Goshawk Trades
@GoshawkTrades • 76,610 subscribers
Algorithmic trader | 7 years live | 10+ algos in production Helping traders go algorithmic (300+ so far). Co-founder @AskEdgarIO (AI agents for funds).
Shorts
Videos
druckenmiller at 27, running $6 billion with no experience, when the shah of iran fell: "I said this is easy, let's put all our money in oil stocks and defense stocks. if I had been a little older, a little wiser, I would have diversified. it went up a lot, so everybody thought I was a genius." "I wasn't a genius. I just didn't know any better."
Goshawk Trades273,519 просмотров • 4 дней назад
jane street doesn't allocate its GPU cluster by committee. it runs a live internal auction and lets researchers bid against each other for it. no approvals. no allocation meetings. no manager deciding whose model matters more this quarter. "our compute cluster is available for everybody. you're tasked with realistically thinking about how to value this job you're running versus all the other things." "and there's not really any guard rails." it's also globally readable and writable. anyone can see what anyone else is running, and anyone can kill anyone else's job. "you can just nuke somebody else's job anytime." which produced exactly the outcome you'd expect at least once: "there was an instance where an unwitting researcher ran a command in a jupyter notebook to adjust the bids on his job. he accidentally ran the command without any arguments — which meant he changed the bids of every single job running on the hive." they fixed that one afterwards. but the visibility is the point. if your job is failing while you're away from your desk, someone else sees it and pauses it for you. "i see that you're running this job and this MFU is just terrible. what's going on?" "a lot of things can be sorted out without it going up the ranks." that cluster is the same one from their data centre film. 4,032 GPUs, 56 racks, 8,000km of fiber, allocated by internal price discovery. they built a market to decide who gets to use the market-making machine.
Goshawk Trades656,470 просмотров • 14 дней назад
a 24-year-old ex-openai researcher ran a fund to $45 billion with eight people, was up 439% net through june, and by the end of july had sold his entire public stock portfolio to ken griffin in a single block trade. griffin described this exact failure mode two years ago, asked why portfolio managers wash out at citadel: "you have a portfolio that is extraordinarily highly concentrated, you have large positions, you cannot demonstrate a clear and concise competitive advantage in why you own those positions." "and there are some people that just, with full information, are unable to help themselves and get to a better portfolio construction." he was also asked the opposite question in the same interview, why citadel keeps working, his third and final answer was this: "it's experience. it's the price paid in losses and pain that converts into wisdom. my leadership team, we've been through a lot of very difficult moments of the markets together. we've learned some very bitter lessons. but it makes us much more effective as investors in periods of turmoil and crisis." citadel started in november 1990. situational awareness started in 2024, long AI infrastructure and short software at roughly 4x leverage. both legs went against it in the same three weeks. there was no bitter lesson priced into that book yet. that's what got bought. he wasn't wrong about AI. he was wrong about the construction of his book.
Goshawk Trades1,653,167 просмотров • 1 месяц назад
an institutional trader who was personally recruited by andy krieger has asked every retail trader he meets the same question for 14 years: what is your edge? "you know how many times I've gotten a legitimate answer to that question? zero. literally not once." the answers he gets: I follow smart money. I have excellent risk management. I know a chart pattern that works all the time. his definition of what they should be saying: "an edge is a mechanical cause and effect relationship in the market that is repeatable enough for you to take advantage of it and profit from it over time." I see the same thing, most have zero edge.
Goshawk Trades124,248 просмотров • 5 дней назад
3Blue1Brown (grant sanderson) asked a jane street researcher whether people modelling markets think about time as continuous rather than discrete. "when you're modeling real world time series, they're not always so cleanly discrete, or as regular." "i do also think when you're modeling real world physical phenomena it is actually quite nice to say time is continuous. we just sample it at some points that we care for." in language the index is the whole object. token 1, token 2, token 3. there is no duration between tokens unless you add one. market data comes with a clock attached. quotes and trades arrive when they arrive, a 1m or daily bar is a grid you laid over them afterwards, and the waiting time between two events carries information: intensity, liquidity, whether anything is happening at all.
Goshawk Trades173,744 просмотров • 9 дней назад
Jane Street just showed the inside of their AI training data center in Texas. 4,032 GPUs. 56 racks. 8,000 km of fiber. liquid cooling running through every server because air cooling can't handle the heat anymore. but the part that got me was the origin story. Ron Minsky, who co-heads their technology group. said their first compute cluster was literally six Dell boxes stacked on top of each other at the end of a desk row. they called it "the hive." the trading systems sat out in the room with the traders because they wanted to be able to unplug them if something went wrong. at one point, someone vacuuming the office unplugged a live trading system in the middle of the day. from six Dell boxes and a vacuum cleaner incident to a liquid-cooled GPU data center processing trades in under 100 nanoseconds. that's a 20-year arc.
Goshawk Trades2,276,047 просмотров • 3 месяцев назад
did a defense contractor just sell me a Nintendo 64? palmer luckey sold oculus to facebook for $2 billion at 21. now runs anduril, a $61 billion autonomous weapons company. and this weekend I find him on youtube, 30 minutes deep, explaining an N64 clone he engineered to last for centuries. gold-plated every connector. removed every component that can physically degrade. built the first bluetooth 6.2 device on earth to get wireless latency down to 0.375ms. then open sourced the entire thing. "i wish that all hardware was designed this way, but i unfortunately only designed the hardware that i saw."
Goshawk Trades550,755 просмотров • 1 месяц назад
Lloyd Blankfein, former CEO of Goldman Sachs, broke down his entire personal trading setup: portfolio: 98% risky assets. 75-90% single stocks. mostly big tech hyperscalers plus "second tier" names slightly below blue chip. "i invest in risky assets. that's what's fun for me." "do you have a team? oh, just me." hardware: no computer. an iPad and a phone. information source: texting and calling people. "somebody will text me. i'll text them. then i'll get tired of tapping things out because of my fat fingers. so i just call people up." frequency: trades every single day. multiple times. "it's taking a lot of discipline not to look at my screen while i'm talking to you right now." "some people listen to music. to me, the market is like music. it's out there. it's going on." has he outperformed the market? "yes, i have for a while. it's because of where i focused, tech, energy, and financial services. i know a lot about financial services having been in financial services." the former CEO of Goldman Sachs. manages his own money. on an iPad. alone. focused on three sectors he actually knows. and beats the market. sometimes the edge isn't the model. it's 40 years of pattern recognition and a phone full of the right contacts.
Goshawk Trades1,067,710 просмотров • 2 месяцев назад
"every single trader out there seems to make the same mistake over and over again. their position size is probably two to ten times more than it should be." martin shkreli built a kelly criterion simulator to test exactly this. "most folks don't actually have an edge when they trade. but if they did, they're trading as if they had a 4x or 5x kelly edge." "if you run the simulator, you will go to zero each time." "even with a 60/40 edge on every trade you make, you'll go bust if you overbet." "and who has a 60/40 edge in the stock market? nobody has a 60/40 edge." even a real edge, sized wrong, still ends at zero. sizing is too commonly the part I see every trader consider second or third.
Goshawk Trades340,101 просмотров • 1 месяц назад
Ken Griffin on the single factor he looks for when hiring at Citadel: "show me an athlete who did well academically." "an athlete because they know what it takes to win and they've had to experience loss." talent is everywhere. what's rare is someone who knows how to lose, recover, and still perform at a high level. same thing separates profitable traders from everyone else.
Goshawk Trades1,218,389 просмотров • 3 месяцев назад
"i notice a lot of people's crypto portfolios are very inflexible." flood runs a family office trading crypto full time. this is how he actually structures a position, and it's the opposite of how most people hold one. "for a hype position, we will typically have a default spot position. we will also potentially be short the future. and then we may or may not have options on, depending on whether we think they're fairly or unfairly priced." "this gives us three different levers to pull as the market gyrates up and down." what that buys you: "i'm long 10 million hype, i'm short one million of futures against it. you'll actually be paid to put on this trade. and then when the market goes down, you have a button to click to not really buy more hype, but cover a directional position." the alternative: "people are like, i own 80% bitcoin, 10% altcoins, 10% cash. okay, if the market goes down much further than you ever thought, you're not going to be able to drastically alter your portfolio composition." "people do not give themselves enough ability to make decisions. they're allocated to something, prices go in either direction, and then they can't make adjustments."
Goshawk Trades71,229 просмотров • 10 дней назад
jane street invented their own trading game to train people. it isn't poker. it's called Figgie, and a former jane street trader, describes it as modelled on an open outcry pit: "almost like the chicago pit traders type trading experience, where you have four or five players and you're bidding for these things." "you're quoting these spreads, i'm three bid spades, i'm offering a five for diamonds, and you can make a lot of your money in this game by capturing the spread that you're quoting." "it really teaches you how to be a market maker in this toy example." SIG went a different route entirely. their training runs on texas hold'em, and new traders play a minimum of 100 hours of it. two of the biggest market makers alive, both training people on card games instead of markets. "you can't model these things completely through, and this is much more akin to the real world, where incomplete information is sort of all over the place." jane street have since put a digital version of Figgie online. you can go play it.
Goshawk Trades132,016 просмотров • 26 дней назад
this guy won a World Series of Poker bracelet, then became a quant at Susquehanna (a firm with over $500B in market making). here he breaks down the decision process behind every trade of theirs: incomplete information, time, and EV. this clip is worth 3 minutes of your day.
Goshawk Trades558,645 просмотров • 4 месяцев назад
When Russia invaded Ukraine, BP, Shell, and Vitol ran from Russian oil. A 27-year-old trader ran toward it. Three years later, Christopher Eppinger made $250 million, owns a €7M villa on the French Riviera, and a private jet. The full story of how he did it below:
Goshawk Trades520,046 просмотров • 4 месяцев назад
Ken Griffin, CEO of Citadel, on how he handled the AI hype inside a $60 billion hedge fund: "i said, 'i want a list of all the GenAI-based projects that we have at Citadel.' he came back with 200 projects." "i said, 'no, no. five. we're going to do five.'" "'what about the other 195?'" "'we're going to kill all those. we're going to do five. we're going to push them really hard. we're going to see what works and what doesn't.'" "we can't afford to let everybody be pursuing random ad hoc projects. it's too expensive. it's too much of a distraction." "of the five that we pursued, we had real success with just over half." "i don't think we incurred any cost of killing 195 projects." this is one of the most important lessons in strategy development too. you don't need 200 ideas in research. you need 5 that you push hard enough to actually validate. most traders spread themselves across too many ideas and end up with none that are live.
Goshawk Trades296,952 просмотров • 2 месяцев назад
a quant researcher at IMC (one of the oldest electronic market makers in the world) on a result that catches most candidates out in interviews. first, why "just predict further out, there's less noise" is wrong: "a number of people suggest they would try and predict one day into the future because there's less noise. i think they're thinking of some kind of central limit theorem idea." "but if you just think about what happens in markets ,the longer you look into the future, the more can happen. the more people can come and trade with you, the more idiosyncratic events." "looking further into the future has more variance and is harder to predict." then the counterintuitive part: "models trained for shorter term predictions can actually be better at predicting longer horizons than a model specifically trained on that target horizon." "you might train two models, one at 5 minutes, one at 10 minutes. and you might find the five minute model is actually better at predicting 10 minutes than the 10-minute model." why: "there's a battle between the signal you're trying to use getting diffused by noise the longer you look. it can be better to train on a shorter horizon which captures more of that short-term signal, and then leverage the conditional relationship between the 5 and 10-minute targets."
Goshawk Trades30,821 просмотров • 7 дней назад
in march 1999 stanley druckenmiller shorted $200 million of internet stocks. within four weeks the position had lost $600 million. so he hired two guys in their twenties who actually knew the names, flipped long, and finished the year up 42% net. in january 2000 he called it a bubble and sold everything. then he sat in cash and watched the two twenty-somethings make 8% a day. "there's the little devil here going: do it, do it. and the little devil here saying: don't do it, don't do it." "i think i missed the top by an hour." he put in $6 billion. it became a $2-3 billion loss and he knew within five days. "so now i'm down 18%, and i had just been down 18% the year before, and i'm literally an emotional mess." so he quit. wrote to his investors, told them they could take their money back, and went to africa with his kids. no television, no financial newspapers. "if you take a break, take a break. don't sit there and trade your own account and mess around." he came back on labor day, opened a newspaper, and saw oil up, rates up, dollar up. he put 350% of the fund into treasuries. made 40% in the fourth quarter. "i will go to my grave believing that if i had sat in my office the whole year grinding away looking at the screen, i would not have seen it. my brain was too messed up in may and it would not have gotten unmessed up watching that price action." the trade that saved the year required four months of not looking at a screen.
Goshawk Trades111,268 просмотров • 1 месяц назад
Ken Griffin, founder of Citadel, the most profitable hedge fund in history, on the moment that left him "shocked and depressed" about AI: "one of our team members built an agentic AI system to recreate academic papers in finance." "we have a legion of young masters and PhDs doing this work. it takes roughly six to eight weeks to reproduce a paper." "my colleague built an AI system that would read a paper, reproduce it, verify the results, produce the results on out of sample data — in about two to three hours per paper." "this is not just a white collar job. this is a master's or PhD-level job. six weeks of work turned into two to three hours." "and then OpenAI just solved a math problem that no one had solved for eighty years." "competitive moats are being filled in at lightning speed." the founder of Citadel watching PhD-level work get compressed from six weeks to two hours. and his response wasn't to cut headcount, it was "i will take every single productivity gain i can get because with the talented people we have, we just have more to go after."
Goshawk Trades184,227 просмотров • 1 месяц назад
Druckenmiller on why most investors can't sell losers: "if the reason i bought a stock is no longer the case, i don't care what i paid for it." "if i bought it at 60 and it's 50 because the market discovered the problem before me, i have no emotion whatsoever." "after a while you develop enough confidence that you're not afraid to clean the slate and start over. because you have the confidence that you'll be successful again." "you're not going to sit there in a lazy position that you're not that sure about anymore. just clean house." "i don't care what i paid for a stock. it's absolutely irrelevant to my investment process going forward." then on buying back higher than he sold: "i don't like it. but i'm perfectly willing to buy something back higher than i sold it. some people can't get themselves to do it. i can." this is the exact same principle behind properly managing strategies. if the edge isn't there anymore, turn it off. don't sit in it hoping it recovers just because you spent three months building it. clean house. start fresh. the next opportunity doesn't care about your sunk costs.
Goshawk Trades216,684 просмотров • 2 месяцев назад