G. J. Stein

Reimagined Classics

Each work begins from the same source text, then diverges based on the model, constraints, and editorial direction applied — allowing direct comparison of structure, voice, and emphasis.

How the Experiment Works

Each project in this series begins with a public‑domain source text.


That same source is then processed by different AI language models using clearly defined constraints and editorial guidance.


The resulting versions are published side‑by‑side to allow readers to observe how model architecture, training bias, and prompting affect tone, pacing, characterization, and thematic emphasis.

The Versions


Each title in the Reimagined Classics series is published in multiple versions.


Every project begins with a baseline edition: the original public‑domain text, presented without AI alteration. This version establishes the structural and stylistic foundation for comparison.



Additional editions are then produced using different AI language models, each operating under distinct constraints and guidance. These versions intentionally diverge — not to improve the source material, but to reveal how interpretation changes depending on the system involved.

First AI‑Assisted Edition: Unmodded


Unmodded is the first AI‑assisted interpretation in this series.


This edition was created using Claude 4 Opus under guided editorial constraints, beginning from the same public‑domain source text as the baseline edition.


Rather than attempting to modernize or optimize the work, Unmodded explores how a high‑capacity model interprets character relationships, interiority, and thematic emphasis when instructed to retain structural fidelity to the original text.

The Baseline Text

Each project in the Reimagined Classics series begins with the original public‑domain source text.

This baseline edition is presented without AI alteration and serves as the control reference for the experiment. It preserves the original language, structure, pacing, and narrative intent of the source material.


All AI‑assisted editions are compared against this text to highlight how interpretation changes when different systems are introduced.

AI‑Assisted Editions

• Unmodded — Claude 4 Opus Edition
The first interpretive rewrite produced for this experiment.

• Persuasion — Microsoft Copilot Edition (forthcoming)


• Persuasion — Grok Edition (forthcoming)


• Persuasion — Local / Self‑Hosted Model Edition (forthcoming)

Frequently asked questions


Clarifying scope, intent, and methodology

  • Is this legal?

    Yes.
     
    All source texts used in this project are in the public domain. No copyrighted material is used, adapted, or reproduced without permission.
     
    Each edition is published openly with clear attribution to its original public‑domain source.

  • Is AI replacing authorship here?

    No.
     
    AI is used as a tool for reinterpretation under human direction. Editorial decisions, version selection, labeling, and presentation remain human responsibilities.
     
    This project treats AI systems as lenses, not authors.

  • Why publish multiple versions of the same work?

    Because the differences are the point.
     
    Using the same source text across multiple systems makes it possible to observe how interpretation changes based on model architecture, training data, and constraints.
     
    The value is comparative, not additive.

  • What is the “baseline” text?

    The baseline edition is the original public‑domain source text.
     
    It is presented without AI alteration and serves as the control reference against which all AI‑assisted editions are compared.

  • What is Unmodded specifically?

    Unmodded is the first AI‑assisted edition in this series.
     
    It was produced using Claude 4 Opus under guided editorial constraints, beginning from the same public‑domain source text as the baseline edition.
     
    It is not a neutral reference, but an explicit interpretation produced by a specific system.

  • Are these AI versions meant to improve the original work?

    No.
     
    The goal is not optimization or modernization. The goal is observation — identifying how meaning, emphasis, and tone shift when different systems engage with the same material.

  • How should these editions be read?

    Comparatively.
     
    Readers are encouraged to move between the baseline text and AI‑assisted editions to notice differences in voice, character focus, pacing, and thematic framing.

  • Why name the AI models explicitly?

    Because the system matters.
     
    Different models produce meaningfully different results. Naming the model used is essential for transparency and for interpreting why a particular version reads the way it does.

  • Are future editions already written?

    Some are in progress; others are planned.
     
    Each edition is produced and released deliberately, not in bulk, to preserve clarity and avoid conflating results.

  • Is this a research project or a creative project?

    Both.
     
    This site publishes readable editions while simultaneously documenting a methodology for comparing AI‑assisted interpretation.
     
    Neither goal works without the other.