On-chain trading, AI & market analysis. Breaking down the market, sharing insights, and exploring the latest in artificial intelligence. Follow for updates π
GHOST FACE
A 20-year-old doesn't exist, but she's the highest-earning "model" on a content studio's roster. Last month, campaigns using her face brought in $140,000 across 8 brands.
She was never a real person to begin with. He generated her at "20" because the data said that age range converts best for the skincare and fashion niches he sells into β not a real birthday, just a parameter in the prompt.
The pipeline: one reference face locked with a fixed seed, aged and lit differently per brand so nobody notices it's the same generation. A video model handles motion, an image model handles the static ad sets, and an agent picks which 40 of the 300 weekly renders actually ship.
Brands request her by a name he made up on day one. None of them have asked to see a real photo, a contract with an actual agency, or proof she's a real person β the deliverables just show up on time, every time.
He's never adjusted her age since the first render. Every campaign, every brand, every year β she's still 20, because changing the seed means starting the consistency training over from scratch.
8 brands paying for exclusivity on a face that's never once aged a day, never had a bad skin day, and never existed outside a folder on his hard drive.
GHOST STUDIO
A guy in Bali runs a content studio that produces 200 AI-generated model videos a week and has never hired an actual model. Last month: $220,000 across 12 client brands.
He doesn't sell the videos. He sells the face β the same generated woman shows up in ads for skincare, fitness apps, and phone cases, licensed to different brands who each think they have exclusivity.
The pipeline: one reference face, locked with a fixed seed. A video model handles motion and lip-sync from a script. He renders in batches of 20 overnight, picks the best 6 in the morning, and a separate agent edits pacing and cuts to match whatever platform it's going to.
Clients don't know she's shared. Contracts say "custom AI talent," not "one asset running for nine other brands simultaneously." He hasn't been caught yet β different niches, different edit styles, nobody's cross-referencing.
He tested real models first. Turnaround was 2 weeks and $3,000 per video. The generated version costs him about $40 in compute and ships same day.
The biggest complaint from clients isn't quality. It's that she looks "almost too consistent" β the same face, same expressions, frame after frame, week after week, because she's never once had a bad take.
12 brands, 1 face, 0 humans on set. Nobody's asked why she never ages.
FACELESS FEED
A guy in London generates 60 TikToks a week and has never turned on a camera. Last month: $94,000.
Every video features the same face. She doesn't exist. He built her once, then never touched the model again β just the prompts.
The pipeline: a video model takes a reference image and a script, outputs 15-second clips with lip-sync tight enough that three separate platforms have never flagged it. He doesn't render one video at a time. He batches 40 scripts overnight, wakes up to a folder.
The scripts come from an agent that scrapes trending audio, trending topics, and his own top-performing hooks from the last 90 days, then rewrites them in her "voice" β a tone he locked in after testing 6 versions against each other for a week.
He posts across 5 accounts. Same face, different niches β one fitness, one relationships, one just reacts to news. Nobody's connected the accounts. The face reads as different enough with different lighting presets.
Engagement outperforms his old real-face account by 3x. Nobody comments asking if she's real. They just comment on what she said.
He hasn't appeared on camera in 5 months. 60 videos a week, one face, zero of it filmed.
GHOST SELLER
A 24-year-old in Austin made $61,000 on TikTok Live in March and has never said a word to a single buyer.
Her name is Nova. 190,000 followers. 340 livestreams. Zero of them hosted by a human.
He built her in 9 days. Face from a diffusion model, voice cloned off 30 minutes of a voice actor he paid $60. Script generated live by an agent, price and pitch adjusted per viewer.
2 screens on his desk. One shows the stream. The other shows the agent reading chat, flagging buyers by spend history.
Someone asks if the bag comes in black. Nova answers in 1.1 seconds, uses their name, references a comment they left 3 weeks ago.
The agent pulled it from their profile. Nova never forgot anything, because she never had anything to forget.
He sleeps. The stream doesn't. Sales peak at 3 AM while he's unconscious.
Platform rules require a live human host. Every 35 minutes the agent triggers a "brb," loops footage, resumes. 340 streams. 0 flags.
He hasn't spoken to a customer in 4 months. 190,000 people think they know her.
He built a girl to sell for him. Now she outsells every human on the platform.
@CryonixCrypto This is about the current market situation β a lot of people are losing on it and don't understand one thing: with every cycle, only the wrapper changes, the mechanism itself stays the same
Today there was a spread opportunity on the FWDI token. Bitget adjusted the index price, which caused the market price to diverge. Those with fast hands managed to open positions with around a 25% spread before the prices converged again. It was a great opportunity to make some good profit.
pump and dump playbook never changes, only the wrapper does.
thin liquidity, one wallet buys first, volume looks "organic" for about 20 minutes, then the same wallet is the first one out.
retail sees the green candle. nobody sees the wallet that already sold.
if the chart pumped 300% in an hour and you don't know why β you're not early, you're exit liquidity.
@0xEthan I think this week will see sideways movement or growth; even though itβs a meme coin, the Robinhood network is gaining momentum right now, and CatCash is currently one of the most popular ones thereβthough I think it will run out of steam within a month.