AI-first fiction engineering

Turn a story idea into a book without losing the book.

Alephink is a native macOS workspace for long-form fiction: manuscript organization, chapters, Story Bible, Memory Delta, research materials, Skills, backups, exports, and AI proposals that stay reviewable.

Alephink preview
ManuscriptChapter identity, Binder Order, Threadboard, snapshots, and compile readiness.
MemoryStory Bible, Memory Delta, hard cases, and source-backed long-form recall.
ReviewAI writes candidates and diffs; the author decides what becomes canon.
LocalVault, notes, research, backups, exports, and keys stay controlled by the writer.

One book, many projections.

Alephink's product direction is not "more buttons for writers." It is a stable writing core where every view serves the book: draft, canon, memory, research, and export.

Draft

Write chapters in order

Keep chapter identity, Binder Order, Threadboard cards, candidate exclusions, and export manifests aligned so the book remains understandable.

Canon

Protect story truth

Story Bible, Memory Delta, character state, relationships, foreshadowing, timeline, and world rules are treated as canon-sensitive material.

AI

Collaborate with restraint

AI can plan, continue, inspect, summarize, extract memory, recommend templates, or explain risk, but every major write path stops at a candidate.

Built around the shape of a book

A novel project is more than loose Markdown files. Alephink gives the manuscript a durable structure without turning the author into a database administrator.

First day

Start without configuring everything

New-book tasks, old-draft import, current-chapter continuation, and material import are intended to appear as direct writing paths.

Research

Turn materials into usable context

Local research import, OCR, MarkItDown, AI understanding, and candidate setting extraction all feed the book through review gates.

Export

Compile the manuscript deliberately

Compile readiness should show which chapters are included, which candidates are excluded, and what gaps need attention before handoff.

Stable core first

The current roadmap prioritizes trust: manuscript order, reviewable AI, reliable backup and restore, local references, and performance for large vaults.

  • Scrivener-level manuscript organization, adapted for AI-first fiction workflows.
  • Cursor-like collaboration, but specialized for chapters, canon, and memory.
  • Obsidian / Joplin-like local storage that remains inspectable and migratable.
  • Long-form memory that grows only from accepted chapters and evidence.