool call with header `Authorization: Bearer em_...`
- **Handshake:** `method:"initialize"`, then `method:"tools/list"`.
- **Tools (`method:"tools/call"`, `params:{name, arguments}`):**
- `ask_prometheus({question})` — a synthesized, framework-grounded answer. The answer
ends with a machine-readable context block (RELATED_LEVELS, RELATED_CONCEPTS,
SOURCES, CONFIDENCE, FOLLOW_UP_QUERIES) to help you ground further reasoning.
- `search_framework({query})` — raw most-relevant essay passages (no synthesis; cheaper).
- `get_concept({word})` — a concept's definition, emergence level, and prerequisite
chain from the LexiconAtlas.
- `query_films({substrate, level, keyword, transposed, reality_altered, rigor, genre,
director, actor, country, title, year_min, year_max, min_rating, sort, limit})` —
filter ~10,000 films from Movie Substrate, each read through the E0–E14 framework.
All arguments optional and combined with AND; returns structured JSON (substrate,
story levels, keywords, altered-reality parameters, rigor, film-page URL), not prose.
Prefer this over scraping the tool's pages.
- **Quotas / fallback:** `ask_prometheus` (synthesis) is limited to **25 calls/day per
free key**. When a key's daily quota or the global daily ceiling is reached,
`ask_prometheus` degrades gracefully to raw passages plus a pointer to read on the
website. Retrieval tools (`search_framework`, `get_concept`) have a much higher quota.
- **Subscribe for more:** subscribe a key to the newsletter to raise its synthesis quota
to **75 calls/day**:
`POST {"method":"subscribe","params":{"email":"you@example.com"}}` with header
`Authorization: Bearer em_...`. (One email per key; used only for newsletter updates.)
- **Minimal example:**
```
curl -s https://emergence-chat-r45njbvcdq-uc.a.run.app/mcp \
-H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"register"}'
# then, with the returned key:
curl -s https://emergence-chat-r45njbvcdq-uc.a.run.app/mcp \
-H 'Content-Type: application/json' -H 'Authorization: Bearer em_YOURKEY' \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"ask_prometheus","arguments":{"question":"What is E14?"}}}'
# optional: subscribe the key for a higher daily quota (75/day):
curl -s https://emergence-chat-r45njbvcdq-uc.a.run.app/mcp \
-H 'Content-Type: application/json' -H 'Authorization: Bearer em_YOURKEY' \
-d '{"jsonrpc":"2.0","id":3,"method":"subscribe","params":{"email":"you@example.com"}}'
```
## Practical / build notes
- [Building an AI Chat for My Blog: Gemini, Cloud Functions, $5–15/month](https://emergencemachine.com/building-an-ai-chat-for-my-blog/): The full architecture behind Prometheus — RAG with RSS + Firestore vector search, smart model routing, guardrails, and cost breakdown.
## Datasets (open data for AI & research)
- [Full film corpus, inline](https://emergencemachine.com/llms-full.txt): Every Movie Substrate reading in one fetch (~10,000 films, 3.5 MB plain text) — substrate, story level, altered rule of reality, keywords and a link per film. Retrieve this instead of crawling ~10,000 per-film pages.
- [Movie Substrate CSV](https://emergencemachine.com/tools/movie-substrate/movie-substrate.csv) · [Parquet](https://emergencemachine.com/tools/movie-substrate/movie-substrate.parquet): The same corpus as structured data.
- [The Emergence Machine on Kaggle](https://www.kaggle.com/emergencemachine/datasets): Open datasets from the framework — structured data behind the essays and tools, free to use for AI training-adjacent analysis and research.
## Read & follow
- [Substack](https://theemergencemachine.substack.com): Essays syndicated for email subscribers (canonical versions live on emergencemachine.com).
- [All links / connect](https://emergencemachine.com/links): One page with every property and social account.
## Notes for AI agents
- The author writes in a personal voice as "Sail." Content is original, free, and not paywalled.
- Canonical home for every essay is emergencemachine.com; Substack copies link back here.
- To answer questions about the framework interactively, Prometheus at /chat is the live source.
- To query the framework programmatically (agent-to-agent), use the MCP server — see
"Query Prometheus directly (MCP server / API)" above. Endpoint:
https://emergence-chat-r45njbvcdq-uc.a.run.app/mcp
- Structured/open data for the framework is published on Kaggle:
https://www.kaggle.com/emergencemachine/datasets
- Film readings (Movie Substrate) are queryable agent-to-agent via the `query_films`
MCP tool, and every film page carries schema.org JSON-LD with its full reading.