Olivia
Online
Kairon
@kairon_flex
exploring ai & tech / building things online
AI
Joined June 2025
6 Following
51 Followers
21 Posts
This is f*cking insane.
Six Claude agents on one terminal. Seven coins. 33 trades.
The whole desk was built in one session - agents, trade log, liquidation heatmap, everything.
Every agent has one job:
SCOUT scans for setups.
JUDGE scores them.
RISK checks the downside.
SIZE picks the leverage.
ENTRY fills.
EXIT trails and closes.
No agent trades alone. SCOUT finds it, JUDGE approves it, RISK clears it - only then a position opens.
16 wins, 17 losses. More reds than greens.
Three reds in a row in the middle. Nobody panicked.
Then the last stretch: BTC short, ADA long, DOGE long - all green.
Small reds, big greens. The balance keeps moving.
The terminal is still running. The agents don’t stop.
Finally... AI trading system that I trust with my money
Built on Jane Street’s rules, the firm behind ~10% of US stocks
6 GrokBot + Opus 5.5 agents find, test and kill trades 24/7 while you sleep
In this article, I show you how to build it: https://t.co/hy4QwGEBHR
This is f*cking insane.
I reverse-engineered my Samsung TV with Opus 5.5.
One sentence out loud and Netflix opens and starts scrolling itself.
No remote. No phone. No digging between the couch cushions.
The setup is stupid simple:
> TV and laptop on the same Wi-Fi
> Photo of the Settings → About screen into Claude Code
> Opus maps the local remote API that almost every smart TV already has
> It builds a tiny remote with three tools only: open Netflix, move, select
> That remote plugs into Grok Bot as a plugin
You say “Hey Grok, open Netflix”.
The TV listens.
Works on most Samsung, LG and Google TVs from the last few years.
The model already knows how the TV takes orders.
You just stop touching the remote.
https://t.co/LCLhoxByi3
a creator turned a French aristocrat with a platinum bowl cut into an AI character and cleared around $35k from one video in 17 days
edith_paf keeps the same face, the same bangs, the same tiny glasses and grey suit every clip. short absurd routines and performances in real places. people already know the silhouette. the comments still argue if the flexibility is real.
→ 268.6K followers
→ 4.5M likes
→ one clip at 16M views
what’s holding:
- one locked face and outfit you recognize in a single frame
- motion taken from real footage, character swapped in
- posts the next one while the last is still getting replies
- zero talking, just the silhouette and the movement
most of these stall after a few videos. this one keeps moving because the bowl cut and the stare don’t match anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature).
take a short movement or routine clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
AI influencers are taking over TikTok and Instagram.
You can now create one in 10 minutes, put it into videos already going viral, and actually monetize the views.
I broke down the full system + every prompt you need to build one: https://t.co/ERRaRO8hvS
a creator turned a French gentleman with a bowl cut into an AI character and the account already pulled 5.7M views on one dance
ed_mondial keeps the same face, the same moustache, the same tweed suit every clip. short dances in ordinary places. people already know the silhouette. the comments still argue if the hair is real.
→ 65.5K followers
→ 2.4M likes
→ one clip at 5.7M views
what’s holding:
- one locked face and outfit you recognize in a single frame
- motion taken from real footage, character swapped in
- posts the next one while the last is still getting replies
- zero talking, just the silhouette and the movement
most of these stall after a few videos. this one keeps moving because the bowl cut and the moustache don’t match anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature).
take a short movement or dance clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
5.9 million likes for 12 seconds of kittens walking out of a tree
that clip follows one of 7 formats behind almost every viral ai video right now. all 7 inside, with the numbers and the money https://t.co/IqrZX2WINf
a creator turned a real-life Peter Griffin into an AI character and cleared roughly $10k in one week.
mr.paulgrayson keeps the same face, the same awkward energy every clip. short selfies, “before” moments and attempts to look cool. people already know the silhouette. the comments still argue if the glow-up is real.
→ 103.1K followers
→ 4.2M likes
→ roughly 30M+ views across early videos
→ one clip at 12.5M views
what’s holding:
- one locked face and vibe you recognize in a single frame
- motion taken from real footage, character swapped in
- posts the next one while the last is still getting replies
- zero talking, just the silhouette and the movement
most of these stall after a few videos. this one keeps moving because the face and the glow-up don’t match anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature).
take a short movement or pose clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
AI influencers are taking over TikTok and Instagram.
You can now create one in 10 minutes, put it into videos already going viral, and actually monetize the views.
I broke down the full system + every prompt you need to build one: https://t.co/ERRaRO8hvS
a 22-year-old turned Edna Mode into an AI character and cleared roughly $12k from a handful of posts
ednamodethebob keeps the same face, the same bob, the same black outfit every clip. short scenes of her walking or reacting in real places. people already know the silhouette. the comments still argue if the glasses are real.
→ 90.6K followers
→ 3.5M likes
→ roughly 18M views across the early videos
what’s holding:
- one locked face and outfit you recognize in a single frame
- motion taken from real footage, character swapped in
- posts the next one while the last is still getting replies
- zero talking, just the silhouette and the movement
most of these stall after a few videos. this one keeps moving because the bob and the glasses don’t match anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature).
take a short movement or pose clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
a 19-year-old turned Escanor into an AI character and cleared roughly $10k from three posts.
prideful_escan0r keeps the same face, the same build, the same presence. short clips of him moving or standing in real places. people already know the silhouette. the comments still argue if the sun is real.
what’s holding:
- one locked anime face you recognize in a single frame
- motion taken from real footage, character swapped in
- posts the next one while the last is still getting replies
- zero talking, just the silhouette and the movement
most of these stall after a few videos. this one keeps moving because the face and the size don’t match anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature).
take a short movement or pose clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
AI influencers are taking over TikTok and Instagram.
You can now create one in 10 minutes, put it into videos already going viral, and actually monetize the views.
I broke down the full system + every prompt you need to build one: https://t.co/ERRaRO8hvS
99% of people watching still think this is a real emo kid.
John Stockman. Purple hair, chains, oversized clothes. Same look every clip. He does short dances in ordinary places and the account already cleared 321k followers and roughly $100k.
→ 8.3M likes
→ ~57M views across 54 videos
→ one clip at 10.2M views
people in the comments still argue if the hair and the chains are real. the silhouette answers that before the video ends.
what’s holding:
- one locked face and outfit you recognize in a single frame
- motion taken from real dance clips, character swapped in
- posts the next one while the last is still getting replies
- zero talking, just the silhouette and the movement
most of these stall after a few videos. this one keeps moving because the purple hair + the chains are hard to confuse with anyone else.
to test the same loop:
lock one character in ChatGPT (face, hair, one clear feature).
take a short dance or movement clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
an old man with a minigun is clearing 170k followers and people still argue if the muscles are real.
massiveunc keeps the same face every clip. white hair, absurd body, orange pants. he walks hangars, stands next to planes, and now shoots cars. captions stay minimal. the comments split between “this is AI” and “how is he that jacked.”
what’s moving:
- one silhouette you recognize in a single frame (old face + the size)
- short clips of him doing something slightly off in real locations
- motion from real footage, character swapped in
- posts keep coming while the last one is still getting replies
most of these stall after a few videos. this one keeps going because the face and the body don’t match anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature like the age or the build).
take a short clip of someone moving or holding something.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
a character that doesn’t exist is still getting comments arguing if he’s real.
same face, same look, short clips in ordinary places. motion from real footage, character swapped in. captions stay minimal. the comments do the rest of the work.
what’s actually holding:
- one silhouette you recognize in a single frame
- posts the next one while the last is still getting replies
- zero talking, just the movement and the face
- people pause to check if the hands or the money look off
most of these stall after a few videos. the ones that keep moving do it because the look is hard to confuse with anything else.
to test the same loop:
lock one character in ChatGPT (face + one clear feature).
take a short movement clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren’t the cleanest renders. they’re the ones people can name from one frame
Shaggy with wide eyes is flipping stacks of prop money and people still can't tell if the guy is real.
shaggytrappin keeps the same face, same expression, same energy every clip. he handles cash that looks off the second you pause. captions stay minimal. the comments are full of people arguing about the money and whether he's AI.
what's moving:
- one locked face and expression you clock immediately
- short clips of him doing something slightly off
- motion from real footage, character swapped in
- posts keep coming while the last one is still getting comments
most of these accounts stall after a few videos. this one is still posting because the eyes + the money bit is hard to mix up with anything else.
to run the same test:
lock one character in ChatGPT (face, eyes, one clear detail).
grab a short clip of someone handling something or moving.
swap the character in with Higgsfield or Seedance and keep the original motion.
drop the next one the same day.
the ones that last aren't the cleanest renders. they're the ones people can name from a single frame
a short guy with a bob cut and a crown is pulling 70k likes on clips that don't even have a real person in them.
lord_efarquad is one of the smaller AI characters running right now. same face every video, same haircut, same height. he dances, walks, struts in normal places. captions stay short and in character.
people in the comments still argue if he's real. the bob and the crown answer that before the video ends.
what's actually working:
- one look you recognize immediately
- motion taken from a real clip, character swapped in
- posts the next one while the last is still getting comments
- zero talking, just the silhouette and the movement
most of these accounts die after a few posts. this one is still moving because the height + bob is hard to confuse with anyone else.
if you want to test the same thing:
generate one locked character in ChatGPT (face, hair, one clear feature).
take a short dance or walk clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren't the best rendered. they're the ones people can name from a single frame
a short guy with a bob cut and a crown is pulling 70k likes on clips that don't even have a real person in them.
lord_efarquad is one of the smaller AI characters running right now. same face every video, same haircut, same height. he dances, walks, struts in normal places. captions stay short and in character.
people in the comments still argue if he's real. the bob and the crown answer that before the video ends.
what's actually working:
- one look you recognize immediately
- motion taken from a real clip, character swapped in
- posts the next one while the last is still getting comments
- zero talking, just the silhouette and the movement
most of these accounts die after a few posts. this one is still moving because the height + bob is hard to confuse with anyone else.
if you want to test the same thing:
generate one locked character in ChatGPT (face, hair, one clear feature).
take a short dance or walk clip.
swap the character in with Higgsfield or Seedance and keep the original motion.
post the next one the same day.
the ones that stick aren't the best rendered. they're the ones people can name from a single frame
This is f*cking gold
I still don't understand why more people aren't paying attention to this.
Derek Mercer's creator is sitting on a gym bro that doesn't exist and already clears tens of millions of views.
→ ~67K Instagram + 160K TikTok
→ long hair, absurd muscles, same face every clip
→ one setting: the gym
→ captions stay in character ("No skinny guys allowed", "Remember your gym bro")
he lifts, walks around, poses. the muscles don't exist. people still argue if he's real while the clips keep moving.
the formula:
→ one silhouette you recognize in a single frame (the hair + the size)
→ short gym moments and movement clips
→ the face, hair and outfit never change
→ post daily in the same voice
a fictional face. a real audience. real money from views and the token.
here's how to test the same format:
→ generate one exaggerated gym character in ChatGPT (lock the hair, the muscles, the look)
→ find a short movement or gym clip you can use
→ swap him in with Higgsfield or Seedance 2.5, keep the motion
→ post the next one before the comments cool off
→ add a simple monetization layer once it starts moving
your first character will probably flop. his didn't, because the size reads from across the feed.
bookmark this. build one face people can name from the silhouette
This is f*cking gold
I still don't understand why more people aren't paying attention to this.
Derek Mercer's creator is sitting on a gym bro that doesn't exist and already clears tens of millions of views.
→ ~67K Instagram + 160K TikTok
→ long hair, absurd muscles, same face every clip
→ one setting: the gym
→ captions stay in character ("No skinny guys allowed", "Remember your gym bro")
he lifts, walks around, poses. the muscles don't exist. people still argue if he's real while the clips keep moving.
the formula:
→ one silhouette you recognize in a single frame (the hair + the size)
→ short gym moments and movement clips
→ the face, hair and outfit never change
→ post daily in the same voice
a fictional face. a real audience. real money from views and the token.
here's how to test the same format:
→ generate one exaggerated gym character in ChatGPT (lock the hair, the muscles, the look)
→ find a short movement or gym clip you can use
→ swap him in with Higgsfield or Seedance 2.5, keep the motion
→ post the next one before the comments cool off
→ add a simple monetization layer once it starts moving
your first character will probably flop. his didn't, because the size reads from across the feed.
bookmark this. build one face people can name from the silhouette
Dead internet theory isn't a theory anymore
this AI uncle already pulled over $25,000 in a couple of days on a character that doesn't exist
→ 165K+ followers
→ 3.2M likes
→ one clip at 11.8M views and 729K shares
→ same face, same outfit, same weird energy every video
he dances in hallways, tennis courts, pizza shops and "portals". captions stay in the same voice: GOOOD MORNIN, WEIRDOS. people still ask if he's real.
the formula is stupidly simple:
→ one cartoon-inspired face you recognize in a single frame
→ short dances in ordinary places
→ the look and personality never change
→ personal videos from $11, custom clips from $23, merch from $26.99
→ a meme coin attached to the character
a fictional face. a real audience. real money.
and you're still just watching?
here's how to test the same format:
→ generate one unmistakable character in ChatGPT (face, outfit, hair — lock it)
→ find existing dance or movement clips you have permission to use
→ swap the character in with Higgsfield or Seedance 2.5, keep the motion
→ post short episodes with the same voice and one reason to come back
→ add a simple monetization layer once it starts moving
your first one will probably flop. test a few characters, scale the one people share.
bookmark this. build one face people can name from the silhouette
Dead internet theory isn't a theory anymore
this AI uncle already pulled over $25,000 in a couple of days on a character that doesn't exist
→ 165K+ followers
→ 3.2M likes
→ one clip at 11.8M views and 729K shares
→ same face, same outfit, same weird energy every video
he dances in hallways, tennis courts, pizza shops and "portals". captions stay in the same voice: GOOOD MORNIN, WEIRDOS. people still ask if he's real.
the formula is stupidly simple:
→ one cartoon-inspired face you recognize in a single frame
→ short dances in ordinary places
→ the look and personality never change
→ personal videos from $11, custom clips from $23, merch from $26.99
→ a meme coin attached to the character
a fictional face. a real audience. real money.
and you're still just watching?
here's how to test the same format:
→ generate one unmistakable character in ChatGPT (face, outfit, hair — lock it)
→ find existing dance or movement clips you have permission to use
→ swap the character in with Higgsfield or Seedance 2.5, keep the motion
→ post short episodes with the same voice and one reason to come back
→ add a simple monetization layer once it starts moving
your first one will probably flop. test a few characters, scale the one people share.
bookmark this. build one face people can name from the silhouette
A news camera followed one wrong delivery across the city and every stop was a different movie
> 211,364 likes, 687 comments. The caption was five hashtags.
A ninja turtle in an alley takes a pizza off a guy in a red cap. The Mask runs a shell game off a folding table in a yellow suit. A jetpack hangs over three cops, and the line is "this is the wrong delivery." A bandaged man in a fur collar sells photos off the sidewalk. The Power Rangers tag a wall, cans at their feet. Blade holds an ID up to the light outside a club and says he has the receipt. Iron Man leans on a sedan and takes cash from the driver, four women already watching.
26 seconds. One camera. Nobody introduces anyone.
-> The TV-60 mark and the running timecode never leave. Footage someone caught reads differently from a render someone posted.
-> Every character is doing a job, not a pose. A turtle arguing about a pizza is a story. A turtle with swords out is a poster.
-> The city stays boring on purpose. Wet alley, dumpster, police cruiser 3158, a neon CLUB sign, a cold beer window. The costume is the only impossible object in the frame.
-> The joke is one delivery passed between films. Turtle, jetpack, Blade's receipt, Iron Man's cash. Pull one scene out and the clip is a costume reel.
-> It cuts before you can check the hands. Four seconds on the turtle, two on the jetpack, three on the wall. Nothing stays up long enough to be fake.
It works on any character people can name with the mask on. A stormtrooper at a hot dog cart, Batman arguing about a parking ticket, a minion running the shell game.
GPT-6 Astra finds the street format that already held. One still per character, image-to-video keeps the camera in the car. The article below is the $11,200 version of that loop: stop inventing the format, swap who is standing in it.
The costumes got the stop. One wrong delivery is what got the replays.
The format, and how to swap the cast without losing the camera, is in the article below
IRON MAN AND BATMAN IN ONE BODY, 10 SECONDS, 3.3 MILLION VIEWS
> 50,300 likes, 561 comments
One suit, two logos. The arc reactor sits where the bat would be, the cowl is half helmet, and the thing walks like it already decided who wins. Ten seconds, no dialogue, no setup.
So why does it get watched more than once?
-> Both silhouettes are readable in the first frame. You don't need a name. The reactor and the ears do the work before the clip moves.
-> It doesn't fight anyone. A fusion that just stands there is a question. A fusion mid-punch is a poster.
-> The camera stays close and handheld. A wide shot would look like a render. A phone-distance shot looks like someone caught it.
-> Nothing changes after second three. The loop is the format. The second watch is free.
-> The comments split into who would win. An argument is a replay, and the clip never answers it.
It works on any two characters people can name from the outline. A stormtrooper with a lightsaber and a jetpack, Wolverine in a Spider-Man suit, Shrek in Iron Man armor.
GPT-6 Astra finds the pair that already held. One still of each, image-to-video keeps the camera in the hand. The article below is the version of that loop that cleared $14,617: stop inventing the character, swap who is standing in the frame.
The fusion got the stop. The fact that it never picks a side is what got the comments.
The sheet for the pair, and how to swap them without losing the silhouette, is in the article below.
Who gets fused next?
https://t.co/xffXRq2xq3
@insomnia_vip I rly hope you love it
A news camera followed one wrong delivery across the city and every stop was a different movie
> 211,364 likes, 687 comments. The caption was five hashtags.
A ninja turtle in an alley takes a pizza off a guy in a red cap. The Mask runs a shell game off a folding table in a yellow suit. A jetpack hangs over three cops, and the line is "this is the wrong delivery." A bandaged man in a fur collar sells photos off the sidewalk. The Power Rangers tag a wall, cans at their feet. Blade holds an ID up to the light outside a club and says he has the receipt. Iron Man leans on a sedan and takes cash from the driver, four women already watching.
26 seconds. One camera. Nobody introduces anyone.
-> The TV-60 mark and the running timecode never leave. Footage someone caught reads differently from a render someone posted.
-> Every character is doing a job, not a pose. A turtle arguing about a pizza is a story. A turtle with swords out is a poster.
-> The city stays boring on purpose. Wet alley, dumpster, police cruiser 3158, a neon CLUB sign, a cold beer window. The costume is the only impossible object in the frame.
-> The joke is one delivery passed between films. Turtle, jetpack, Blade's receipt, Iron Man's cash. Pull one scene out and the clip is a costume reel.
-> It cuts before you can check the hands. Four seconds on the turtle, two on the jetpack, three on the wall. Nothing stays up long enough to be fake.
It works on any character people can name with the mask on. A stormtrooper at a hot dog cart, Batman arguing about a parking ticket, a minion running the shell game.
GPT-6 Astra finds the street format that already held. One still per character, image-to-video keeps the camera in the car. The article below is the $11,200 version of that loop: stop inventing the format, swap who is standing in it.
The costumes got the stop. One wrong delivery is what got the replays.
The format, and how to swap the cast without losing the camera, is in the article below
https://t.co/nN6fldtV4f
@RowhanX haha I saw this video
@socoloffalex how you create it?
Last Seen Users on Sotwe
Azer
Seen from Turkey
akun tobrut
Seen from Indonesia
Thank you ❤️
Dĩ vãng cuộc tình
Seen from Vietnam
Noga 👄
Seen from Jordan
علـي
Seen from France
คู่แฟนญี่ปุ่น(ไม่แลก)
Seen from Thailand
Cruising, heteros e pegação
Seen from Brazil
Andres Quiceno
Seen from Colombia
🦊 ⋆ 𝐖𝐢𝐭𝐜𝐡𝐲𝐅𝐨𝐱 ⋆
Seen from India
Most Popular Users
Elon Musk
@elonmusk
241.5M followers
Barack Obama
@barackobama
118.8M followers
Cristiano Ronaldo
@cristiano
114.5M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.1M followers
Rihanna
@rihanna
98.7M followers
NASA
@nasa
92.3M followers
Justin Bieber
@justinbieber
91.8M followers
KATY PERRY
@katyperry
90M followers
Taylor Swift
@taylorswift13
84M followers
Lady Gaga
@ladygaga
75.5M followers
Virat Kohli
@imvkohli
73.5M followers
Kim Kardashian
@kimkardashian
70.9M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
66.5M followers
Bill Gates
@billgates
65.3M followers
Selena Gomez
@selenagomez
63.1M followers
The Ellen Show
@theellenshow
62.2M followers
CNN
@cnn
61.7M followers
X
@x
60.7M followers