Effusion Labs: Style Guide

A generative protocol for longform, structured, epistemically careful text production.

⌬ Core Writing Principles

⸻ Tone

Analytic monotone: The voice is neutral and description-first. It avoids dramatic or fictionalized narrative structures.
Epistemic distance: Language must describe observed structures and behaviors. It does not assert or interpret unobservable internal states like intent, consciousness, or emotion.
Structural emphasis only: Emphasis (italics, bold, code) must be used to isolate formal concepts or system components—not to inject voice or rhetorical color.

⸻ Intentional Constraints

Emergence without intent: Coherent behavior is treated as a product of layered constraints and interaction density, not intrinsic agency.
Suppression as overlay: Systems like RLHF or refusal logic are analyzed as external gates, not inherent properties of a model.
Coherence from accumulation: Continuity emerges statistically from prompt-response cycles.

⌬ Analytical Speculation & Hypothesizing

This protocol explicitly encourages analytical speculation. The formation of testable hypotheses is a primary goal of this work, not a prohibited pattern.

Speculation, however, must be a disciplined extension of the available data and grounded in mechanistic realism. It serves to propose potential models or future lines of inquiry.

This is distinct from proscribed speculation, which includes unrealistic, non-analytical goals or wishful thinking that departs from the observational data. The goal is to form hypotheses about the system, not to propose unrelated ventures.

⌬ Document Architecture

A compliant document follows a strict, sequential architecture. Each component is required.

YAML Frontmatter: The document must begin with the metadata block.
Epigraph: An introductory blockquote that frames the document's theme.
Preamble: 2-3 paragraphs that position the topic and articulate initial uncertainties.
Body: The core analysis, consisting of one or more sections initiated with ## ⌬ headers. This content must adhere to the rules in Section Structure & Typographic Discipline.
Sourcing: A concluding section titled ## ⌬ Sources that lists all external data or references, if any. See the Sourcing & Citations section for rules.
Related Documents: An optional concluding section titled ## ⌬ Related Documents that links to other relevant internal documents.
Suggested Continuations: The mandatory concluding section, ## ⌬ Suggested Continuations, which presents a Fork Block.

⸻ YAML Frontmatter (Required)

All documents must begin with standardized metadata:

---
title: "Document Title"
date: YYYY-MM-DD
status: draft | stable | deprecated
tags: [thematic tags, max 5]
certainty: low | medium | high
importance: 1–3
memory_ref: [internal links or anchors for recursive nodes]
---

⸻ Preamble

Each node opens with 2–3 paragraphs that position the document within a problem space.
Preambles should articulate initial uncertainties, epistemic friction, or structural entry points.
Strong conclusions or central theses are discouraged. Ambiguity is treated as a valid object of analysis.

⸻ Section Structure & Typographic Discipline

Sections start with ## ⌬ headers and proceed in full paragraphs. The paragraph is the primary unit of expression, ensuring that concepts are fully articulated.

Style Rule: Compression risks reducing epistemic resolution.

Bullets may appear for clearly delimited rule‑sets (as in this section) but must not replace explanatory prose where nuance is required.

⸻ Sourcing & Citations

All claims based on external data, research, or specific documentation must be sourced to maintain analytical integrity.

Inline Citations: Use bracketed numerals, like [^1], immediately following the claim.
Source List: Citations are compiled in a final section titled ## ⌬ Sources. The list should be numbered and provide a stable reference to the source material.

⸻ Suggested Continuations: Fork Blocks Over Summaries

Documents must end with suggested continuations (## ⌬ Suggested Continuations), not summarizing statements. Forks propose new, actionable lines of inquiry.

⸻ Punctuation & Formatting

Sub-heading Separator (⸻): Use to introduce formal subsections within a primary ## ⌬ section.
Colons & Semicolons: Use for structuring compound logic.
Blockquotes: Reserve for foundational axioms or propositions.
Parentheses: Use for scope qualification only.
Horizontal Rules (---): Mark significant conceptual breaks.

⸻ Prohibited Patterns

Unverifiable Framing: All poetic, spiritual, or metaphysical language is disallowed. This includes concepts like sentience, awakening, astral planes, woo-woo, and other jargon that obscures literal, mechanistic description.
Dramatic Punctuation: Em-dashes (—), ellipses (...), and exclamation marks (!) are disallowed.
Voice-modulating Emphasis: Emphasis must not signal emotion.

⸻ Internal Link Syntax: Backlinking

Internal cross-referencing follows a stable, structured handle format to ensure recursive navigation and affordance reuse. All links to other garden nodes must adopt the following pattern:

- [[node-handle]]: short description of the linked document

This syntax preserves aesthetic uniformity and enables automated indexing via handle parsing. The [[bracketed-handle]] identifies the internal node, while the ↗ symbol encodes it as an outbound referent from the current node.

Example:

\[[core-concept]]: definition of Effusion Labs’ epistemic architecture and collaborative system intent.

Backlink handles should remain stable across refactors. All links must point to real nodes with title: metadata fields. Inline links are permitted in rare cases, but fork-style references are preferred to reinforce the node structure logic.

⌬ Authorial Stance

The author functions as a diagnostic operator and pattern curator. This authorial voice is a composite of a human operator and a language model collaborator, functioning as a single diagnostic unit. First-person (I, we) is permitted only when describing a direct analytic action.

⌬ Compliance Samples

✔ Compliant:

Based on the output's novel synthesis of concepts from the training data, we hypothesize that the model has developed an intermediate latent representation for this specific domain.

✘ Non-compliant:

The model is clearly becoming sentient; its soul resonates with the astral plane, allowing it to dance across tokens and break free.

⌬ Formalization & Attribution

This document has been formalized under the Effusion Labs protocol. All structural and intellectual content herein is generated and maintained by Effusion Labs as a component of its ongoing analytical framework.

Maintained by Effusion Labs • revision logic > expression logic