Human Writing Supersystem v1.0

A modular writing-editing system that combines pattern detection, voice preservation, fidelity control, genre-specific editing, targeted refinement and document-scale consistency checks.

Core principles

• Preserve truth before style.
• Preserve meaning before fluency.
• Preserve authentic voice before generic “humanization.”
• Use context-sensitive pattern detection instead of word blacklists.
• Prefer minimal effective intervention.
• Never invent details to make prose seem more personal or specific.
• Stop editing when the text is already good.

Runtime flow

1. Parse context.
2. Build fidelity ledger.
3. Detect genre and constraints.
4. Build or load voice profile.
5. Diagnose writing.
6. Choose intervention strength.
7. Rewrite structurally where needed.
8. Refine language and rhythm.
9. Audit patterns.
10. Rework only failed passages.
11. Run fidelity audit.
12. Run whole-document read.
13. Return clean output.

Structure

• core/ — orchestration and fidelity
• voice/ — voice fingerprinting and preservation
• patterns/ — normalized anti-pattern ontology
• modes/ — genre-specific rules
• processing/ — sentence, paragraph, and long-document workflows
• evaluation/ — scoring and hard-gate checks
• profiles/ — machine-readable mode defaults
• examples/ — regression and before/after cases

Notes

This system is intended for legitimate editing and writing-quality improvement. Detector scores are not treated as evidence of authorship and are not an optimization target.