[•]STATUR.
Dossier · by Dr. Louis Korczowski

How Ideas Go Viral on Social Media: We Mapped the 356M-View France vs Japan Meme War on X

327 viral posts in eight languages, read one by one with their images and videos. 18 beliefs in conflict, the camp each one served, the duels they fought, the third parties who joined, and six steps to map your own idea war.

Ask your AISummarize this playbook and apply it to your situation: ChatGPTClaudeMarkdownAI skills on GitHub
TL;DR

On 17 October 2026, between 550,000 and 800,000 people marched in France for education funding and purchasing power, and the evening ended in looting in several cities. A One Piece flag and a Naruto cosplay in the protests set off a war between French-speaking and Japanese-speaking X. We read 327 viral posts one by one, in eight languages, with their images and videos (356.5M views), and tagged each one with the belief it carries and whether it supports or attacks it. In Japan, the riots became an immigration debate: 40% of the Japanese-side views went to beliefs such as "It's a Muslim invasion". A post citing the Justice Ministry's arrest data beat every reply that defended that belief, and the belief still reached about 9 times more people. The French side answered with stereotypes about Japan, mostly as jokes. 23% of the views came from neither camp. Below: the map, the duels, the third parties, and six steps to map your own war of beliefs.

327viral posts read one by one, images and videos included
356.5Mviews, posts from 06 to 26 October 2026
18beliefs in conflict, induced from the posts
23%of the views posted by neither camp
The map

Where each idea stands.

Each bubble is an idea. Its size is the views it earned, its ring the share serving each camp. Battle lines put French-side ideas on the left and Japanese-side ideas on the right. Idea space puts ideas that mean similar things close together. Lines are quotes: one idea answering another, dashed orange when it attacks.

Serves the French sideServes the Japanese sideNeutral or third party
The origin

How one cosplay started a war of ideas.

Most ideas on the map trace back to four posts written on 18 and 19 October.

DateWhat happenedReach
17 OctBetween 550,000 and 800,000 people march in France for education funding and purchasing power. The evening ends in looting in Paris, Lyon, Marseille and other cities.
18 OctA French post shows a Naruto cosplayer (the character Obito) posing at a burning car, captioned as an anime fire attack.about 2.3M
18 OctA Japanese account answers: stop soiling Japanese culture to do stupid things.2.3M
18 to 19 OctFrench answers: withdraw every anime inspired by Europe, and a list of Japanese "vices" (rape culture, pedophilia, suicide, submission to the US) that becomes the script of the jokes to come.5.1M and about 2.3M
19 to 24 OctJapanese X frames the riots as a Muslim invasion (a few posts already pushed it before the riots). On 23 October a Japanese post cites the Justice Ministry note: 70% of those jailed are French nationals, nearly all born in France.31.3M for the belief
21 to 25 OctThe French joke wave (a joke claiming the anime hero Vegeta is Algerian: 17.9M). X auto-translates every post, and spectators from other languages join.172.8M on the French side in those five days
The beliefs

The 18 beliefs in conflict, ranked by views.

Each belief is written as something a person can believe, whatever the packaging: a joke, an argument or a statistic can carry it. Serves: the share of its views that served each camp. Stance: the share of its views that supported or attacked it. Elo: the rating earned in quote duels between beliefs (section below), 1,500 at the start.

BeliefMost viewed post (paraphrased)ViewsPostsServesStanceElo
The Japanese are pervertsJapanese sexuality is so deviant that even a sperm cell ends up in a sex doll62.8M37FR 85% · JA 15%Supports 85% · Attacks 15%1474 (1-3)
Don't use our manga for your riotsVegeta is Algerian: the diaspora claims anime heroes as its own57.8M49FR 50% · JA 43% · Neutral 6%Supports 43% · Attacks 53%1562 (5-1)
It's a Muslim invasionVegeta eats a whole pig, so he cannot be Algerian (Muslims do not eat pork)31.3M27FR 13% · JA 76% · Neutral 11%Supports 76% · Attacks 24%1591 (8-1)
France won the meme warOkay France you win, calm down29.2M40FR 86% · JA 6% · Neutral 8%Supports 86% · Attacks 6%1483 (0-1)
The Japanese are suicidalA Japanese hiker whose feet do not touch the ground: suicide by hanging as a punchline25.9M11FR 100%Supports 100%1531 (2-0)
The protesters are just lootersA lone French otaku defends an anime shop from 'invaders'22.2M12JA 99%Supports 99%-
Auto-translation turned X into a battlefieldAuto-translation revealed what Japan really thinks17.4M17FR 42% · Neutral 58%Supports 100%1468 (0-2)
France is collapsingFrance is collapsing: luxury stores looted in daylight, this is multiculturalism15.6M15JA 89% · Neutral 11%Supports 89%1484 (0-1)
Japan's kindness is a facadeJapan was never as friendly as its fiction: it took 17 October for you to notice15.2M18FR 54% · JA 18% · Neutral 29%Supports 54% · Attacks 35%1487 (1-2)
No belief: why are they fighting?Brazilians, Thais and French all mock Japan, but East Asian netizens look even meaner14.8M16FR 34% · Neutral 66%--
The Japanese are submissiveThe Japanese worker thanks his boss for leaving 3 minutes early from a 65-hour week14.8M10FR 100%Supports 100%1454 (1-5)
Japan started itFrench speakers swarmed a Japanese voice who asked them not to misuse Naruto, like sharks smelling blood8.7M11FR 14% · JA 68% · Neutral 18%Supports 14% · Attacks 72%1542 (4-1)
The media hide the truthA video call with a French political influencer to say what Japan's old media will not7.5M11JA 100%Supports 100%-
Japan never paid for its warThe Japanese must have no history classes at school6.6M8FR 100%Supports 95%1517 (2-1)
The whole world is against JapanKorean users mock Japan yet side with it against Muslims: a Japan-Korea truce6M7FR 39% · JA 61%Supports 39% · Attacks 61%1484 (0-1)
The rioters are not real FrenchFrance's multicultural crew raised on manga beats a Japan obsessed with whiteness5.9M10JA 95%Supports 95%1487 (2-3)
The Japanese are racistJapanese tweets: 'you are black'; French tweets: devastating jokes5.4M14FR 86% · JA 14%Supports 86% · Attacks 14%1456 (0-3)
Nobody beats the French at trollingSomeone tell the French there is an ammunition crate here4.9M9FR 95% · Neutral 5%Supports 100%-
The French are cowardsFrench people when the Nazis attacked: France surrendered fast3.3M3JA 100%Supports 100%1481 (0-1)
No belief: racial insultRacial joke about Japanese football players1M2FR 100%--
What made beliefs travel

Seven lessons from the battlefield.

Each lesson is read from the map and the duels. They describe this war, and they are the hypotheses to test on the next one.

01

In Japan, the riots became an immigration debate.

40% of the Japanese-side views (50.8M) went to four beliefs: it's a Muslim invasion (23.8M), France is collapsing, the media hide the truth, the rioters are not real French. The debate ran alongside Japan's own: on 1 October 2026 Japan tightened its permanent residency rules (Jiji).

02

The facts won the duels and lost the reach.

A Japanese post citing the Justice Ministry note on those jailed after the riots (70% French nationals, nearly all born in France, France 24) reached 2.7M. Every reply that defended the invasion story reached fewer people: five duels, five wins for the data. Posts supporting the belief still earned 23.8M, about 9 times more.

03

Racism was answered with stereotypes.

Half of the French-side views (93.8M) went to three beliefs about Japan: the Japanese are perverts (53.1M), suicidal (25.9M) and submissive (14.8M). French-side posts accusing Japan of racism earned 4.7M. Each stereotype fits in one image a stranger understands in a second, with or without translation.

04

A manga plea met a wall of mockery.

"Don't use our manga for your riots" earned 25M on the Japanese side. French mockery of it earned 29M and won 11 of 12 direct duels. A camp is not one voice: Japanese-speaking critics of the plea earned 1.5M, one noting that Japan got angrier at a One Piece flag than at a Japanese robbery gang that called itself Luffy.

05

The joke is the packaging, the belief is the cargo.

The same belief travels as a joke and as an argument. Supporters of "It's a Muslim invasion" earned 11.9M with jokes and 11.6M with arguments. Counting formats tells you how a belief travels. Counting beliefs tells you what people were asked to believe.

06

Third parties decide who looks like the winner.

23% of the views came from neither camp. English-speaking spectators quoted the French jokes with verdicts that France had won, adding 20.7M views of verdicts. One of the most viewed French-side jokes (7.6M) was posted by an English speaker who says he was posing as French.

07

The platform is a third party.

X translated every post in both directions. Spectators made the translate button the punchline, and other national fights ran the same week (Turkey vs China, Saudi Arabia vs Argentina, Lebanon vs Algeria): 17.4M views for the belief that auto-translation turned X into a battlefield.

The duels

When a post answers another.

A quote is a duel: someone takes a post and answers it. The ratio is the reach of the answer divided by the reach of the original. Above 1, the answer won. Duels inside a belief pit support against attack. Duels between beliefs are rolled into an Elo rating per belief, as in chess.

Original ideaAnswered withRatio
Vegeta is Algerian: the diaspora claims anime heroes as its own
@liistrk · French · 17.9M
Vegeta eats a whole pig, so he cannot be Algerian (Muslims do not eat pork)
@monokabu_g0628 · Japanese · 8.3M
0.47
Japanese sexuality is so deviant that even a sperm cell ends up in a sex doll
@BWhey69 · French · 11.5M
Grok, which country commits more child rape per capita, France or Japan?
@Jonathan_H07 · English · 494K
0.04
The Japanese worker thanks his boss for leaving 3 minutes early from a 65-hour week
@namekoisvnr · French · 10.1M
French people when the Nazis attacked: France surrendered fast
@in9hvozgfZcP5IP · Japanese · 1.5M
0.14
An American friend explains that Muslim immigrants carry the most radical views
@amatsuuni · Japanese · 4.8M
Auto-translation revealed what Japan really thinks
@fsooll5 · Arabic · 3.1M
0.64
France recorded more rapes in 5 years than Japan in 60
@ShitpostRock · English · 4.8M
Because in Japan it is legal: answer to a rape statistic, implying child abuse is legal
@maai_as_ · French · 1.4M
0.28
Hiroshima's bomb was named Little Boy: atomic trauma and pedophilia merged in one joke
@pomleloupWP0 · French · 5.5M
More people agree with the French, but that is not strength
@jjdtajaga · Japanese · 195K
0.04
How to read it. Most answers reach fewer people than the post they answer, so a ratio of 0.47 is a strong showing: the pork answer to the Vegeta joke reached 8.3M people. In the full data, only one rebuttal beat its original: an insult that out-reached the Japanese plea to stop soiling Japanese culture (ratio 1.06). It is counted in the ratings and left out of this table, like every insult.
The third parties

23% of the views went to posts by neither French nor Japanese speakers.

Views of the posts we read, by the community that posted them (language and self-identification). Communities are not nationalities.

CommunityViewsShareWhat they brought
French144.7M41%The jokes, and a minority that defended Japan.
Japanese130.5M37%The manga plea, the Muslim invasion belief, the correction by data.
English57.4M16%The jury: verdicts, amplification of the French jokes, a few pro-Japan statistics.
Arabic18.3M5%Jokes about Japan's submission to the West, the broader translation world war.
Spanish2.8M1%Spectators, and a statistic on the Japanese side of the rape debate.
Thai2.5M1%Explainers and translated best-of lists.
Portuguese163K0%Old World Cup grudges.
Chinese11K0%War-crimes retorts.
The playbook

Map your own war of beliefs in six steps.

The method behind this page, for a meme war, a political campaign or a brand crisis. A language model can do most of the reading. A human checks the beliefs and every uncertain post.

01

Collect the posts that carried the reach

Views decide what enters the corpus: 500,000 and more, 250,000 for third-party languages.

Do this
  • Search the platform in every language with a minimum-likes filter and a start date.
  • Open the Quotes tab of the 10 to 15 most viewed posts and keep the most viewed quotes.
  • Log every exclusion with its reason.
For your AI

Copy this prompt into your assistant:

I am mapping the ideas of an online controversy: [describe it, start date, camps, languages]. Build a corpus plan: search queries per language with a minimum-likes filter, the quote pages to open, the fields to log for each post, and the exclusion rules. Use the method at https://statur.co/how-ideas-go-viral.md.

Skill: statur-idea-war-map · install: npx skills add lkorczowski/content-strategist-AI-cookbook --skill statur-idea-war-map

02

Read every post with its picture

Write the implied idea: the joke lives in the image.

Do this
  • Look at the image or a few frames of the video.
  • Write a literal caption, then the implied idea in one neutral sentence.
  • Give a confidence from 1 to 3 and send every 1 to a human.
For your AI

Copy this prompt into your assistant:

Here are [N] posts with their images: [paste or attach]. For each, write a literal caption, the implied idea in one neutral sentence (the claim the post makes, including the meaning of the joke), and a confidence from 1 to 3. Never quote insults.

Skill: statur-idea-war-map · install: npx skills add lkorczowski/content-strategist-AI-cookbook --skill statur-idea-war-map

03

Name the beliefs

A belief is what people believe. The joke is only the packaging.

Do this
  • Group the implied ideas into 15 to 25 beliefs, each written so a person can say "I believe that...".
  • Test every label: a joke, a serious post and a debunk can all carry it; one post can support it and another attack it; the label names no format (joke, meme, verdict).
  • Keep posts with no belief (questions, pure insults) off the map, and have a human edit the list before measuring.
For your AI

Copy this prompt into your assistant:

Here are [N] implied ideas: [paste]. Propose 15 to 25 beliefs, each written as something a person can believe ("The Japanese are suicidal", never "Suicide jokes"). Check each label: can someone say "I believe that"? Can a joke and a serious post both carry it? Can one post support it and another attack it? Then assign every idea to one belief with a stance: supports, attacks or neutral. Flag the ideas that fit none.

Skill: statur-idea-war-map · install: npx skills add lkorczowski/content-strategist-AI-cookbook --skill statur-idea-war-map

04

Label camp, community and relation

The camp is the side the idea serves, read from the post. Never infer anyone's origin.

Do this
  • side: the camp served, or neutral.
  • stance: supports, attacks or neutral toward the belief.
  • register, the packaging: humor, argument, testimony, data.
  • For quotes: rebut, amplify, verdict or restate.
For your AI

Copy this prompt into your assistant:

For each post in this table [paste], fill side (camp served or NEU), stance toward its belief (supports, attacks, neutral), community (language and self-identification), register (humor, insult, argument, info, testimony, meta) and, for quotes, the relation to the original (rebut, amplify, verdict, restate).

Skill: statur-idea-war-map · install: npx skills add lkorczowski/content-strategist-AI-cookbook --skill statur-idea-war-map

05

Measure reach, duels and Elo

Ratio = views of the answer divided by views of the original. Above 1, the answer won.

Do this
  • Sum views per belief, per camp and per stance.
  • Compute the ratio of every rebuttal: inside a belief, count support against attack; between beliefs, roll them into an Elo rating.
  • Compute the share of views posted by third parties.
For your AI

Copy this prompt into your assistant:

Run the idea map script of the statur-idea-war-map skill on my annotations.csv and beliefs.csv, then explain: which beliefs carried the reach, whether supporters or attackers won inside each belief, which beliefs win duels, and which ones belong to third parties.

Skill: statur-idea-war-map · install: npx skills add lkorczowski/content-strategist-AI-cookbook --skill statur-idea-war-map

06

Map it and report

Battle lines for the story, idea space for the meaning, a table for the record.

Do this
  • Place beliefs by camp share (battle lines), sized by views.
  • Optionally place them by meaning: sentence embeddings, then MDS.
  • Report: who won the attention, which beliefs carried it, the duels, the third parties, the origin, the limits.
For your AI

Copy this prompt into your assistant:

Using map.json from the previous step, write a one-page report in this order: who won the attention, which ideas carried it, the five biggest duels, the third parties, the origin of the conflict, and the limits of the method.

Skill: statur-idea-war-map · install: npx skills add lkorczowski/content-strategist-AI-cookbook --skill statur-idea-war-map

Method and limits

How this was measured, and what it cannot say.

Corpus

59 seed posts (every tracked post above 500K views, plus posts found by hand), 146 posts found by searching X in eight languages (six returned new posts), posts quoted by the seeds, two origin posts seen in screenshots, and 115 quotes read in the Quotes tab of the 15 most viewed posts. 327 posts, 356.5M views. Twelve viral posts were excluded after a relevance audit: off topic or camp unclear. Every post shown on the live dashboard is vetted the same way.

Reading

Every post was read with its image or video frame, because the joke often lives in the picture (MemeCap shows that vision-language models still struggle with visual metaphors, Hwang and Shwartz, 2023). The beliefs were induced from the posts, following concept induction with language models (LLooM, Lam et al., 2024).

Map

Ideas are embedded with a sentence model; belief centroids are placed by metric MDS on cosine distance, like the intertopic map of LDAvis. Battle lines place each belief by the share of its views serving each camp.

Can it be automated?

Partly. Clustering raw post text with a multilingual sentence model recovers almost none of the beliefs (adjusted Rand index 0.10). Clustering the implied ideas reaches 0.19. The rest takes reasoning about the image and the joke, the step topic models such as BERTopic skip and prompt-based methods such as TopicGPT start to cover.

Limits. Seed posts were partly chosen by hand and favour the most viral posts. Views measure reach. Agreement is another question. One annotator so far, a language model reading each post with its image, reviewed by the author: a second human reading of 10% of the posts is planned. A camp is a reading of a post. We never assume the origin or religion of a person. Post text is never displayed: ideas are paraphrased, and insults are counted but not linked. Full method and tests: preprint. Data and code: GitHub.
For your AI

The same method, written for your assistant.

Skills in the open Agent Skills format, maintained on GitHub so every update reaches you. Upload them in Claude, paste them in ChatGPT, or install them with one command.

Map the beliefs of an online controversy: corpus, multimodal reading, belief naming with a four-question test, stance, camps, quote duels, Elo and the battle-lines map.

Rules every skill follows. Drafts only: the AI never sends or publishes for you. No passwords: you sign in yourself. No promised views.
Live data

The war goes on, and so does the measure. Follow the daily tug of war dashboard, updated every morning from the X API, read the preprint, and download the open data and code on GitHub.

Questions

How ideas go viral, answered.

How do things go viral on social media?

In this war, posts went viral when they compressed one idea into one image a stranger understands in a second: a joke built on a stereotype, a frame that confirms a fear, or an old historical wound. Reach also came from people outside the fight: 23% of the views went to posts by neither camp, many of them spectators judging who was winning.

What are memetic warfare examples?

The French vs Japanese war on X in October 2026 is a public example: two language communities fought with memes and reframes after a Naruto cosplay appeared at a riot. Each side reused a small set of beliefs ("The Japanese are perverts", "It's a Muslim invasion"), carried by jokes as much as by arguments, and third parties amplified or redirected them.

How do you measure which idea wins?

Read the quotes. When a post answers another with a different idea, divide the views of the answer by the views of the original. Above 1, the answer won the duel. Rolling every duel into an Elo rating per belief shows which beliefs hold up when they are attacked, which is different from which ideas earn the most views.

Did X auto-translation change the war?

It let everyone follow it. Spectators in English, Arabic, Spanish and Thai read the posts in their own language and joined, and the same week saw other national fights (Turkey vs China, Saudi Arabia vs Argentina): 17.4M views in our corpus for the belief that auto-translation turned X into a battlefield.

Can AI map an idea war automatically?

Partly. Sentence embeddings of raw posts miss the hidden meaning of jokes. A model that reads the image and writes the implied idea gets much closer, and a human still needs to check the belief list and the uncertain posts. The companion skill statur-idea-war-map packages that workflow.

Version 1.1 · last updated 2026-10-28: Ideas become beliefs: 327 posts re-tagged with 18 beliefs written as what people believe, a stance per post (supports, attacks, neutral), duels inside each belief, and political findings (immigration, arrest data, stereotypes, manga). Changes are logged in the changelog. Author: Dr. Louis Korczowski.

Bring the three ideas you repeat on every sales call.

Work with me
Two-minute application. I reply within 24 hours.