Thought equivalence across languages

This method is adapted from Legion's current Taiwanese Mandarin transcreation practice. It is language-independent; a generated skill supplies the actual locale rules and usage references.

What remains invariant

A charitable, intelligent reader of the target edition should reconstruct substantially the same central thought, reasoning, intended implications, emotional movement, epistemic posture, authorial presence, and relationship with the reader.

Sentence resemblance, vocabulary coverage, and back-translation can expose particular mistakes. They do not establish that the same thought reached the reader.

Map the source thought

Before composing the target prose, identify the purpose and central proposition. Map how claims relate: an observation supporting an inference, premises supporting a conclusion, an example establishing a distinction, or an objection changing the scope of a claim. Include relationships the reader is expected to infer.

Preserve each important claim's status: observation, memory, inference, speculation, judgment, or deliberate uncertainty. A qualifier may govern a whole paragraph. Keep its scope rather than distributing cautious language across every sentence.

Record voice, intimacy, emotional movement, humor, profanity, rhythm, ambiguity, and deliberate friction. Identify source-language devices that need another expression to have the same effect. Note exact quotations and identifiers separately.

The map describes thought and reader effect. It is the basis for composition, not a draft of target-language sentences.

Compose and compare

Compose naturally from the map and voice brief, letting the target language determine sentence length, order, paragraphs, transitions, and rhythm. Cultural adaptation preserves the source's function within the latitude the author granted. Distinguish an authorized new explanation from an explanation the source actually contains.

Then reopen the source. Examine missing thoughts, invented claims, altered agency or causality, changed certainty, flattened voice, and language still governed by source syntax. Repair the target edition. Revisit the map when it misrepresented the source.

Independent thought reconstruction

For substantial work, give a fresh reader only the target edition and enough audience and genre context to read it naturally. Keep the source, thought map, source title, and the editor's concerns outside that reader's context.

Ask the reader to reconstruct the central proposition, reasoning, intended implications, confidence, normative commitments, emotional movement, voice, reader relationship, and deliberate ambiguities. Ask where the text becomes difficult to follow. Their job is to report what reached them, not propose a rewrite.

Compare that reconstruction with the source map. Material divergence identifies a problem even where individual sentence translations are defensible. A separate target-only usage pass can address locale, register, translationese, punctuation, terminology, and cultural fit.

Independence is a property of the reviewer's context, not a role label. Use available authorized review capabilities. If independent context is unavailable, perform the useful comparisons you can and record the missing review accurately.

Edit the fixed target prose

Once a complete draft is fixed and its generation has ended, edit it in a later invocation. Preserve the argument and voice while removing thinking tics and translationese. Use the installed thinking-tics method when available; preserve its draft/edit separation either way.

Make locally justified changes. A substantial rewrite becomes another draft requiring its own later edit. Accept a successful passage as written. A further review is useful when a material change warrants it, rather than merely because another review can be requested.

Preserve provenance

Keep source and edition revisions, language and locale, authorship, actual reviewers and methods, unresolved material questions, and publication status in the form appropriate to the work. Small labels need little machinery; a consequential essay benefits from its thought map and review decisions.

Credit human review only when a human performed it. Present AI authorship and review plainly. The result is ready when the intended thought and voice survive, the language fits its readers, exact material is intact, and the stated review history is true.