Seth Godin's Purple Cow Is AI Slop Now. It Collected $235K in 10 Days. Here's How
We lost the war on slop.
Most people still have no idea what that loss is worth.
And the best explanation is a purple cow from a 2003 marketing book.
Meet Jean Phil. @jean_philanthrope on Instagram, @JeanPhilMadame on X. Blond bob. Curled moustache. Brown suit. He drinks coffee, eats oysters, then starts shadowboxing in the middle of a café.
He does not exist. Every frame is generated.
His first clips went up on September 17, 2026.
-> Park clip: 5.6M views in 3 days
-> Café clip: 2.5M views in 1 day
-> Kitchen shadowboxing: 13.8M views
-> 240K Instagram followers in 3 weeks, on 11 posts
A memecoin appeared on day 3. Dexerto put its value above $8M. The wallet listed as its creator collected ~1,970 SOL in fees in 10 days. Roughly $235,000.
Those are wallet receipts, not audited profit. Still. One character. One template. Three weeks.
No studio. No crew. No face that ages, gets tired or asks for a contract.
This article does two things. First it explains why a moustached Frenchman who was never born beat thousands of real creators. Then it gives you the build: 5 steps, 4 prompts, $155.
The plan
-> 1 · The cow Seth Godin saw in France
-> 2 · From cow to herd in 7 days
-> 3 · Why a man who doesn't exist holds real attention
-> 4 · The money
-> 5 · The build: 5 steps, 4 prompts
-> 6 · What it costs
-> 7 · Where it breaks
1 · The cow Seth Godin saw in France
In 2003 Seth Godin published Purple Cow: Transform Your Business by Being Remarkable. The book opens on a family road trip through France.
For the first few miles everyone stares out the window at storybook cows grazing beside the highway. Soon nobody looks up. The new cows are the same as the old cows. What was charming became common. Then it became invisible.
A purple cow would be different. You'd notice it. You'd tell someone.
That is Godin's test for remarkable: worth talking about. And his warning: very good is no longer enough. Very good is what everyone ships, so nobody mentions it.
For 20 years, earning it took a person. Reputation. Experience. Risk. An edge nobody else had.
Now open your feed. It is the French highway. Thousands of flawless AI faces, all very good, all the same cow. Then a Frenchman with a moustache shadowboxes in a kitchen, and the clip pulls 13.8 million views.
Jean Phil is a Purple Cow by the book. The one thing Godin didn't predict is the price. Not a career. A prompt and a ready template.
2 · From cow to herd in 7 days
Here is the part the book couldn't cover.
In 2003 a competitor needed months to copy your cow. In 2026 they need an afternoon.
Derek Mercer showed up days after Jean Phil. Long dark hair, cartoon muscles, chains.
him at 80K followers in 7 days on the same formula. Within a week creators were posting full breakdowns, and the two characters already had a staged rivalry and competing tokens.
If your play works on Monday, by Sunday there are hundreds of clones with the same structure.
The Purple Cow becomes a herd of slop.
So here's the real question. If everything around you is fake, does it matter?
The audience already answered. Nobody in the comments asks for a birth certificate. They ask for the next episode. The algorithm measures retention. It never checks for a pulse.
Godin saw the cycle coming. His advice once a cow works: milk it for everything it's worth, and start building the next one before the first wears off. In his world that cycle took years. Now it takes weeks.
That leaves two positions. Stand in the herd and complain about slop. Or learn to produce cows faster than the herd can copy them.
3 · Why a man who doesn't exist holds real attention
Three mechanics. All of them are visible in his clips.
-> One-frame recognition. Hair, moustache, suit. You know it's him before anything happens
-> Borrowed motion. The movement comes from real videos that already went viral, so the body language reads as human
-> The argument. "Is he real or AI?" makes people replay, zoom in and fight in the comments
Then the platforms joined. On October 2 @higgsfield launched a product called AI Influencer and used the Jean Phil meme in the announcement. The loop is closed: creators publish, viewers share, platforms amplify because every hit sells their tool, more creators try.
That is the window. The tools are cheap, the format is proven, and most people still believe this takes a studio.
4 · The money
-> Token creator fees: ~$235,000 in 10 days around one character
-> Fee math at a hypothetical 0.5% creator share: $10M trading volume = $50,000. $200M = $1M
-> Scale scenario: 10 characters × $10,000/month = $1.2M/year before costs and taxes
The last line is a scenario, not a promise. Repeating a hit ten times is much harder than writing the equation, and a first batch can earn $0.
What changed is the price of an attempt.
5 · The build
Step 1. Design a face people recognize in one frame
"A skinny guy with brown hair, glasses and a suit" is very good. Which means invisible.
You need 2 to 4 loud traits that survive any angle, any room, any motion. Hair. One facial feature. One outfit. One gesture he repeats in every clip. If you can't describe him in a single sentence, nobody will describe him to a friend.
You are my character designer for photorealistic AI characters built for short vertical video.
Give me 3 very different original human characters. Each must be recognizable at thumbnail size in under 1 second.
For each one, write:
- Name and one-line vibe
- Hair: exact cut, shape, colour
- Face: one standout feature and a default expression
- Outfit: exact garments, colours, fit. Identical in every video
- Signature gesture repeated in every clip
- The 2-4 traits a viewer remembers after one second
Rules: a believable real person. No fantasy. No cartoon. No celebrity lookalikes. No generic influencer.
Never write "quirky haircut". Write "silver flat-top exactly 4 inches tall, razor-sharp edges".Step 2. Lock the identity with a character sheet
Generate the hero image in Nano Banana 2 or Nano Banana Pro inside Google Flow. Then build one reference sheet and never generate from a text description again. Every future video pulls from this sheet. Skip it and the face drifts by clip 3.
Use @image1 as the ONLY identity reference.
Create one photorealistic character reference sheet of the exact same person.
Top row, full body: front, three-quarter, side profile, back.
Bottom row, close-ups: neutral headshot, three-quarter headshot, signature expression.
Same outfit, same hair, same proportions in every panel.
Plain light-grey studio background. Soft even light.
Natural skin pores, small imperfections, real fabric texture.
No beautifying. No redesign. No text, labels or logos.
Step 3. Stop asking ChatGPT for ideas
TikTok shows you videos with 5M, 10M and 20M views for free. That is your research department.
-> Train a fresh FYP for 4 days on viral short-form
-> Save every clip above 1M views with one person and a steady camera
-> Ask one question: does my character make this better?
The content can be random. A dance, a street clip, a guy talking in a car. The character is the constant.
Step 4. Swap your character into proven videos
Tools: Higgsfield Genjutsu, Kling Motion Control or Seedance 2.5. One image in. One viral video in. Your character out.
Use @image1 as the ONLY character identity reference.
Use @video1 as the full motion, camera, timing, audio and scene reference.
Replace the main person in @video1 with the character from @image1.
Lock identity: same face, hair, outfit, accessories and body proportions in every frame.
Copy the original performance exactly: body movement, gestures, facial expressions, eye direction, timing.
Keep the original audio untouched and lip sync to it.
Keep the original background, lighting, camera movement and cuts.
Match the source video quality. No glossy, cinematic or AI-perfect look.
No face morphing. No hair flicker. No extra fingers. No duplicate accessories. No on-screen text.Inspect every output. Mouth off? Hair flickers? Regenerate. Start with clips of 5 to 8 seconds and plan on 4 attempts per finished clip.
Step 5. Publish 10 clips, then put an offer behind them
Post 10 clips over 10 to 14 days. Same name and avatar on TikTok, Instagram and X. Change the location and the ending. Keep the character.
Write "AI CHARACTER" in the bio. The reveal is the hook: people click to find out how he was made.
Watch three numbers: completion rate, shares, profile visits. The only question that matters is whether viewers want another episode.
Once posts hold 50K+ views, pick a route:
-> Sell the method. The account is the proof
-> Affiliate links for the tools you use
-> Brand deals inside the same clips people already watch
-> Sell the whole system to brands as a done-for-you service
-> Token creator fees. Highest ceiling, highest risk
My AI character: [describe].
Audience so far: [country, age, what they comment].
Average views per post: [number].
Rank these 4 routes for this exact account: digital product, affiliate, brand deals, done-for-you service for brands.
For each: first-month revenue range, setup time in hours, and the one thing most likely to kill it.
Then write the bio line and the link-in-bio offer for the top pick.
6 · What it costs
-> 50 images in Nano Banana 2: ~$3.35
-> 320 seconds of video at $0.318/sec: ~$102
-> Other AI calls: $20
-> Editing and music: $30
-> Total for 10 clips: ~$155. Budget $200 to $300 with retries
Add 30 to 50 hours of your time for the first batch, learning included.
No actors. No studio. No crew. No face.
7 · Where it breaks
-> Mistake 1: A very good face. Pretty and generic is the cow nobody looks at
-> Mistake 2: Cloning Jean Phil. You become cow number 1,000,001
-> Mistake 3: Skipping the character sheet. Face drift kills recognition by clip 3
-> Mistake 4: Automating before proof. Agents can't fix content nobody wants. Do 10 clips by hand first
-> Mistake 5: Treating a token as income. Launching a coin does not create demand, and market cap is not your money
The bottom line
Godin's family got bored of perfect cows in one car ride. Your audience gets bored in 2 seconds.
The Purple Cow used to cost a career. Now it costs one afternoon and $155. That is why there will be a herd. And that is why the herd is not the threat.
The threat is being one more very good cow beside the highway.
One character. Ten clips. A fixed budget. Then the next cow.