x-three-translations

Three Translations — Turning Internal Language into External Language

R — Original Text (Reading)

From internal language to external language. First translation: change the launch into help—"we launched a new feature" → "this feature can help you turn an 80-page report into a 3-page outline." Second translation: change capability into scenario—"supports long context" → "read an industry report in one go and spot competitor changes." Third translation: change conclusion into evidence—say less "works great," and include more real screenshots, input and output, steps, comparisons. The focus is "what others can take away," not "what I said."

— Xiangyang Qiaomu @vista8, X viral secrets sharing · Three Translations

I — Methodological Skeleton (Interpretation)

Creators naturally write in "internal language"—talking about what they did (launch), what the product has (capability), and how the results are (conclusion). These have meaning to the creator but give readers nothing to grasp.

The Three Translations are three steps of perspective shift, each moving the focus from "what I said" toward "what others can take away":

Launch → Help: Don't tell readers what you did; tell readers what this thing can help them do. "We launched a new feature" is a launch; "helps you turn an 80-page report into a 3-page outline" is help.
Capability → Scenario: Don't list product capabilities; embed them in a real task the reader will encounter. "Supports long context" is a capability; "read an industry report in one go and find competitor changes" is a scenario.
Conclusion → Evidence: Don't just say "works great"; give evidence readers can verify for themselves—screenshots, input and output, steps, before-and-after comparisons.

The Three Translations are not "making the wording smooth," but switching to the information receiver's perspective. The criterion: after reading, can the reader directly take away an action?

A1 — Applications in the Book (Past Application)

Case 1: Xiangyang Qiaomu's "announcement-style → help-style" rewrite

Problem: AI product/tool tweets easily get written as "we launched X feature," like posting an announcement, and readers don't know what it has to do with them.
How the methodology was used: Apply the first translation (launch→help) to "we launched a new feature," rewriting it as "this feature can help you turn an 80-page report into a 3-page outline." The focus shifts from "what we did" to "what you can do with it."
Conclusion: Announcement-style expression only conveys information; help-style expression conveys the possibility of action. Only the latter has spreadability.
Result: This case is used as a benchmark example of the Three Translations, helping readers understand the first conversion of "internal language→external language." In Xiangyang Qiaomu's 3,861-post data, posts with "actionable" signals (resources/steps/entry points) had a significantly higher probability of entering the top 10%.

Case 2: X's official Article guide "Show, don't just tell"

Problem: In its Article writing guide, X officially points out that a common mistake creators make is only stating conclusions ("works great") without giving evidence.
How the methodology was used: The official principle "Show, don't just tell" is proposed—for any claim, immediately follow it with evidence (data, personal stories, before/after images). This is essentially the official version of the third translation (conclusion → evidence). Original guide text: "For any claim you make, follow it immediately with evidence of why it's true (stats, personal story, before/after, etc.)".
Conclusion: The official guide and Xiangyang Qiaomu's three translations independently validate the same principle—conclusions must be paired with evidence.
Result: This guide was released as the standard method for X official Article writing, targeting all Premium users.

Case 3: The second translation of "long context feature" (capability → scenario)

Problem: The user asks, "How do I turn 'we launched a long context feature' into a strong tweet?"
Use of the methodology: The second translation (capability → scenario)—embed the capability "supports long context" into the reader's real task: "read an industry report in one go and identify competitor changes." Then connect to the third translation (conclusion → evidence): include before/after screenshots showing the efficiency difference before and after using long context.
Conclusion: Capability is the product perspective; scenario is the reader perspective. Readers do not pay for capabilities; they pay to solve their own problems.
Result: This is a new question constructed during the V2 validation phase, proving that the three-translation framework can handle variants beyond the original cases.

A2 — Trigger scenario (Future Trigger) ★

In what situations would users need this skill?

There is already a draft but it reads like an announcement: The user wrote a tweet/product launch copy whose content is "we released X/support Y/upgraded Z," feels no one will retweet it, and wants to make it more attractive.
Feature introduction is not attractive: The user wants to write a product capability as a tweet, but it comes out like a feature list, and they do not know how to make readers feel "this is relevant to me."
The tweet got no response, self-checking the reason: The user posted a tweet with very low engagement and suspects it is a problem with the way it is expressed (the actual reason is that it only talks about themselves and not the reader).
Turn "the effect is very good" into credible content: The user wrote conclusive expressions such as "the effect is very good/very powerful/the experience is excellent" and needs to add evidence.
Rewriting tweets for long context/new feature launches: The user wants to translate technical feature descriptions into scenarios readers can perceive.

Language signals (activate when these appear in the user's words)

"This tweet reads like an announcement / like a press release / like a product update note"
"How do I turn 'we launched X' into a strong tweet / make it attractive"
"How do I write a feature introduction / how do I write a feature so people will read it"
"The effect is very good but no one responds / why is no one retweeting"
"This passage is too much like self-indulgence / too much like talking to oneself"
"announcement tone / feature list / show don't tell / translate to reader value"
"how to make this less like a product announcement"

Distinction from adjacent skills

Difference from x-four-saves : The four-province model evaluates "whether this is worth posting" (valuation), while the three translations rewrite "already decided to post but the expression is wrong" (rewriting). The four provinces are screening before posting; translation is expression optimization after screening.
Difference from x-five-piece-checklist : The five-piece set checks whether the content is complete (all five elements present), while the three translations focus on the language quality of two of them, "value promise" and "evidence." The five-piece set is a structural check; translation is language conversion.
Difference from x-short-content-craft Difference: The Short-Form Content Workshop drafts from scratch (Hook-Body-CTA structure selection), while Three Translations performs a perspective shift on an existing draft. The former solves "how to write," the latter solves "what if it comes out sounding like an announcement."
vs. x-content-archetypes Difference: The Four Archetypes determine "what type of content to post," while Three Translations determines "how to translate the language of the same type of content from internal to external."

E — Executable Steps (Execution)

After the skill is activated, the agent should execute the following steps:

Identify announcement-style expressions (internal language)
- Scan the user-provided draft and mark three types of internal language signals: release-style ("we released/launched/upgraded"), capability-style ("supports X/has Y feature"), and conclusion-style ("the effect is great/very powerful").
- Completion standard: Mark at least 1 instance of each type (if present), or clearly tell the user "the current draft has no announcement-style expressions and does not need Three Translations."
- Stop condition: If there are no announcement-style expressions in the draft, skip to step 4 and directly output "the draft is already in external language and needs no translation."
Perform three translations sentence by sentence
- Perform the corresponding translation for each marked point:
  - Release→Help: Ask "What can this help the reader do?" and rewrite "we did X" as "this can help you do Y."
  - Capability→Scenario: Ask "In what real task would the reader use this capability?" and embed the feature description in a concrete task.
  - Conclusion→Evidence: Ask "How can the reader verify this conclusion themselves?" and replace "the effect is great" with screenshots/numbers/steps/before-and-after comparisons.
- Completion standard: Every marked point has a corresponding translation result, and the translated content focuses on "what the reader can take away."
Verify translation quality and output the rewritten result
- Self-check the translated content: After reading it, can the reader directly take away an action or a verifiable piece of evidence? If not, return to step 2 and retranslate.
- Output: Original → translated comparison (indicating the type of each translation), and provide the final rewritten version.
- Completion standard: The output includes a comparison table + final version, and the final version contains no remaining announcement-style expressions.
(Optional) Provide suggestions not covered by the translation
- If you find that the draft has structural issues in addition to language issues (such as a missing Hook/CTA), remind the user that they can further use x-short-content-craft or x-five-piece-checklist。
- Completion standard: Only remind; do not overstep by performing the work of other skills.

B — Boundary ★

Do not use this skill in the following situations

When drafting new content from scratch: Three Translations is a rewriting tool for an existing draft, not a drafting tool. To start from scratch, use x-short-content-craft to choose the type and structure.
When the content itself is already in external language: If the draft is already written from the reader's perspective (with scenarios, evidence, and help-style expressions), no translation is needed. Forced translation will over-embellish it.
Pure information lookup/news reporting: Some content is naturally an announcement (such as official press releases or version update logs) and does not need to be translated into help-style expressions—its purpose is simply to convey facts.

Failure modes the author warns about in the book

ce18 (low-quality Quote "too strong, so awesome"):Low-quality Quote contains only exclamations with no substantive content; essentially, it has not done the third translation (conclusion→evidence)—it only says "so strong" (conclusion) but provides no incremental evidence (screenshots/steps/comparisons). Readers see it as traffic-farming and report and block it. The value of a Quote lies in adding incremental viewpoints; a Quote with no increment is equivalent to parasitic noise.

The author's blind spots / era limitations

The cases of the three translations mostly come from the AI/product research track (tool recommendations, feature releases). In other tracks (such as purely story-driven or emotional content), the third translation (conclusion→evidence) does not necessarily apply—the power of story-driven content lies in resonance rather than evidence.
Xiangyang Qiaomu's experience is based on personal 3.4G of data, which carries a risk of overfitting—not all "announcement-style expressions" need translation; audiences in some fields really do need an official tone (such as B2B and enterprise clients).

Adjacent methodologies that are easily confused

"Popularized expression":The three translations are not about translating professional terminology into plain language (that is dimensionality reduction), but about switching the perspective of the information recipient (from creator to reader). The two can be done at the same time, but they are not the same thing.
"Show don't tell":This is a subset of the third translation (conclusion→evidence), covering only one step of the three translations. The three translations also include two perspective shifts: publishing→helping and capability→scenario.

Related skills

composes-with x-five-piece-checklist: The three translations rewrite language quality, and the five-piece set checks structural completeness. The two work together—after translating, use the five-piece set to verify whether the evidence satisfies the "three can" criteria.

Audit information

Verification passed: V1 ✓ / V2 ✓ / V3 ✓
**Test pass rate: to be tested (see test-prompts.json)
Distillation time: 2026-07-14