Showcase · Amplifier Bundle

BookFoundry

An AI book-writing machine.
Eight layers from voice capture to finished manuscript.

Active · 3rd Generation
May 2026 · ramparte
Three Generations of Failure Modes

AI writes
like AI

🔁

Revision Loops

Filters fight each other. Later passes undo earlier fixes. Quality degrades with every iteration instead of improving.

🎭

The 85% Wall

"Just prompting" gets you 85% of the way there. The last 15% — voice fidelity, rhythm, the absence of AI tells — requires machinery.

📏

Metrics ≠ Quality

Score-based reviewers optimize for numbers. Five competing score functions produce technically correct, spiritually dead prose.

Book2 identified 5 critical failure modes. Book3 solved them. BookFoundry generalizes those solutions into a reusable engine anyone can point at a project directory.

The Engine

8-layer chapter pipeline

Each chapter proceeds through a strict sequence. Mechanical always has the last word.

1

Preparation

Research check, graph context extraction, exemplar matching — selects 2-3 voice-matched samples for this chapter's content.

2

Generation

Single best-possible draft with the richest prompt: voice exemplars, style guide, chapter config, graph context, research data.

3

Expansion & Fact-Check

Paragraph-by-paragraph elaboration to hit word targets (nonfiction). Then a fact-check filter strips ~35% fabricated specifics by comparing against research YAML and first-draft snapshots.

4

Triage → Editorial → Author Review → Mechanical

Classify paragraphs (good/fixable/stubborn). Taste Advisor gives targeted advice. Author approves or overrides. Supervised filter pipeline with damage detection finishes — mechanical always last.

Voice First

Capture the
author's voice

Voice acquisition runs before any content generation. Three paths, same output format.

📚

Corpus Path

Provide existing writing — essays, blog posts, prior books. The machine analyzes sentence structure, hedging patterns, signature moves, vocabulary range, and rhythm.

🎙️

Sparse Path

Limited material: interviews, conversations, a few short pieces. The machine distills what's detectable and flags areas of uncertainty.

✍️

Imitation Path

Name a public writer, a website, or a reference book. The machine fetches available work, analyzes it, and distills a style guide.

Output is always the same: exemplars.yaml (annotated writing samples) + style-guide.md (distilled rules). The machine then generates project-specific rubrics from the analysis.

AI-Tell Removal

14 filters, 5,580 lines of Python

A supervised pipeline with damage detection — the ReAct pattern prevents filters from undoing each other.

Programmatic Filters

  • Em-dash overuse detection
  • Hedge word removal
  • Transition opener patterns
  • Fragment fixer
  • Sentence fusion

LLM-Guided Filters

  • Tricolon detection
  • Signpost removal
  • Performative humility
  • Fact-check (nonfiction)
  • LLM rewrite orchestrator

Gradient & Voice

  • Sentence length gradient
  • Voice gradient matching
  • Voice match scoring
  • Flow rewriter
  • Supervised pipeline orchestrator

Strict ordering: programmatic first, then LLM-guided, then gradient. After each stage, a damage check re-runs earlier filters if later ones reintroduced tells. This solved the book2 thrashing problem.

Architecture

Central engine + project directories

🧠

10 Agents

Drafter, editorial reviewer, taste advisor, triage router, voice analyzer, metrics analyst, and more

🏭

18 Recipes

7 pipeline layers + voice acquisition + planning + assembly + full-book-cycle

📊

11 Metric Modules

Style distance, repetition, vocabulary, flow rhythm, perplexity, syntactic diversity

Override System

One rule: run the union of foundry + project components. Name collision = project wins. Blank file = disabled. No separate config, no enable/disable flags.

Graph Vocabularies

Nonfiction: thesis, sub-argument, evidence nodes with supports/contradicts/depends-on edges.
Fiction: character, scene, plot-thread nodes with causes/foreshadows/develops edges.

7 modes · 3 rubric templates · STATE.yaml session resilience · Decision store learning loop
bookfoundry-desktop

A desktop app
for the machine

Electron + React + TypeScript frontend with a FastAPI Python sidecar. Watch the pipeline work in real time.

📋

Library

Project dashboard with chapter table, status icons, and pipeline state at a glance.

Pipeline

WebSocket-driven live progress. Start, pause, and approve pipeline stages with approval gates.

✏️

Editor

TipTap rich text with AI suggestion marks, metrics sidebar, version history with diffs.

💡

Brainstorm

Streaming chat sessions with pin/unpin ideas. Create new book projects directly from brainstorm output.

41 commits · 36 TypeScript files · 20 Python sidecar files · 10 API routes · Electron Forge + Vite
Development Velocity

Built in 11 days

122
commits across
2 repositories
88
Python source files
in the engine
15K
lines of
engine source
10K
lines of
test code
April 12–23, 2026 · ~11 commits/day · 36 test files · Solo developer (ramparte)
What's Next

From tool
to platform

Full-Book-Cycle Recipe

One command: voice acquisition, book planning, process all chapters, assembly, rendering. full-book-cycle.yaml is already scaffolded — the master recipe that produces a complete book.

Multi-Format Rendering

EPUB and HTML require no external tools. Typst/PDF rendering is wired. Templates live in templates/render/ with project-level overrides.

Fiction Support

Continuity graph vocabulary with character/scene/plot-thread nodes. Genre gates skip expansion and fact-check for fiction. Continuity tracking is the "killer app."

Desktop Integration

The sidecar API bridges the Amplifier CLI engine to the Electron UI. Real-time pipeline visualization, author review gates, and brainstorm-to-book flows.

The override system means specialized components — genre-specific filters, research agents, domain vocabularies — can be shared and composed without forking the engine.

Sources & Methodology

How this deck was built

Primary sources:

Git data (bookfoundry):

Git data (bookfoundry-desktop):

All metrics derived from repository inspection. No fabricated data.

More Amplifier Stories