R — Original text (Reading)
Sample of 3,861 main posts, average engagement score 188.4, median only 42. Most posts perform ordinarily; a small number of posts carry most of the growth. Distribution on public platforms is inherently long-tailed. The real question is not how to make every post go viral, but how to systematically increase the probability that content enters the top 10%.
— Xiangyang Qiaomu, X Viral Secrets Sharing · The Truth About Growth: Long-Tail Distribution
I — Methodological skeleton (Interpretation)
This is a strategic mental model for public platform distribution, with a problem shift at its core.
Distribution on public platforms is naturally long-tailed, not normal—most people think "as long as the content is good, it should get traffic," but the data shows that among 3,861 posts the median is only 42 engagement, while the mean is pulled up to 188 by a few viral hits. This means: it is normal for the vast majority of posts to perform ordinarily, and that is not your failure.
Key shift:Replace "how do I make every post go viral" (an impossible question) with "how do I systematically increase the probability of entering the top 10%" (an engineerable question). This shifts the direction of effort from "gambling on a single viral hit" to "building a repeatable system"—choosing the right content archetype, matching posting timing, and strengthening evidence credibility are all levers for increasing probability, not formulas that guarantee virality.
It is also an anti-anxiety tool: a post that performs ordinarily is not a failure, but the expected median.
A1 — Application in the book (Past Application)
Case 1: Xiangyang Qiaomu's 3,861-post sample statistics
- Problem: Understand the true distribution pattern of content distribution on X, and answer "is it actually normal for a post to perform ordinarily"
- Use of the methodology: Use 3 years and 3.4G of self-owned data to calculate the engagement score (likes + 3× reposts) of 3,861 main posts, and plot the distribution
- Conclusion: Mean 188, median only 42, a typical long tail—a few posts carry most of the growth, and most posts performing ordinarily is a structural norm
- Result: The strategy shifted from "gambling on a single viral hit" to "systematically increasing the probability of entering the top 10%"; subsequent content choices (resource-entry type 51%, tool-tutorial type 39%) and data reviews were all organized around this goal
Case 2: Xiangyang Qiaomu's growth nodes (consecutive viral hits in 2026.1)
- Problem: In a long-tail distribution, how does growth actually happen—linear accumulation or node breakthroughs
- Use of the methodology: Observe the growth curve of one's own account and find that growth does not happen evenly every day, but triggers node-style breakthroughs when "units worth bookmarking appear consecutively"
- Conclusion: Under a long-tail distribution, growth depends on a small number of high-quality posts breaking out in concentration, rather than every post contributing evenly; the goal is to make this batch of "units worth bookmarking" appear consecutively
- Result: Consecutive viral hits in January 2026, with the account growing from 100 followers to 110,000—validating that "systematically increasing the probability of entering the top 10%" is more effective than "chasing a viral hit with every post"
A2 — Trigger scenarios (Future Trigger) ★
In what situations would a user need this skill?
- Feeling anxious because a post didn't blow up — After posting several pieces of content with very low engagement, they start doubting themselves: "Am I just not cut out for X?" "One post didn't blow up and I'm questioning my whole life."
- Evaluating overall performance after posting a batch of content — "I posted 20 in a row and only 2 broke 100. Should I give up?" They need to calibrate expectations using long-tail benchmarks.
- When setting growth goals — When the user asks "How do I make every post blow up?", the goal needs to be converted from an impossible question into an engineerable one.
- Feeling a gap when comparing with others' viral hits — Seeing others' viral hits makes them feel far behind; they need to understand that under a long-tail distribution, "other people's viral hits are also the minority."
Language signals (activate when these appear in the user's words)
- "I posted 20 in a row and only 2 broke 100. Should I give up?"
- "How do I make every post blow up" / "how to make every post go viral"
- "One post didn't blow up and I'm questioning my whole life" / "I've posted so many and nothing's getting any traction"
- "Am I just not cut out for X?" / "my posts get no engagement"
- "Why is my post engagement so low?" / "only 2 out of 20 hit 100, should I quit"
- "Others blow up with one post, but nothing works for me"
Distinction from adjacent skills
- Difference from
x-data-review: data-review is "how to review data to find the 80/20" (operational method), while this skill is "how to set growth goals and mindset" (mental model). Data review is the means; long-tail mindset is the premise—first accept the long tail, then use data review to optimize the system. - Difference from
x-foryou-algorithm: foryou-algorithm is "understanding how the recommendation mechanism works" (algorithmic knowledge), while this skill is "how to adjust mindset and goals in the face of distribution results" (strategic mindset). One is the underlying mechanism, the other is the strategic shift in facing results. - Difference from
x-content-archetypes: content-archetypes is "what type of content to choose to raise the distribution ceiling" (content selection means), while this skill is "accepting the premise that most content won't blow up" (mindset baseline). Archetype selection is one of the levers to raise probability; long-tail mindset is the premise for using the lever.
E — Executable steps (Execution)
Once the skill is activated, the agent should execute the following steps:
-
Redefine the problem: from "every post blows up" to "increase the probability of entering the top 10%"
- Explain to the user that public platform distribution is a long-tail distribution (3,861-post sample: mean 188 / median 42); most posts performing ordinarily is the norm
- Completion standard: The user can restate "one post not blowing up is the expected median, not a failure"
-
Calibrate current performance using long-tail benchmarks
- Have the user list the engagement data of a recent batch of posts (e.g., 10-20) and evaluate against the median of 42
- Completion standard: The user can judge whether their content is "within the normal long-tail range" or "systematically below the median"
- Stop condition: If the user only has data for 1-2 posts, skip to step 3 (the sample is too small to calibrate; directly explain the levers for increasing probability)
-
Point out the three levers for increasing the probability of entering the top 10%
- (1) Choose the right content archetype — resource-entry type (about 40% probability of entering the top 10%) and tool-tutorial type (about 24%) are far higher than ordinary expression type (about 9%)
- (2) Match the posting time — post tool tutorials at midday, opinion cases in the evening, and high-save content late at night
- (3) Strengthen evidence credibility — add visible (screenshots) / clickable (links) / calculable (numbers) evidence
- Completion criterion: The user can identify their current weakest lever and know the direction for the next optimization
-
Distinguish "median normality" from "systemic problems"
- Most posts performing averagely = long-tail normality, no adjustment needed; only when several consecutive batches (e.g., 2-3 weeks or more) all fail to reach the top 10% does it indicate the system needs adjustment
- Completion criterion: The user can distinguish "a single post not going viral (normal)" from "several consecutive batches with no viral hit (system needs adjustment)"
B — Boundary ★
Do not use this skill in the following situations
- Single-content production decisions — "Help me write a tweet about Claude Code" should go through x-short-content-craft / x-five-piece-checklist. The long-tail mindset is for evaluating the performance of a batch of content, not for writing a single piece.
- Specific data review operations — "Help me analyze which types had high engagement this week and how to adjust next week" should go through x-data-review. This skill provides a mindset baseline, not a data review operational process.
- Understanding recommendation algorithm mechanics — "How does the For You algorithm score things" should go through x-foryou-algorithm. This skill does not explain how the algorithm works.
Failure modes the author warns about in the book
- ce16 Survivorship bias — The long-tail distribution does not deny the existence of luck; it only adjusts the goal. "Systematically increasing the probability of reaching the top 10%" is not a deterministic statement that "if you follow this formula, you will go viral." Treating a probability increase as a guaranteed viral formula is another manifestation of survivorship bias.
- ce07 Obsessing over traffic anxiety — The long-tail mindset is a tool for easing anxiety, but if the anxiety reaches the point of "anxious if I don't post for a day, questioning life if one post doesn't go viral," it means you have been hijacked by traffic. At this point, you should not only adjust your mindset but also return to content ability itself — platform techniques will change, but content ability is what you can take with you.
The author's blind spots / era limitations
- The 3861-post sample comes from Xiangyang Qiaomu's personal account (AI/product research track, predominantly male Chinese-speaking audience, specific algorithm versions from 2023-2026), and the median of 42 may not apply to all tracks. Different tracks, different audiences, and different periods have different data distributions; directly applying the specific numbers is equivalent to applying one person's overfitted model to another person's data (ce15).
- The long-tail distribution is a statistical observation, not a causal law. "Systematically increasing probability" can raise expected value, but it cannot guarantee the result of any single post.
Adjacent methodologies easily confused with this one
- x-data-review(80/20 review) — The long-tail mindset is the premise of "accepting that most posts are average"; data review is the operation of "finding the 20% of posts that contribute 80% of the value and doubling down on them." First have the long-tail mindset, then do data review; the order cannot be reversed.
- x-content-archetypes(Four archetypes) — Archetype selection is one of the levers for increasing the probability of reaching the top 10% (Step 3), but the long-tail mindset itself does not involve specifically how to choose archetypes; it only provides the motivation for "why choose archetypes."
Related skills
- depends-on
x-data-review: Data review depends on the long-tail mindset coming first — first accept that most posts being average is normal (median 42), then use 80/20 to find the 20% to double down on. First have the mindset, then do the review; the order cannot be reversed. - composes-with
x-foryou-algorithm: Algorithm mechanics (why something gets recommended) + long-tail mindset (how to adjust when facing results); understand the mechanics to adjust your approach, then accept the long-tail results.
Audit information
- Verification passed: V1 ✓ / V2 ✓ / V3 ✓
- Test pass rate: To be tested (see test-prompts.json + test-results.md for details)
- Distillation time: 2026-07-14