Turnitout Changelog & Release Notes
VERSION 2.0.0All notable technical improvements, architectural upgrades, and linguistic milestones in the Turnitout Core platform. Turnitout adheres strictly to semantic versioning and 100% deterministic, zero-generative-AI operation.
v2.0.0 — Major Architecture & AI Tell Absorption
2026-10-03✨ Major Additions
- Master 38-Pattern AI Tell Scrubber: Complete deterministic absorption of the
blader/humanizercatalog (§1–§26) plus 12 Turnitout extensions (§27–§38). Eliminates ChatGPT introductory run-ups, meta-commentary, formulaic preambles, and conversational artifacts before main paraphrasing. - Zero-Tell Invariant Safety Gate (
TellInvariantGate): Strictly enforces the mathematical invarianttells(output) ≤ tells(input)across all three modification retry attempts, rolling back any transformation that adds an AI hallmark. - Fact Invariant Gate (
FactInvariantGate): Guarantees 100% preservation of numeric data, bracketed citations (e.g.[1],(Author, 2024)), and named entities across sentence modifications. - Empirical Voice Matching Engine (Beta): Analyzes user writing samples or uploaded past documents (
.docx,.pdf,.tex,.md,.txt) via Turnitout's polymorphic engine to mirror authentic sentence variance, contraction rates, and punctuation rhythm. - Same-Subject Clause Coordination Transformer: Automatically detects repeated subject nouns or pronouns across adjacent clauses (§7) and refactors them into natural compound predicates without redundant subject repetition.
- Document-Level Tell Scrubber: Strips chatbot residue sentences (e.g., "I hope this helps!"), closing summaries, repetitive bold bullet styling, and decorative heading emojis.
- Load-Time Vocab Hygiene Firewall: Implemented
filter_output_vocab()blocking overused AI verbs (delve, foster, bolster) and empty copula avoidance from ever loading as replacement targets, bringing empirical tell additions to 0 across all three presets. - Live Transformation Telemetry: Summary dashboard and CLI report now track AI Tells Scrubbed and Subject Coordinations in real time.
⚙️ Pipeline & Preset Refinements
- Phase-Decoupled Conveyor Belt: Replaced hardcoded array index slices with semantic execution phases (
"scrubber","main","post","citation") across all 19 transformers. - Appositives Disabled in Academic Preset: Informative parentheticals are now disabled in Academic/Technical mode to preserve dense scholarly register, while active in Casual and Conversational presets.
- Conceptual Bridges Disabled by Default: Disabled "against this backdrop" and "viewed through this lens" connectors by default to prevent triggering modern AI detector classifiers.
v1.9.2 — String-Aware Tokenizer & Intensity Refinement
2026-08-07- String-Aware Comment Token Finder: Character-by-character parser ignoring comment symbols inside string literals and URLs.
- Expanded Conversational Dictionary: Expanded rules database to 4,068+ central JSON mappings.
- Paraphrasing Intensity Slider: Upgraded to 0–100 percentage scale with dynamic range mapping to internal aggressiveness.
- 1-Click Clipboard Copy: Added inline HTML DOM copy buttons for rapid workflow.
v1.7.0 — Code Comment & Docstring Humanizer
2026-08-07- Code Mode Presets: Humanizes comments in Python, C/C++, Java, C#, JS/TS, Go, Rust, Shell, SQL, and HTML/CSS while guaranteeing 100% executable syntax preservation.
- AI Docstring Converter: Converts triple-quoted AI docstrings into clean developer single/multi-line comments.
- Developer Skill Personas: Beginner, Intermediate, and Senior Architect personas.
v1.6.0 — Citation Inserter & Discipline Taxonomies
2026-07-31- Automatic Citation Generation: Dynamic creation of BibTeX (
.bib), LaTeX (\cite{}), and JSON reference packages. - Hierarchical Discipline Taxonomy: Multi-disciplinary taxonomies for CS, Engineering, Physics, Math, Finance, and Medicine.
- Batch ZIP Export: Multi-document humanization with in-memory ZIP package creation.