zostaff
@zostaff • 17,248 subscribers
phd holder/quant researcher/HFT system dev: @chartcultfun
Shorts
Videos
THE WORLD'S MOST DANGEROUS HACKERS ARE F*CKING AMATEURS. I TRACED THEIR ENTIRE $625M HEIST IN 53 SECONDS the fucking Korean hacker group lazarus that the FSB was hunting for 4 years just sucked my dick I pasted one address into tracehound. 53 seconds later I was looking at the dumbest money trail I've ever seen. the "privacy maxis" will hate this. the "mixers make you anonymous" crowd will hate it more. let me show you how bad it is > THEY GOT CAUGHT BEFORE THEY EVEN STOLE ANYTHING the most feared hacking group on the planet needed gas money to rob a bridge. 1.06 ETH. guess where it came from. BINANCE. a withdrawal. from an exchange. that keeps logs of every account that requests one. the $625M heist started with a receipt. > THE WHOLE INDUSTRY WAS ASLEEP FOR 6 DAYS half a billion dollars walked out the door and NOBODY NOTICED. 6 days. not 6 minutes. 6 DAYS. "decentralized security" btw. > THEN THEY BUILT A MONEY SHREDDER 103,854 ETH -> 4 wallets 4 wallets -> 15 burners 15 burners -> TORNADO CASH looks smart right? looks professional? that's exactly what gave them away. > 300 DEPOSITS. IDENTICAL. LIKE A ROBOT. every single burner did EXACTLY 20 deposits into the mixer. not 19. not 21. TWENTY. fifteen times IN A ROW. $175,000,000 pushed through tornado by a script so predictable you could set your watch to it. you don't hide from a pattern by REPEATING IT 300 TIMES. > THEY COPY-PASTED THEIR OWN MISTAKE SIX TIMES 6 burners funded with EXACTLY 3,003.3 ETH. each one mixed EXACTLY 3,000. each one left EXACTLY 3.3 ETH sitting there. the same leftovers. six times. like signing your name at the crime scene. > THE SCARIEST HACKERS ON EARTH CLOCK OUT AT NIGHT across every flow I traced: ZERO moves between 18:00 and 02:00 UTC. they stole a fortune and still kept a 9-to-5. > THEY DIDN'T EVEN WAIT burner gets funded -> first deposit into the mixer -> MINUTES. no patience. no randomness. just a shift at work. so let's recap the "world's most dangerous hackers": > paid for the heist from a binance account > ran the same script 300 times > left the same 3.3 ETH behind six times > kept office hours amateurs. f*cking amateurs. now read this twice: TORNADO CASH HIDES WHERE THE MONEY GOES. IT DOES NOT HIDE HOW YOU MOVE IT. the gas source. the amounts. the leftovers. the rhythm. the sleep schedule. it's all on-chain. public. permanent. and it took 53 seconds to pull it out. 22 wallets. 35 flows. 4 findings. if a nation-state with unlimited budget can't stay invisible... what makes you think the guy who drained your wallet can? that's tracehound. paste an address. watch them run. the blockchain never forgets. it's just been waiting for someone to look. name the hack. I'll trace it live.
zostaff163,027 görüntüleme • 6 gün önce
CLAUDE MADE ME 3 TRADING BOTS IN 15 MINUTES +$2,503 in my wallet the next day, I quit my job that same day. I wrote one prompt and answered a few questions. Claude took three strategies - MACD, RSI + VWAP, CVD divergence. Assigned each bot its own. The first one catches momentum - sees when volume picks up and gets in before the crowd wakes up. The second one trades reversals - waits for everyone to panic and bets against them. The third one scans divergences - when price says one thing but money does another, it follows the money. Built the structure itself, wrote the backtest, ran each strategy on historical data, set up the risk manager, deployed - all on its own. Three bots, three accounts, each trades differently - they don't know about each other. Started with $1, then $5, then $10, then $50, then $2,503 in a day. Citadel, Jane Street, Two Sigma have been trading with bots for years - they don't feel fear, greed, or FOMO - they only listen to algorithms. The market rewards systems.
zostaff1,952,215 görüntüleme • 7 ay önce
Michael Lewis, author of The Big Short and Moneyball, says 160 million dollars a day is being skimmed off the top of the U.S. stock market. After the book came out, the FBI opened a tip line for Wall Street traders. It's taken by HFT firms. Computers that are milliseconds faster than you. You decide to buy Microsoft. They see your order before the exchange does. They buy first. They sell it back to you at a higher price.
zostaff1,008,920 görüntüleme • 5 ay önce
AI FOOTBALL ANALYSIS. A FULL COMPUTER VISION SYSTEM. BUILT ON YOLO, OPENCV, AND PYTHON. You upload a regular match video. No sensors, no GPS trackers, just camera footage. The neural network finds every player, referee, and ball on its own. Every frame, in real time. KMeans clustering breaks down jersey colors pixel by pixel. The system splits players into teams automatically. Without a single manual hint. Optical Flow tracks camera movement. Separates it from player movement. Perspective Transformation converts pixels into real meters. Speed of every player. Distance covered. Ball possession percentage. All calculated automatically. Four hours of tutorial from zero to a working system. The model is trained on real Bundesliga matches. Runs on a regular GPU. Python code - take it and run. Sports analytics is no longer behind closed doors. AI leveled the playing field.
zostaff1,077,438 görüntüleme • 6 ay önce
The founder of Renaissance Technologies - the hedge fund that makes 66% a year, runs the most secretive trading floor on Earth, and has never accepted outside money - once stood in front of 500 mathematicians and explained exactly how he did it. Jim Simons. The Einstein Lecture. American Mathematical Society. 1 hour 20 minutes.They left it on a small university channel. Almost nobody knows it exists. No one's talking about it. Bookmark it before they do. Then read the article below.
zostaff535,308 görüntüleme • 5 ay önce
I MADE MY 3 CLAUDE BOTS COMPETE AGAINST EACH OTHER Same bankroll, different strategies, no overlap. Started at 7:17 PM, stopped 24 hours later. One fast, one patient, one reads patterns. MACD - 147 trades, hold time 8 min, sees volume spike and gets in before the crowd, win rate 61%, Sharpe 1.4, P&L: +$389 RSI + VWAP - 23 trades, hold time 4 hours, waited for extremum, RSI above 80 VWAP confirmed - entered against the crowd, win rate 74%, Sharpe 2.1, P&L: +$641 CVD - 31 trades, hold time 47 min, price goes up but money flows out - sells, price drops but money flows in - buys, win rate 58%, Sharpe 1.8, P&L: +$937 24 hours - +$1,967 Not a single trade overlapped, three bots one market three different answers. CVD won not because it's right more often but because when it's right it takes more. Correlation between them 0.12, when one loses the others don't lose with it. Diversification by logic.
zostaff639,378 görüntüleme • 7 ay önce
I BUILT A BOT THAT PREDICTS FOOTBALL MORE ACCURATELY THAN BOOKMAKERS 3 probability sources. ML model + Bet365 odds + Polymarket. When all three diverge - that's edge 5 seasons of EPL, La Liga, Bundesliga. 7,600+ matches. Each with goals, shots, possession, corners, cards, odds ELO rating using the FIFA formula - accounts for opponent strength, goal difference, home advantage. Not just W/D/L but the context behind every win xG proxy from basic stats - shots on target * 30% conversion + shots off target * 3%. Teams scoring more than they should - regression is coming Rolling averages over 5 matches, fatigue factor, head-to-head history, day of the week Claude API analyzes context the model can't see - motivation, pressure, derbies XGBoost + Random Forest + Logistic Regression in an ensemble. Walk-forward backtest, not random split Bookmaker says 55% home. Polymarket says 48%. Model says 52%. KL-divergence between sources = signal. The bigger the gap + the fatter the edge All three agree - I skip, zero edge. Two against one - I enter on the majority side Kelly sizes the position, Claude explains why
zostaff513,567 görüntüleme • 6 ay önce
The MIT professor who trains quants for Citadel, Two Sigma, and Renaissance just gave a closed-door keynote at Oxford in front of Man Group's $151B team. They left the recording on a public server. Probably by accident. 1 hour. Free. No one's talking about it. Bookmark it before Oxford notices. Then read the article below.
zostaff384,076 görüntüleme • 5 ay önce
a sports analytics guy at MIT DM'd me after my last post "NBA stat models need 10 years of play-by-play data. you're telling me Claude does it from one weekend of prompts?" i told him Claude doesn't model games. it models markets. the Lakers don't need to be predicted. the Polymarket contract on the Lakers does. when the market is at 62% and my detector says 78%, that's the edge. i never touch the game itself. 3,000+ stars. open source framework. one prompt. 48 hours. 8 detectors running on every sports contract live. last 14 days across NBA playoffs: > 62 contracts entered > 49 winners > +$6,217 net he replied once: "so you arbitraged fan emotion" exactly. fan emotion is a price signal. the market underweights blowouts and overweights narrative. my bot only reads the first one. +$31,447 lifetime. 9 weeks. $1,600 seed. copytrade: he never DM'd again. but his team started a research project on "retail sentiment decay in sports contracts" three weeks later they publish. i'm nine weeks ahead of them.
zostaff402,728 görüntüleme • 6 ay önce
I SAW THIS ARTICLE AT 11:27 PM AND DIDN’T SLEEP UNTIL 4:11 AM Read it three times, then just... started building. Took the Avellaneda-Stoikov quoting logic. Wired it to the Hawkes process for order flow. Added the VPIN circuit breaker exactly like the article says. Ran 500 simulations tonight: > Mean P&L: +$312/session > Sharpe: 1.87 > Win rate: 68% > VPIN saves: 11 sessions > Max drawdown: -$890 The VPIN part is insane btw, 11 times it pulled my quotes before informed flow ran me over, without it sharpe goes negative, just like that. Kyle's lambda estimation is literally 30 lines of python, i had no excuse not to build this. Side effect, now i can't sleep.
zostaff437,532 görüntüleme • 7 ay önce
🚨 HOW I BUILT A FULLY AUTOMATED TRADING BOT IN 34 MINUTES WITHOUT WRITING A SINGLE LINE OF CODE > One person built a trading bot in 34 minutes > Didn't write a single line of code manually > Just talked to Claude Code > The bot detects market regime on its own: bull, bear, sideways > Switches strategies for each regime automatically > 13 safety checks before every single trade > Stop-loss, circuit breakers, position limits, correlation check > Walk-forward backtest - not curve fitting, real simulation > Connected to Alpaca - free paper trading > 45 out of 45 tests passing > Full pipeline: data -> regime -> signal -> risk -> order -> log Hedge funds pay millions for this, He paid $0
zostaff365,011 görüntüleme • 6 ay önce
SOME GUY BUILT ABSOLUTE GOLD AND ALMOST NOBODY KNOWS ABOUT IT repo: gemsearch. a chrome extension that puts a spider right into your x feed. it scans posts in real time and finds emerging narratives while you sleep, eat or trade something else how it works inside: > the spider hooks into your feed and pulls every post into a local dashboard > jev module automatically groups posts by narrative. not by tickers but by meaning. if 40 people are posting about ai agents with wallets thats one cluster > site crawler visits project sites from posts and pulls context. docs, team, repo, what is actually built > 4 grok agents verify each narrative separately: first evaluates signal, second checks product, third looks for risks, fourth decides if theres enough data to enter > if all four give ok its a verified narrative > dashboard shows everything in json. export and build your own strategies on top but the real juice is what comes next > built in launcher fires a token by narrative through pumpfun in one action > https adapter connects to any solana launchpad if pump doesnt work for you > 5 attempts per day. 0.025 sol per launch. not spam but precision strikes on verified narratives connected via api to slimefamilyxyz which does big work in the era of narratives. slime finds growth and tracks the wave, gemsearch verifies through four groks and can launch a token by narrative in one action. together its a pipeline that does the work for you that is impossible to do by hand while everyone was looking for alpha in chats someone was building a tool that finds it on its own 🕷
zostaff50,432 görüntüleme • 24 gün önce
Jane Street, Goldman Sachs, JP Morgan, BlackRock, Hudson River Trading, Two Sigma, D.E. Shaw. The most expensive engineering teams in the world released their financial tools on GitHub. Here are 7 repos, one from each. 1. Jane Street, janestreet/magic-trace 5.3k stars. Process tracer powered by Intel PT. When your profiler is blind, magic-trace sees every CPU instruction. 2. Goldman Sachs, goldmansachs/gs-quant Derivative pricing the GS traders use at their desks. MIT licensed. 3. JP Morgan, finos/perspective What JPM traders use to watch markets in real time. A $24k/year terminal, for free. 4. BlackRock, blackrock/lcso Rust optimizer for portfolio problems. Where scipy gives up, this works. 5. Hudson River Trading, hudson-trading/corral Structured concurrency for C++20. The foundation of HFT infrastructure at one of the largest U.S. trading firms. 6. Two Sigma, twosigma/flint Time-series joins on Apache Spark with temporal tolerance. Built for billions of ticks. 7. D.E. Shaw, deshaw/pyflyby Auto-import for IPython and Jupyter. D.E. Shaw also funded the development of IPython itself. Bookmarked it
zostaff294,142 görüntüleme • 5 ay önce
Vinod Khosla, VC legend ($40B), on stage in India: "This whole IT service industry and BPO industry will be gone." Millions of engineers whose careers run on outsourced tickets. He gave one door. GitLab's founder got cancer, refused doctors, designed his own drugs from scratch. Alive and healthy. No medical degree, just agency. He calls it founder mode. It used to be a career choice. Now it is a survival filter.
zostaff188,939 görüntüleme • 4 ay önce
20 years ago Jane Street's entire compute cluster was six Dell boxes stacked on the floor at the end of an office row. They called it the Hive. Last week they let Dwarkesh Patel walk through their new Texas data center. 4,032 GPUs. Each rack pulls 140 kilowatts. A normal data center rack pulls 10. 8,000 kilometers of fiber inside one building, and the fastest links are still copper because light moves more slowly through glass than electrons through metal. Bookmark it tonight. Then read the article below.
zostaff236,781 görüntüleme • 5 ay önce
10 repos that mass replace a $100,000/year football analytics department. all free. all open source. -> replaces Hawkeye and Second Spectrum YOLO tracks every player and ball from any broadcast. assigns teams by jersey color. calculates speed, distance, possession. from a TV feed. no sensors. -> replaces entire quant sports desk stacked ensemble: LightGBM + XGBoost + Neural Networks + Random Forest. scrapes FBRef automatically. ELO with dynamic K-factor. Poisson xG. MongoDB backend. the most complete open-source football prediction pipeline on GitHub. -> replaces paid prediction platforms ($30/mo) full GUI app. 7 ML algorithms. downloads data from football-data. co. uk. predicts upcoming fixtures. exports to Excel. one click. -> replaces manual feature engineering XGBoost with 354 hand-crafted features. works for any European league. data straight from football-data. co. uk. plug and predict. -> replaces value bet scanners ($50/mo) ELO + expected goals + offensive/defensive ratings. compares model probability vs Vegas lines. flags when you have edge. -> replaces bookmaker calibration tools Gradient Boosting tuned to output probabilities that match real bookmaker odds. not just accuracy - calibrated confidence. -> replaces StatsBomb xG subscription xG model from KU Leuven researchers. LogReg + XGBoost pipelines. supports Wyscout, StatsBomb, Opta data. academic grade. -> replaces xG analytics dashboards xG on StatsBomb open data. SHAP explanations for every prediction. proper calibration. tested on FIFA World Cup 2022. -> replaces basic prediction models Poisson distribution for goal simulation. the classical approach that still beats most ML models on draw prediction. -> replaces Premier League prediction services XGBoost + AdaBoost + SVM on EPL data. detailed EDA. confusion matrices. honest 56% accuracy - because football is hard. like + bookmark you'll need this when you build your first football prediction bot
zostaff224,902 görüntüleme • 6 ay önce
AI BASKETBALL ANALYSIS. A FULL COMPUTER VISION SYSTEM. BUILT ON YOLO, OPENCV, AND PYTHON. Take any NBA broadcast. Any camera angle. Any resolution. Feed it into the system. YOLO finds every player and the ball. Frame by frame. No manual annotation. No pre-labeled data.The model just sees the court and understands it. Zero-shot classification looks at jersey pixels and decides who plays for whom. Two teams separated in milliseconds. Without ever being told the team names. But here's where it gets wild. Court keypoint detection identifies the geometry of the playing surface. Homography transforms the broadcast camera into a top-down tactical map. Real meters. Real coordinates. From a flat 2D video. Now every player has a position on a real court. Every movement becomes a data point. Every pass becomes a vector. Speed. Distance. Possession time. Pass networks. All extracted from nothing but pixels. No GPS. No chips in the ball. No million-dollar Second Spectrum setup. A Python script and a GPU. The full pipeline is open source. The tutorial walks through every line of code. From detection to transformation to analytics output. Teams used to pay six figures for this data. Now you build it in an afternoon. Computer vision didn't just enter basketball. It democratized it.
zostaff231,762 görüntüleme • 6 ay önce
Marc Andreessen, co-founder of a16z: "The price of AI is dropping faster than Moore's Law. The per-unit cost of every AI input is collapsing." The GPU shortage everyone is panicking over? In his read it flips to surplus within a decade. Shortage is the number one cause of surplus. The thing you can't get today gets built until it's cheap. The man who funded half the internet says the bottleneck strangling you now is temporary by design.
zostaff163,825 görüntüleme • 4 ay önce
Dylan Patel, founder of SemiAnalysis: "The upper bound on how much compute can be produced by 2030 is around 200 gigawatts a year." The entire world has about 20 gigawatts of AI deployed right now. The ceiling is 10x what exists today, and it still isn't enough to feed what Sam, Elon, Dario and Demis are racing to build. Everyone argues about which model wins. The real limit is a number measured in gigawatts, and it's already maxed out years in advance.
zostaff163,567 görüntüleme • 4 ay önce