x-four-saves

Four-Saves Model — Content Valuation

R — Original Text (Reading)

"The value of content lies not in how much information you convey, but in how many steps you save the reader. 1. Save searching: readers don't have to fish for an entry point in a sea of information; 2. Save understanding: readers don't have to chew through complex concepts themselves; 3. Save trial and error: readers don't have to step in every pit themselves; 4. Save expression: readers can forward this directly to others."

— Xiangyang Qiaomu, X爆款秘籍分享.md · Four-Saves Model

I — Methodological Skeleton (Interpretation)

The Four-Saves Model is a content valuation tool, not a writing template.

It shifts the evaluation perspective from creator-centered ("how much information did I say") to reader-centered ("what can the reader skip"). Whether a piece of content is worth posting does not depend on the amount of information, but on whether it saves the reader at least one of the following four steps:

Save searching — Readers don't have to find the entry point themselves in a vast amount of information; you give it directly.
Save understanding — Readers don't have to chew through complex documents or concepts; you have already translated them into words they can understand.
Save trial and error — Readers don't have to step in every pit themselves; you provide the steps, path, or pitfall-avoidance guide.
Save expression — Readers can forward this directly to others without having to reorganize the language themselves.

Core judgment: content that saves not even one step is not worth posting. The more steps saved and the rarer they are, the higher the content value. Spending 30 seconds on this self-check before posting is more efficient than waiting for data after posting.

A1 — Application in the Book (Past Application)

Case 1: Feishu Museum Document Post (Save Searching)

Problem: How do you turn a scarce resource into highly spread content?
Use of the methodology: Xiangyang Qiaomu shared a resource post that organized 1.55 million items from museums worldwide into a Feishu document. What readers got was a directly usable entry point, without having to scrape the websites of 155 museums one by one — a typical "save searching."
Conclusion: By saving readers the searching step, the content has posting value.
Result: 263,000 views, a representative case of the resource-entry type with a median interaction of 2,965.

Case 2: Cat Catch Chrome Extension Post (Save Searching + Save Trial and Error)

Problem: How can a tool recommendation post get both views and usage?
Use of the methodology: Shared the Cat Catch extension with 600,000 installs, emphasizing the specific scenario of "automatically sniffing audio and video downloads" and the installation entry point. Readers don't have to search for the extension themselves or try out which one works — saving both searching and trial and error.
Conclusion: Content that saves two steps (searching + trial and error) has more spread power than content that saves only one.
Result: 207,000 views, 286 reposts, 1,319 bookmarks.

Case 3: ClaudeCode Beginner's Guide Post (Save Understanding)

Problem: How can a technical tool guide attract non-technical readers?
Use of the methodology: Shared a Claude Code usage guide written by a non-programmer, emphasizing "written from a beginner's perspective" and "even people who don't know how to code can understand it." Readers don't have to slog through the official docs—saves comprehension.
Conclusion: Translating complex concepts into words beginners can understand is saving comprehension; the content is worth posting.
Results: 214,000 views, 254 reposts, 1,319 bookmarks.

A2 — Trigger Scenarios (Future Trigger) ★

In what situations will users need this skill?

Value self-check before posting — After writing a piece of content (or about to post one), hesitating over "is this worth posting" or "will anyone read it."
Comparing multiple options — Having several topics in hand, unsure which is more worth doing, and needing to compare value using a unified standard.
Reviewing "why no one responded" — Posted something with poor results, wanting to determine whether it "didn't save a single step" or "saved steps but didn't get the message across."
Evaluating others' viral hits — Seeing a viral hit, wanting to break down why it took off, using the Four Savings framework to analyze its source of value.
Deciding content investment level — A topic looks like it could be written at length, but the input-output ratio is uncertain; first self-check how many steps it saves.

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

"Is this worth posting / should I post it"
"Does it have content value / does this have value"
"Why is no one reading it / why is no one responding"
"How many steps does it save / what does it save readers"
"worth posting / content value / should I post this / save steps"

Distinction from adjacent skills

Difference from x-five-piece-checklist : The Five-Piece Set checks the structural completeness of a tweet (whether the value promise/scenario/entry point/evidence/reason to bookmark are all present); the Four Savings model checks whether this piece of content is worth existing (how many steps it saves readers). The former is "is it complete," the latter is "is it worth it." When the user says "help me check if it's complete" → Five-Piece Set; when the user says "is it worth posting" → Four Savings.
Difference from x-content-archetypes : The Four Archetypes classify by "what readers can do with it" (resource entry type/tool tutorial type/AI tool discovery type/ordinary expression type), predicting the spread ceiling; the Four Savings model is a valuation check (how many steps saved), deciding whether it's worth posting. The former answers "which type is this," the latter answers "should it be posted."
Difference from x-three-translations : The Three Translations is a rewriting tool (translating announcement-style internal language into external language readers can take away); the Four Savings model is a valuation tool (doesn't rewrite content, only judges value). The former is "how to change it," the latter is "is it worth it."

E — Executable Steps (Execution)

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

Self-check the Four Savings item by item
- For the content to be evaluated, answer one by one: (a) Does it save readers search? (provides entry points/resources that are usually hard to find) (b) Does it save readers comprehension? (translates complex concepts into understandable words) (c) Does it save readers trial and error? (provides steps/paths/pitfall-avoidance guides) (d) Does it save readers expression? (readers can directly forward it to others)
- Completion standard: All four items have a clear "yes/no" judgment, no ambiguity allowed. If "yes," note what it saves.
Count which steps are saved, judge whether it's worth posting
- Saves 0 steps → not worth posting, recommend giving up or redoing.
- Saves 1 step → borderline value, see whether this step is scarce enough (if others can easily provide it too, it's not worth it).
- Saves 2 steps or more → worth posting, and the more steps saved, the greater the propagation potential.
- Completion criteria: Give a clear "post / don't post / post after revision" conclusion, and mark which steps are saved.
- Stop condition: If the content saves 0 steps and the user has no willingness to revise, directly recommend not posting, and skip to step 3 to give the reason for abandoning it.
Give revision suggestions (if needed)
- If the judgment is "don't post" but the content has a foundation, point out which step to add to turn it around: missing search → add a scarcity entry point; missing understanding → add beginner translation; missing trial and error → add concrete steps; missing expression → add a quote that can be forwarded directly.
- Completion criteria: Revision suggestions must be specific enough to say "what content to add to save step N," not vaguely say "write it better."

B — Boundary ★

Do not use this skill in the following situations

When it has already been decided to post and only structural completeness needs to be checked — use x-five-piece-checklist to check whether the five-piece set is complete; the four saves do not solve structural problems.
When content needs to be classified into which archetype it belongs to — use x-content-archetypes for classification; the four saves do not classify.
When the expression needs to be rewritten — use x-three-translations for internal→external language translation; the four saves do not rewrite.
When reviewing data performance — use x-data-review for 80/20 review; the four saves are pre-post valuation, not post-post review.

Failure modes the author warns about in the book

Applying the four saves as a mechanical metric (ce15): Xiangyang Qiaomu himself notes "at first glance it seems a bit overfit, for reference only." The four saves are an inspection tool, not dogma — don't turn "saves four steps" into a hard metric that must be filled to four. As long as a piece of content saves a scarce step (e.g., the Feishu museum only saved search and got 260,000 views), it is worth posting. Forcing all "four steps saved" instead dilutes the content.
Treating overfit experience as a universal formula: The four-saves framework itself is transferable, but the author's specific conclusions based on personal 3.4G data (e.g., that a certain type of content spreads better) are overfit results from his personal sample. Different tracks/audiences/periods have different distributions, and the thresholds cannot be copied directly.

The author's blind spots / era limitations

Survivorship bias: The four-saves model comes from summarizing hits by successful people, and lacks a control group — posts that also "saved search" but did not go viral are not included. The four saves are a necessary-condition reference, not a sufficient-condition guarantee.
Correlation is not causation: "Saving steps" and "high interaction" may only be correlated, not causal. Whether content can go viral is also affected by uncontrollable factors such as timing, algorithms, and audience size.

Adjacent methodologies that are easy to confuse

x-five-piece-checklist (five-piece set): The most easily confused. Distinguishing mnemonic: ask "is it worth it" with the four saves, ask "is it complete" with the five-piece set. The four saves are the valuation threshold (pass or not), and the five-piece set is the structural audit (complete or not).
x-content-archetypes (four archetypes): The four archetypes tell you "which type this belongs to and what its propagation ceiling is"; the four saves tell you "whether this is worth existing." First use the four saves to judge whether it is worth posting, then use the archetypes to judge how far it can go once posted.

Related skills

depends-on x-five-piece-checklist: The five-piece set's check of structural completeness depends on the four saves' prior valuation — first use the four saves to judge "whether it is worth posting," then use the five-piece set to check "whether it is complete." The four saves are the valuation threshold, and the five-piece set is the structural audit.

Audit information

Verification passed: V1 ✓ / V2 ✓ / V3 ✓
Test pass rate: 100% (fallback self-assessment, independent blind testing recommended) (see test-prompts.json for details)
Distillation time: 2026-07-14