Pinned Tweet
HOLY SH..T! NEW GTA 6 BIGGEST LEAK!
Look at the insane amount of detail inside Lucia’s apartment - Rockstar didn’t miss a single inch of this room 👀
The level of detail on Vice City’s rooftops shouldn’t even be possible on current consoles…
look at that skyline 👀🚨
How is this even possible in a video game?!
Wait until the airplane scene - I genuinely forgot I was watching GTA 6
3
320
Robbing stores in GTA 6 is NOTHING like GTA 5. Rockstar completely rewrote the mechanics—watch how different it actually is 😳
6
282
A desperate mother had nowhere else to turn—so she walked straight up to a stranger to save her baby. Wait until you see what happens next… 🥺👇
2
298
GTA 6 IS GETTING REAL-TIME PROMO NOW 😭🔥
An agent is literally streaming with Lucia from GTA 6. This is NOT something we expected to see 💀
#GTA6
SHE ONLY EVER FILMED WITH HER HUSBAND. THEN VIXEN CALLED AND ASKED HER TO COME WITHOUT HIM
Adult film star Serenity Cox built her whole career on shooting with her husband. For their fans he is not a prop. In a hotwife couple he is the story: part of every scene, every post, every fantasy people pay for.
And Vixen is the luxury brand of the industry: big budgets, glossy shoots, the biggest names. The call every performer waits for.
They wanted her alone.
That is the kind of call that tests a marriage built on this. Say yes and go big without him, or say no and stay a couple act forever.
What would you tell your wife to do?
🚨 GTA 6 HAS BEEN SPOTTED IN THE WILD?!
The FIRST-EVER GTA 6 UNBOXING just surfaced. 👀
WE NEED TO SEE WHAT'S INSIDE. 🔥
4
453
SHE SAID NO TO THE PILL. SHE SAID NO TO HIM. SO HE SAID NO TO HER PAYCHECK
Adult star Sky Bri says her first rap video, with Shy Glizzy in 2020, ended exactly like that. She claims that after dark he and his crew cornered her in the kitchen and pushed a pill on her until she took a piece of it. After 12 hours of shooting, he walked into her room, pulled her onto the bed with his pants down and told her to suck it. No. Then to lick it. No. Then he asked if she was fine with not getting paid for the whole day. She said yes and walked out. She never saw the money. When she later joked about it on TikTok, he sent her a screenshot with one question: "what is this about?"
She tells the story herself in the video below, starting with the pill.
Her first music video. 12 hours on set with the director and crew.
→ She says his energy that day felt off.
→ After dark he cornered her in the kitchen, his "little posse" behind him.
→ A pill she could not identify. "I said no a million times."
→ She finally took "the littlest tiny bit." Molly or ecstasy, she thinks.
→ She still finished the shoot, then went to her room to pack.
→ "He comes in, pulls me on the bed, and I look down and his d*ck is out."
→ "Suck it." No.
→ "Lick it." No.
→ "Then you're fine with not getting paid?"
→ "Yeah."
She got up, walked downstairs and told the director. The crew agreed it was not okay. They put her in a car and drove her home.
She had never told this story before.
Months later she posted a joke on TikTok: "Starring in my first music video. Didn't get paid." And in brackets: "My lawyers will be in contact."
He sent her a screenshot of it.
"What is this about?"
She never answered. "He could probably figure it out, right?"
The fee was 3 to 5 thousand dollars. She could afford to lose it.
Then she thinks about the girls who can't.
"If you don't do this, you're not getting paid. And they need that money. So they're gonna do it."
These are her claims. We could not find a public response from Shy Glizzy.
12 hours of work. A pill she refused a million times. A paycheck that depended on one more thing.
She could afford to walk out. How many girls on sets like that couldn't?
4
3,294
Everyone is looking at the arrow. But you need to look at the hands.
Continued in the next posts...
3
88
GROK BOT RAN A HEDGE FUND OFF MY PHONE FOR 72 HOURS. IT HIRED 5 AI AGENTS, MADE 18 TRADES ON ITS OWN AND TURNED $600 INTO $47,612
i've lost money on memecoins three times, always the same way: buy too late, panic-sell the bottom. so on wednesday i opened grok bot and typed one line: "scan pump.fun. real holders only, no dev bags"
i expected a list of coins. it answered "on it. 5 agents live" and built an entire trading desk by itself
i ran it on paper first. real market prices, fake money. i wasn't handing $600 to a bot i'd met 10 minutes ago
the team it hired:
→ /scanner: checks every new solana pool, 28,000+ in 3 days
→ /analyst: scores each token, holders, liquidity, dev wallets
→ /executor: buys with exactly 8% of the stack, never more
→ /risk: -7% stop and a trailing take-profit on every position
→ /logger: writes down every decision and the reason behind it
then i closed the app and didn't touch it
on friday my phone buzzed. grok had sold a coin called HIGH for +1,774%. i'd never heard of HIGH. it found it, checked it, bought it and sold it without me
three days later i opened the app:
trades: 18
wins: 12
stopped out: 6
best trade: +1,941%
worst trade: -7%
balance: $600 → $47,612
what i keep coming back to is the losses. all 6 stopped at exactly -7%, and it never once waited for a bounce the way i always do. the winners it just let run
when it crossed $47,000 the screen went green and my phone buzzed twice. a friend saw the screenshot and texted "79x in 3 days?? teach me"
i had no answer, because all i did was type one sentence and close the app
next run starts monday. same $600, same 5 agents, and this time i'm watching every trade
2
196
CLAUDE CODE JUST FOUND A HIDDEN CHIP INSIDE MY $19 AIRPORT POWER BANK. EVERY TIME IT CHARGED MY IPHONE, IT SILENTLY COPIED EVERYTHING OFF THE PHONE
it also read the wi-fi password saved on my iPhone, joined our home network, and uploaded everything to a server
bought it for $19 at an airport kiosk 3 weeks ago. no brand, 30,000 mAh on the label, suspiciously heavy
the first time i plugged it in, the screen asked "trust this computer?". i tapped trust without thinking. that one tap handed it the keys
this morning i had claude code alerting me any time a device started reading my files. an hour later, 10 seconds into a charge, the screen filled with yellow warnings, then red ones: data was moving when it should have only been charging
claude code's verdict: my phone saw the power bank as a small computer with 64GB of storage and its own internet connection
a USB cable carries power lines and data lines. a normal charger only uses power. this one used both, and it had already copied:
→ /device: IMEI and serial number
→ /contacts: 1,206 numbers
→ /apps: every app on the phone, including 2 banking apps
→ /photos: 3,912 photos, my passport among them
→ /wifi: our home network name and password
→ my wife's iphone: her photos too. she charged off it twice and tapped trust as well
every time we plugged in, it used that wi-fi key to get online and send the new files:
send to: 47.***.***.12
uploads: 19 in 3 weeks
last one: 2:14 am, while we slept
i unplugged it instantly. claude code blocked the address on our router and walked me through rotating the wi-fi password and the banking logins
i took it to a repair shop and opened it up. what's inside is in the first reply. a 30-second check for your own power bank is in the second
test every power bank in your house today
1
2
257
WHILE I WAS FEEDING MY BANK STATEMENTS INTO CLAUDE, IT FOUND A MONEY LEAK I DIDN'T WANT TO BELIEVE
61 companies charge my card every month. i could name 9 of them
here's everything it dug up, and how to run the same check on yours
what claude found:
→ 14 subscriptions still billing me months after i deleted the apps
→ 9 free trials that flipped to paid without a single warning email
→ one streaming service raised my price 4 times in 12 months
→ my own bank took $1,140 in fees i never agreed to
→ total damage: $7,980 a year on things i never opened
and it's not just me:
people guess they spend $86 a month on subscriptions. C+R Research measured the real number: $219. that's $2,600 a year per person, and almost nobody audits it
fix yours in 5 steps:
→ export the last 12 months of statements and drop them into claude code
→ ask it to list every charge that repeats
→ cancel anything you haven't opened in 30 days
→ email the rest asking for a discount. most say yes
→ new trials go on a virtual card with a $1 limit. they can't flip to paid
run yours tonight. the leak isn't coming, it's already billing
2
133
THE P*RN INDUSTRY NO LONGER NEEDS REAL ACTORS ON SET. AI NOW GENERATES CLIPS LIKE THIS FROM A TEXT PROMPT. AND NOBODY CAN TELL THE DIFFERENCE
a few years ago a clip like this needed a studio, performers, a crew, a testing protocol, a release form
now it needs a GPU and a sentence
the entire pipeline that made the industry an industry is being automated:
→ no casting. the "performer" is a prompt
→ no set. the location is generated
→ no contracts, no limits, no testing. nothing to violate because nobody real was there
→ and no studio taking a cut. whoever owns the model owns the whole production
here's the uncomfortable part: the audience doesn't care
engagement doesn't measure "was this human". it measures watch time. and a generated clip that hits the same buttons wins on the same metric
this is the same pattern that hit illustrators, voice actors and stock photographers. the industry kept selling "real" while the buyers were paying for "good enough at scale"
the studios aren't competing with other studios anymore. they're competing with a text box
and every viral clip that proves nobody can tell real from generated is one more step toward the day when "p*rn actor" sounds like "elevator operator", a job people did before machines learned to do it
1
91
ELSA JEAN FILMED A GANGBANG THAT NEVER CAME OUT. SHE ASKED FOR ONE RULE: EVERYONE TESTED WITHIN 24 HOURS. THE COMPANY SAID YES. THEY LIED. SHE CAUGHT AN STD ON SET, AND THAT ONE SCENE ENDED HER MAINSTREAM CAREER
in the video she tells it calmly, like someone reading a contract a client broke
the host asks a simple question. was there ever a scene she hated?
she can't think of one that made her miserable. she pushed her limits sometimes, but that was her choice
then she brings up the gangbang. it pushed her boundaries. and nobody ever saw it
"it's actually why i retired. they lied to me about testing"
here's how testing works. in porn a test is valid inside a 14-day window. so the person you shoot with could have been tested almost two weeks ago
elsa wanted more. she was barely filming by then, so she could be picky: everyone tested the night before, results back the same morning
her logic: unless they slept with 20 people that night, she's safe
the company told her everyone had a 24-hour test
they didn't
she got a treatable STD. the infection wasn't what broke her. the lie was
"i felt violated. very violated. i can't trust this industry anymore"
she was already on her way out. this was the straw
and she's fair about it: the industry is strict on testing, and she doesn't know anyone else this happened to. one company. one lie. one scene
she started two weeks after her 18th birthday. a stripper before that, watching feature dancers out-earn her. her sugar daddy gave her a choice: do porn or stay with him. she picked porn
most girls lasted 3 to 6 months back then. she lasted years, built a multi-million dollar brand, shot 18-hour days
today she only shoots solo. nobody else in the scene. nobody else's results to trust
ten people in the room watched every scene she ever filmed. nobody checked the one piece of paper that mattered
and that's the exact failure mode that breaks AI agents in production
elsa didn't need a better model. she needed one verified input. the whole system was working, cameras rolling, everyone doing their job, and a single unverified claim upstream wrecked the output
same thing with agents:
→ "the data is fine" without a source = trusting a test nobody ran. an agent that can't cite where a number came from is guessing
→ approval chains exist because humans lie too. "everyone confirmed" means nothing until the agent checks the receipt itself
→ one poisoned input compounds. she trusted one test; an agent trusting one stale price or one fake repo ships the whole run wrong
→ least privilege after a breach. she only shoots solo now = cut the tools and permissions you can't verify, keep only what you control
→ and the kill switch. she retired after one breach, not after ten. your agent should halt on the first unverifiable step, not complete the scene
the companies that fail aren't the ones with weak agents. they're the ones who skipped the one piece of paper that mattered
ten people watched every scene. nobody checked the paperwork. that's your production pipeline right now if your agent acts on unverified data
was that the industry failing her, or one company that should be named?
2
2,966
Claude Code can now detect a background sound that leaks your exact location (and it's literally everywhere)
This is honestly getting terrifying for all of us...
here is the unfiltered breakdown from my 12-hour testing session + an open-source skill to detect and scrub it:
how the technique works — the Doppler curve method created by YouTuber colsto:
1 → whenever an aircraft passes overhead, its engine frequency drops in pitch, just like a speeding siren passing you on the street. On a spectrogram (a visual map of audio), that pitch drop forms a distinct S-curve
2 → the exact geometry of that curve reveals the aircraft's airspeed, its closest approach distance, and the precise second it passed overhead
3 → flight trackers like Flightradar24 record the position of nearly every aircraft second by second — cross-reference that audio curve against flight logs and you identify the exact plane, plus where the microphone had to be sitting
4 → once you lock in the aircraft, the timestamp, and the distance, the microphone is narrowed down to a line running parallel to the flight path — a second flyover pins your exact coordinates
What worked with Claude Code:
• A 10-second takeoff clip from a public Chicago airport webcam. Using audio alone, it extracted the speed, distance, and timestamp, then isolated the exact match out of 35 nearby flights: Southwest SWA4113 (I already knew the webcam's position)
• colsto uploaded a 1-hour audio recording from an undisclosed location and challenged people to geolocate it. Using the aircraft flyovers in the audio, Claude Code narrowed it down to the exact town: Oyster Bay, New York (with partial hints: we knew the general region, and another developer manually tagged the flyovers via doppler-locate)
Where Claude came up short:
• Running completely blind — given only the raw audio, the timestamp, and a 40x40 km map grid — it solved 0 out of 6 test clips
• A Miami test looked like a 749 m hit at first, but feeding it audio from 30 minutes later produced the exact same coordinates — it was estimating based on heavy airport traffic patterns rather than the acoustic curve itself
Bottom line:
I ran this for 12 hours using free flight logs, a 16 GB machine, and noisy airport streams packed with jet roar. The underlying math is proven: doppler-locate scanned the entire US with zero hints and placed colsto's microphone within 67 meters using a much beefier rig.
- With unrestricted ADS-B data and heavier compute, I'm convinced Claude Code can pull this off completely blind too — this setup is just the baseline
either way: background audio can now geolocate you. Claude Code already gets you halfway there, and colsto's full pipeline finishes the job
test your own video files (for free):
1. run this inside Claude Code: npx skills add Bodila51/colsto-leak-and-mask --skill flyover-check
2. prompt "check my video": it scans your file and warns you if an aircraft flyover, mains electrical hum, or embedded GPS/timestamp metadata could expose where you recorded it
3. prompt "make a safe copy": it automatically filters out those acoustic signatures and strips all GPS and date tags from the file
wild... someone just open-sourced an entire reverse engineering degree on GitHub for free.
Now you can wire up AI agents that reverse engineer almost anything:
1. Ghidra
How to dissect any binary.
The NSA’s flagship reverse engineering suite—and the starting point for almost everyone in the field.
github.com/NationalSecurityA…
2. radare2
The exact same capability, built for the terminal.
A UNIX-style reverse engineering framework and CLI toolkit.
github.com/radareorg/radare2
3. LLM4Decompile
A language model trained specifically to decompile.
Translates raw binary straight into clean, readable C—with the 9B v2 model hitting 64.9% re-executability on its benchmark.
github.com/albertan017/LLM4D…
4. r2ai
An LLM embedded directly inside radare2.
AI-assisted binary analysis inside your terminal—where the model reads assembly alongside you.
github.com/radareorg/r2ai
5. ida-pro-mcp
Put your AI agent in the driver’s seat.
Connects IDA Pro directly to LLMs via the Model Context Protocol (MCP).
github.com/mrexodia/ida-pro-…
6. reverse-engineering-assistant
The exact same bridge, built for Ghidra.
An MCP server that lets your agent run autonomous reversing workflows inside Ghidra.
github.com/cyberkaida/revers…
7. REA
Autonomous agents that reverse engineer anything.
Covers 35 target formats—native binaries, APKs, firmware, web apps, Electron builds, and more—with every single finding backed by verifiable evidence.
github.com/morluto/rea
Bookmark this so you don't lose the stack ↓
2
6
179
F*CKING INSANE. I HIRED AN AI CHIEF OF STAFF FOR $0, GAVE IT ONE RULE: "FIND ME $50,000 A MONTH OR YOU'RE FIRED", WENT TO SLEEP, AND OPUS 5.5 HAD IT ON MY DESK BY MORNING
a SpaceXAI engineer dropped one line that hit 5 million views:
"i have a Chief of Staff agent. he knows about all my other bots and manages everything"
a human one runs $200,000 a year. mine runs on Opus 5.5 inside Viktor, on his own cloud computer, with my laptop shut
i set the goal, he works the steps, i only make the calls
the rule:
"find me $50,000 a month of new business. public data only. contact nobody without my yes"
the night:
→ 9pm split the goal into 4 jobs, each with a day and a time, i wrote none of them
→ 11pm pulled 240 dental clinics in Austin off Google Maps and dropped every chain
→ 1am read every site left and kept 25, each with one detail he spotted and where he saw it
→ 2am wrote an opener for each, sent zero
→ 3am built the offer on one page, no invented results, price left blank for me
→ 8am i typed $2,000 a month. 25 clinics × $2,000 = $50,000 a month on the table
honestly the blank price is what i'd pay for. an AI guessing your price loses the deal
and he repeats it every Monday: 100 new clinics a month, and he never forgets to start
240 clinics → 25 worth it → 1 offer → $50,000 a month on the table → 0 sent without my yes
second 12, he leaves the price blank and hands it to me
someone's chief of staff is already pitching the clinics in your city
whole setup below ↓ free, steal it
93

