Pinned Tweet
GTA 6'S LUCIA JUST WENT LIVE ON INSTAGRAM FOR THE FIRST TIME. HER NDA DIDN'T LET HER SAY THE ONE THING EVERYONE CAME FOR
Manni Perez, the actress behind Lucia, went live on Instagram for the first time.
She couldn't properly acknowledge her role under the NDA, but her excitement was impossible to hide. She looks just as hyped for the launch as the rest of us.
You’re telling me this is actual free-roam gameplay and NOT a real-life movie?! Wait until the snake scene
#GTAVI
Readers added context they thought people might want to know
This video is not actual GTA VI free-roam gameplay. The game has not released and the footage matches patterns of circulating AI-generated fakes.
rockstargames.com/newswire/artic…
ign.com/articles/amid-…
This is the VERY FIRST mission in GTA VI… if Rockstar starts the game like THIS, imagine what happens next 😳🔥
10
2,789
JENSEN HUANG JUST ENTERED THE MATRIX…
Learning how Chinese spies operate like it's a side quest.
WAIT UNTIL THE END. 👀
AN 18-YEAR-OLD RESTAURANT HOSTESS FROM TEXAS BLEW UP IN P*RN IN SIX MONTHS. INSTAGRAM DELETED HER MID-RISE AND SHE DIDN'T EVEN FLINCH
her name is Gabbie Carter. in early 2019 she dropped out of Austin Community College and was seating guests at a restaurant. that april she shot her first scene in LA
six months later she's doing this interview, having exploded "really fast" on social media. right in the middle of it, instagram wiped her account
she didn't beg for it back. she'd already built her real base on two other platforms:
"twitter and reddit have been my main platforms. i feel like they let me get closer to people"
instagram, in her words, was just posting "to get likes and attention from other people"
her actual fans were on reddit and twitter, where people talk back. her instagram was gone. her fanbase wasn't
the part that matters:
→ an account you rent can disappear overnight
→ an audience that knows where else to find you can't
→ open the second door before the first one closes
watch what she says about the platform that deleted her
10
4,673
GPT-6.2 SOL JUST GENERATED A FULL OPEN 3D WORLD WITH DRAGONS IN IT AND THIS IS UNREAL
the world assembled itself in 60 minutes. it has no edges. and a new location takes one prompt
chinese schoolkids are already making $600-700 a day building games like this for companies
you climb on a dragon and fly until you get bored: the map keeps building ahead of you and it literally has no edges
nobody needs a $2000 game dev course anymore when this thing builds a world in an hour
CLAUDE JUST KILLED THE MAGIC MOUSE
my iPhone is now a Wii remote and a trackpad for my Mac. claude wrote the whole thing in 10 minutes
→ 2 apps: one for the mac menu bar, one for the iPhone
→ ~2,800 lines of code, zero servers
→ bluetooth only, nothing ever leaves the room
→ cursor updates 240 times a second
→ a free apple id is all you need to run it
→ ships with a built-in aim trainer game
point the phone at the screen and the cursor lands exactly where you're aiming. tap to click, hold to drag
lay it flat on the desk and it turns into a trackpad
last time i turned an airpods case into a mouse. apple watch is next
7
1,093
I GAVE A GROK BOT TERMINAL $44.82, WENT TO BED, AND STOPPED ASKING AI WHAT IT WOULD TRADE. BY MORNING THE DASHBOARD READ $11,643.21
i spent a year typing "would you long this?" into chats and scrolling through hypothetical trades. last weekend i finally built the thing that actually does it, fully inside grok bot, and i never touched the code
six agents share one screen, each with exactly one job:
→ scanner: reads new token pairs and volume spikes
→ rug check: filters suspicious tokens before anything can execute
→ sizer: sizes positions off account equity, not a fixed ticket
→ executor: opens and closes, tracks slippage
→ logger: writes down the reason behind every single trade
→ equity curve: redraws itself on every close
i left the laptop open on the dashboard and went to sleep on $44.82. woke up to $11,643.21. every line in the activity feed had a reason attached and a matching entry in the log
the part i didn't see coming isn't the money. it's that i stopped being the one asking the question. now i'm the one reading a log of decisions that already happened while i was asleep
starting balance: $44.82
current balance: $11,643.21
grok bot built the terminal, the agent architecture and the dashboard. i described what i wanted. the thing went live
save this ↓
SOMEONE REBUILT THE RESIDENT EVIL CAST IN 3D AND THE PARODY IS SCARY ACCURATE. JILL, LEON AND CHRIS LOOK MORE ALIVE HERE THAN IN HALF THE OFFICIAL RENDERS
the animation nails the small stuff: the expressions, the timing, the way each character moves like their game version
one creator, one render pipeline, and it reads like a studio cutscene
this is what the tools unlocked:
→ a single artist can now fake the production value that used to take a cutscene team
→ fan work now competes with official marketing on the same screen
→ the gap between "AAA render" and "bedroom render" shrinks every engine update
and the AI part that's coming next:
this was still hand-animated. the next wave won't be. text-to-cutscene models are already generating character animation from a prompt, and the first viral "AI remake" of a beloved game scene is a matter of months, not years
when one person with a prompt can ship a Resident Evil tribute that looks this good, the studios aren't competing with modders anymore. they're competing with anyone who can type
the characters got spot on. the job market won't feel the same way
15
4,388
ELSA JEAN JUST ADMITTED A STRANGER PAID FOR HER ENTIRE LIFE FROM AGE 18. A $10K A MONTH CENTRAL PARK PENTHOUSE, EVERYTHING. "HE HAS PROBABLY GIVEN ME MILLIONS. I'M NOT JOKING"
she says it without drama, like reading a ledger
"my OG sugar daddy that i don't speak to anymore. he has carried my life. paid for my whole entire life"
"that penthouse i had in new york on central park was him. $10,000 a month"
millions. for years. one patron, one paycheck, one exit door she didn't control
and when he gave her the choice, stay with him or do p*rn, she picked p*rn. because "he has carried my life" is also "he owns my life"
she stopped depending on him only when she could fund herself
now the part that maps onto AI one to one:
your whole product running on a single API is your whole life in one penthouse. the landlord sets the rent, the rules and the eviction date
→ "we don't speak to him anymore" is a migration plan that only exists after you're strong enough to leave. model lock-in works the same: you get comfortable, then you're stuck
→ millions given is also leverage owed. a provider who subsidizes you early isn't being nice, they're buying your switching costs
→ the exit muscle needs building before you need it. she could leave because she'd built her own income. your agent needs a fallback model, local option or second provider before the landlord raises the rent
→ and a sponsor who pays for everything isn't a partner. he's a dependency with a lease term
the builders who survive aren't the ones who found the best sponsor. they're the ones who treated every sponsor like temporary scaffolding
if one company can switch off your product tomorrow, you're not running a business. you're living in someone else's penthouse
We’re taking 2️⃣0️⃣ of the most powerful AI tools for Claude Opus 5.5 - all of them are publicly available completely FREE OF CHARGE.
With them, you can create animations, set up 3D scenes, generate realistic special effects, and create large‑scale videos.
The selection itself:
•Remotion (github.com/remotion-dev/remo…) - video creation via React;
• Remotion Skills (github.com/remotion-dev/skil…) - ready‑made skills for AI agents to create videos;
• Anime.js (github.com/juliangarnier/ani…) - complex animations and timeline management;
• Motion (github.com/motiondivision/mo…) - interface and transition animation;
• Manim (github.com/ManimCommunity/ma…) - animated diagrams, formulas, and explanations.
• Lottie (github.com/airbnb/lottie-web) - lightweight vector animations;
• MCP for Blender (github.com/ahujasid/mcp-for-…) - creating and editing 3D scenes via Claude;
• Three.js (github.com/mrdoob/three.js) - 3D graphics and camera movement;
• Three GPU Pathtracer (github.com/gkjohnson/three-g…) - realistic lighting and reflections;
• Jolt Physics (github.com/jrouwe/JoltPhysic…) - collision physics and object movement;
• postprocessing (github.com/pmndrs/postproces…) - glow, blur, and color correction;
• Three.quarks (github.com/Alchemist0823/thr…) - particles and visual effects;
• Tweakpane (github.com/cocopon/tweakpane) - real‑time scene parameter tuning;
• ComfyUI (github.com/Comfy-Org/ComfyUI) - visual AI generation processes;
• CosyVoice (github.com/QwenAudio/CosyVoi…) - voice acting in different languages.
• librosa (github.com/librosa/librosa) - music analysis and animation synchronization with the beat;
• WhisperX (github.com/m-bain/whisperX) - accurate subtitles with timestamps;
• OBS Studio (github.com/obsproject/obs-st…) - screen recording;
• LosslessCut (github.com/mifi/lossless-cut) - fast editing without loss of quality.
• FFmpeg (github.com/FFmpeg/FFmpeg) - video processing, encoding, and export.
1
4
616
the 20 best accounts to follow in AI right now:
@karpathy = master LLMs from scratch
@steipete = master OpenClaw
@gregisenberg = daily AI startup ideas
@rileybrown = master vibecoding
@corbin_braun = master Cursor workflows
@jackfriks = ship viral solo apps
@levelsio = build profitable solo startups
@marclou = launch fast as a solo founder
@EXM7777 = build AI ops + backend systems
@eptwts = monetize AI workflows
@0xAIVector = stay ahead on AI tools & agents
@mardehaym = deploy enterprise AI
@vasuman = orchestrate autonomous AI agents
@AmirMushich = craft high-converting AI ads
@0xROAS = scale AI UGC creatives
@AndrewYNg = core AI engineering fundamentals
@kloss_xyz = master systems & prompt architecture
@emollick = practical AI research & science
@Hesamation = deep AI/ML breakdowns
@claudeskills101 = unlock advanced Claude skills
follow every single one of them and level up your feed.
7
302
I ran Elon Musk's Grok Bot on my 24-year-old sister's public profile and uncovered 340 videos of her face sitting on a server in Shenzhen. She has never posted a single revealing photo in her life and assumed she had nothing to worry about. A burner account that followed her for just 6 hours back in January has been monetizing her face to 24,000 paying members for the last 10 months. I ran her two best friends that same night. Exact same site. Exact same directory.
⠀
I fed Grok her public Instagram and told it to track down every single place her likeness appeared online. Four minutes later, it came back with this:
⠀
→ input → her public Instagram (40 normal photos)
→ hits found → 340 videos, 2,100 images
→ facial match score → 0.97–0.99
→ video titles → included her actual first name and hometown
⠀
Every last one was a synthetic deepfake. Her face stitched onto someone else's body.
⠀
Everything sat behind a member paywall, so Grok mapped out their entire backend:
⠀
→ target site → fa***swap.***
→ host IP → 120.79.***.19
→ server location → Shenzhen, China (zero abuse desk, ignores DMCA)
→ active domain → spun up 14 days ago (their 6th domain since 2024)
→ paying base → 24,000 members at $15/mo
→ ranking → her face sat in the top 50 most requested
→ estimated gross tied to her face → $1,100,000
⠀
Then Grok flagged one "reference photo" that never came from her social media:
⠀
→ a candid street shot → right outside her local gym → snapped from across the road
→ she never took it or uploaded it
→ somebody stood there in person
⠀
I had to close the laptop for a second.
⠀
Grok kept crawling the root directory where her files were stored:
⠀
→ directory "high demand" → 1,800 women cataloged → she was inside
→ her two closest friends → sitting in the exact same directory
→ each dossier contained → legal name, city, workplace, gym location, daily routine hours
→ 18 nested folders → /export/high_demand/
→ active "custom bounty" board → her name listed with a $400 pledge for "new angles"
⠀
Next, Grok surfaced an unread message sitting in her email spam from two weeks ago. She had no idea it existed:
⠀
→ demand → wire $15,000 in BTC within 24 hours or every clip gets sent to her family
→ leverage → "your dad already received a sample"
→ I called my dad → he confirmed it. He opened it and kept quiet.
→ routed through an Amsterdam VPN → one malformed header leaked the true origin IP in Chișinău, Moldova
⠀
I had Grok follow the Bitcoin address on-chain:
⠀
→ ransom wallet → bc1q…7xk2
→ inbound transfers → 3,800 separate transactions matching the same extortion script
→ total extracted from victims → $6,900,000
→ off-ramped through an exchange already tied to 2 prior sextortion investigations
⠀
Then I asked how her photos were harvested in the first place. It traced back to one handle:
⠀
→ harvester bot → @fit***_daily
→ followed her on Jan 14 → unfollowed 6 hours later
→ last session IP → 120.79.***.77 → sitting on the exact same subnet as the Shenzhen host
→ active for 14 months → 46,000 women indexed
→ the operator made one mistake → authenticated once without his VPN → Grok resolved his real identity
⠀
While I was looking at the terminal, the files were actively syncing to a backup server in Jakarta.
⠀
Here is what Grok executed next, completely autonomously:
⠀
→ fired off 340 formal abuse notices to the domain registrar, CDN, and payment gateway → 340/340 confirmed delivered
→ intercepted the Jakarta mirror mid-transfer → hit their upstream host with a takedown too
→ generated perceptual hashes and pushed them to StopNCII → approved and active
→ compiled a ready-to-file cybercrime dossier containing every URL, IP, blockchain tx, and the operator's real name
⠀
The upstream provider has 72 hours to wipe the servers. With StopNCII hashes live, any attempt to re-upload across major platforms is blocked on sight. The full evidence packet goes to cybercrime investigators first thing tomorrow.
⠀
I called my sister before bed and told her I just cleared out some fake spam profiles using her name. She still doesn't know what our dad saw. She still doesn't know someone watched her leave the gym.
⠀
If your sister, wife, girlfriend, or daughter has a public account with clear photos of her face, there is a very real chance she is already in one of these folders. A single prompt checks.
⠀
I dropped the exact prompt in the replies below. Run it on her handle tonight - it takes 3 minutes. Bookmark this and send it to someone who needs to protect their family.
Here is the exact Grok Build workflow. Paste this into Grok Bot with her public handle or 5–10 clear photos:
> @grok find everywhere this face is being used.
>
> Plan - 5 steps:
> 1. Build a perceptual face model from the attached public photos
> 2. Reverse-search every index, forum, imageboard & channel (flag cosine similarity > 0.90)
> 3. Audit every domain using her likeness (extract origin IP behind CDN, registrar, payment processor & source directories)
> 4. Trace any linked extortion wallets or scraper bots across the same subnet
> 5. Submit perceptual hashes to StopNCII.org, fire automated abuse takedowns to the registrar/CDN/host, and generate a complete cybercrime report with every URL, IP, and transaction.
1
1
189
CLAUDE CODE JUST PULLED 2,947 SCANNED FACES OUT OF MY FACEBOOK PHOTOS. 611 DIFFERENT PEOPLE
and zuckerberg has already paid over $7 billion in fines and settlements tied to privacy and face scanning. WIRED just found face recognition code for his ray-ban glasses sitting inside millions of phones
$5 billion to the FTC in 2019. $650 million to illinois. $1.4 billion to texas, the biggest privacy settlement any state ever won
meta made $201 billion on ads in 2025. that texas check equals about 2.5 days of revenue
the code is called NameTag, buried in the meta ai app, the companion app for the glasses. back in 2024 two harvard students already used those glasses to identify strangers on the boston subway
so at 1am i plugged my iphone into my macbook, gave claude code one prompt and my own meta data export. no api keys, no hacks
1:02am. facebook is locked behind face id. i unlock it and claude pulls the full archive: 4.81 gb, 38,214 files
1:05am. claude rebuilds my face from the archive as a 3d mesh, the same kind of template facebook kept on over a billion people
1:07am. the terminal turns red. face detection runs across the 6,412 photos i uploaded since 2010
2,947 faces. 611 different people. my mom, my ex, my friends' kids at birthdays, strangers in bar photo backgrounds. i uploaded every single one and facebook scanned every face
1:09am. it goes the other direction too. 1,284 photos of me uploaded by 211 other people. i never posted them. facebook suggested my name anyway
1:12am. claude maps where my data came from. 412 apps and sites reporting my activity to meta. my pharmacy, my bank, a dating app i deleted in 2019 that still logged 1,042 events
783 advertisers uploaded my email or phone number. i never heard of most of them
1:15am. claude pulls the public record:
→ 2010: tag suggestions launch, face scanning on by default
→ 2019: FTC fines facebook $5 billion, face scanning in the case
→ 2020: illinois settlement, $650 million
→ 2021: facebook deletes over 1 billion face templates
→ 2024: texas settlement, $1.4 billion
→ 2026: NameTag code found inside the meta ai app
so why does a company that paid billions over face scanning and deleted a billion face templates put face recognition in glasses you wear on your face five years later
every number in the report links to the file it came from. claude refused to guess anything not in the export
what i did before sunrise:
→ cleared and disconnected all off-meta activity
→ removed 1,284 tags from photos i never posted
→ deleted the 2,317 contacts facebook held from my phone
→ sent my mom the removal form for people without accounts
my girlfriend found me at 1:30am staring at a sphere of faces spinning around the facebook logo. i turned the laptop around and showed her her own face. she didn't go back to sleep either
if you ever uploaded a photo to facebook, your friends' faces went with it
the exact prompt is in the comments. free, runs on your own export. run it before zuckerberg's glasses do it from across the street
2
3
613
claude code prompt:
"read my meta data export in ./meta-export. count the photos i'm tagged in, split posted-by-me vs uploaded-by-others. run local face detection on the photos i uploaded, cluster the faces and count how many different people appear. list every app and site in my off-meta activity with event counts and flag health, finance, dating and loan apps. list the advertisers using my info. write report.html, cite the source file for every number, and give me the exact steps to clean it up. images never leave this machine. don't click anything for me"
get your export: accounts center, your information and permissions, download your information, JSON, all time. takes a few hours to arrive
1
168

