Sources and methods

How source material becomes something readers can actually use.

This is the working method behind the site: collect puzzle material, keep the source trail visible, use AI where it helps, then edit the result like a real person writing for other enthusiasts.

01

Source capture

Source chapters, capture output, extracted images, and page ranges remain traceable in the local content pipeline.

02

Source boundaries

Each public guide separates what the source directly supports from my editorial explanation, inference, or modern mathematical framing.

03

Translation

Public pages use edited English adaptation. Machine capture and literal translation are working material, not finished public articles.

04

Creator edit

Each public guide is rewritten for curious English readers, with a human voice, short paragraphs, and useful context.

05

Visual explanation

Diagrams and interactions should clarify the puzzle object or mathematical idea. They should not serve as generic decoration.

06

Corrections

Factual, translation, attribution, and diagram errors should be corrected when reported, with the updated page date changed when appropriate.

07

AI assistance

LLMs may help summarize, compare terminology, draft explanations, and prototype interactions, but public claims cannot depend on unverified model output.

AI-assisted work

AI can help in the workshop. It does not get the final word.

Large language models may help translate, summarize, compare terminology, and adapt explanations for English readers. Their output is checked against the source text before publication, especially for dates, names, mathematical claims, and cultural attributions.

A native-style editorial pass can make the copy sound more natural, but it still has to respect the source. If a sentence becomes smoother by becoming less accurate, the sentence loses.