Mohit1053/Humanizer

Multi-prompt AI text humanizer to reduce detection scores on GPTZero and ZeroGPT

What's novel

Multi-prompt AI text humanizer to reduce detection scores on GPTZero and ZeroGPT

Code Analysis

12 files read · 3 rounds

A collection of nearly-identical standalone Python scripts that read rows from a CSV, send each text to a local Llama3 model via Ollama with a prompt asking it to rewrite the text to sound more human, clean up the response with fragile string-splitting, and write results back to a CSV.

Strengths

The resume-from-interruption pattern (saving after each batch, skipping already-processed rows) is practical and well-implemented, and the prompt iteration process in optimize_prompts.py shows a reasonable experimental workflow for finding a better prompt.

Weaknesses

The entire codebase is 6-7 copies of the same ~50-line Ollama API call with slightly different prompt strings, with zero shared code, no package structure, no actual tests, a .env.example that nothing reads, and CI that can never fail; the 'humanization engine' is a single prompt template with no algorithmic substance.

Score Breakdown

Innovation

3 (25%)

Craft

42 (35%)

Traction

5 (15%)

Scope

40 (25%)

Signal breakdown

Innovation

Not Fork+1

Code Novelty+0

Concept Novelty+1

Craft

Ci+5

Tests-1

Polish+1

Releases+1

Has License+5

Code Quality+8

Readme Quality+15

Recent Activity+7

Structure Quality+5

Commit Consistency+2

Has Dependency Mgmt+5

Traction

Forks+0

Stars+6

Hn Points+0

Watchers+0

Early Traction+0

Devto Reactions+0

Community Contribs+0

Scope

Commits+5

Languages+3

Subsystems+5

Bloat Penalty+0

Completeness+7

Contributors+5

Authored Files+12

Readme Code Match+3

Architecture Depth+5

Implementation Depth+8

Evidence

Commits

Contributors

Files

Active weeks

TestsCI/CDREADMELicenseContributing

Repository

Language

Python

Stars

Forks

License

MIT