squa retweeted
There is an (unpatchable?) supply-chain zero-day in PyTorch autograd that allows a regularization term to use the model's weights as RAM and SGD as its CPU clock to run any computation during training.
squa retweeted
Cell phone addiction is an absolute thermonuke to the brain and nobody 'takes it seriously' because it is not clear that there is any solution in a world where the revealed preference of everyone across the globe, regardless of income, race, age, any factor whatsoever, would rather be on their phone than do *anything* else. Deciding "I use this too much" and meaningfully altering one's behaviour doesn't seem to have much success for any of the very small minority of people with the sentience or self-reflection to realize how bad it is. Most people seem to think of it as an unproductive habit at most, rather than an addiction. Any attempt to speak seriously about this is met with derision, accusations of pretentiousness. There is no real solution to it. The future of human cognition fifty years down the line is horrifying if you stop and think about it for even a moment.
squa retweeted
$3.50 to $189 in 52 years = 8% CAGR
I'm speechless.
Readers added context they thought people might want to know
Disney World tickets in the ‘70s only covered park admission. Rides cost extra.
Disney World tickets today cover park admission AND all rides.
allears.net/walt-disney-wo…
Also, Disney World’s operating costs have increased.
squa retweeted
delete IG from your phone now. i thought it was twitter sucking all my time and mental energy (which it does) but i think IG is actually worse. short form video is poison. delete!!
squa retweeted
what I love about black people is they’re more prone to speaking their mind
he understood as much of Nietzsche as your average right wing redpill bro or your average SF tech bro
only difference is that all those chuds sit and pretend like they know what Nietzsche is talking about while they don’t know shit
you even got billionaires like Marc Andreesen going on podcasts pretending like they understand Nietzsche
but in reality Marc understands less of Nietzsche than the guy in this video
as being straight up is the true essence of the Nietzschean, thus this man right understands Nietzsche better than all of San Francisco and he hasn’t even realized it
he read a couple of pages and already understood Nietzsche in a way that those lying technohomos never will
squa retweeted
Something to keep in mind--especially when dealing with humans--is that the evidences about an individual totally outweigh the prior about a category they falls into.
squa retweeted
45 years on they still don't get it.
I just had a deeply alarming 45 minute sync with our Talent Acquisition team regarding their new AI-driven resume screening tool.
They reported that the software's reduced our time-to-hire by 60% over the last month.
I asked them to walk me through the specific algorithmic parameters they established for filtering out unqualified applicants.
The HR manager smiled and said they just told the AI to optimize for digital natives with high runway potential and a youthful energy.
I stopped typing and asked her to run a demographic report on every single candidate the software's rejected since September.
The data showed a 100% rejection rate for any applicant who graduated college before the year 2010.
I explained that automating age discrimination doesn't make it suddenly legal, it just makes it highly scalable and incredibly easy to prove in court.
She seemed genuinely confused and argued that the AI made the decision, not her, so we're insulated from liability.
I had to explain that the EEOC doesn't currently accept the robot defense for federal civil rights violations.
I'm currently drafting a memo to halt all hiring until we manually review 4,000 resumes.
squa retweeted
lack of people pairing up into couples and getting married reduces social cohesion in unexpected ways. we got new neighbors. saw the guy outside, said hi. turns out his sister died the day he moved in. obviously, him and i never discussed this. but his wife told my wife - only the two women would communicate that.
so my wife got them flowers - and gave them to his wife. now, we’re basically friends overnight. but there’s no way the two men in the situation (him and i) would have done this absent the women. i wouldn’t get a guy flowers (bizarre, to me) and he wouldn’t (from his perspective) dump this information on some random guy he doesn’t know. you begin to understand the complex formalities present in cultures like victorian england or feudal japan that carved clear avenues and back channels for things like this.
squa retweeted
Let me be specific.
1) In a democracy where one person has one vote, the larger generations like the Boomers have more voting power than the smaller generations. Since a fiat money system allows for greater deficit spending than a hard money system like gold or gold-backed paper, Boomers were able to “borrow money from the future” by using deficit spending to gradually, over time, expand government programs like Medicare since its creation in 1965. On top of that, the deficit spending caused inflation, particularly asset inflation like in real estate. Given old people’s higher propensity to vote and the great size of the Boomer generation, they protected the value of their real estate by preventing the building of new housing. This results in younger generations buying homes later in life and thus starting families later and thus lowering TFR overall.
2) Social Security only works if the population keeps growing. Otherwise, it’s a ponzi. You cannot have an inverted demographic pyramid where successively smaller younger generations work to hold up, on their shoulders, massive older generations who no longer work.
3) Every policy decision the Boomers made when they were in power was to kick the can down the road or to protect themselves. Subsequently, each business cycle of booms and busts brought us closer to the interest rate zero bound. When Volker took rates to the stratosphere, debt-to-GDP was 25%-39% so, although politically unpopular, it was mostly fine. Today, as we approach the era of fiscal dominance, this is not just as politically unpopular but also mechanically much more difficult. In 2008, the market was not allowed to clear and reset. During COVID lockdowns, future economic prosperity for the youth was sacrificed for marginally more safety for old people.
4) All of these things create distortions in the mating market. Since young men compete with old men for young women, the difference in material conditions due to the above points leads to greater prevalence of age gap relationships and harem-style structures. This creates resentment in both young men and young women which ferments into a propensity toward communism. It also causes lower TFR. I have a much longer explanation of this anthropology for another time.
Ultimately, there was nothing uniquely selfish or evil about the Boomers. The real problem was democracy plus fiat so long as some generations are big and others are small. But now that were are here, in this worldstate, it is fair to say that the Boomers have assets that don’t belong to them.
So you see @ConceptualJames, I’m actually even more anti-Communist than you. SS and Medicare are communist. Deleting them is capitalist. At least, old people on the dole shouldn’t be able to vote.
The grand standardization
I think you can even go deeper than this. Ideologies like nationalism have dissipated.
Give someone an opportunity to move countries to make more money and live a better life, most people will take it and not look behind.
The internet has created a global status market. You can see how people live everywhere, and compare yourself to people far outside your country or profession. That weakens the old national and occupational identities
squa retweeted
beyond 200 tok/s, Ahmdal’s pressure shifts back to cpu for running all those tool calls. the cup gets faster, good, the pressure shifts to all the underperforming & inefficient user space software. then, to the operating system itself.
high tok/s implies rewriting everything
squa retweeted
Dear BitMEX Users,
Today, we share with a very heavy heart that BitMEX exchange will shut down its operations, effective 23 September 2026 at 04:00:00 UTC.
The owner and operator of BitMEX, HDR Global Trading Limited, has made the difficult decision to close operations following a strategic review of the business.
It may not look the same today, but we are proud of our 11+ year legacy and the role we played in shaping the crypto industry. We invented the 100x leverage perpetual swap, which for most of you, was the first step to your crypto trading journey. It is now the most traded financial product in the crypto industry, adopted by thousands of users and exchanges. And we remain proud of our robust security infrastructure, which has allowed us to maintain a flawless track record of 0 customer funds lost to hacks in our entire operating history.
We want to reassure you that your assets remain fully safe and under your control during this transition period. This announcement is just to give enough time to ensure a smooth withdrawal process for everyone.
From today we strongly encourage all users to close their positions and withdraw their funds as soon as convenient. For more details on the full process, please read our blog: bitmex.com/blog/bitmex-closu…
BitMEX was once home to some of the greatest traders today. Our team has dedicated tremendous effort and passion into building the platform into what it is, and we are glad to have reached some of you during your time with us. To everyone who has traded, supported, and grown alongside us - thank you for your trust over the last 11 years.
The BitMEX Team
squa retweeted
In 2015 I wrote an interview question for Jane Street based on a real task I had given to new engineers. The problem was ostensibly simple but few candidates did well. The question was retired a few years ago so here it is, with related thoughts on LLMs building trading systems.
US securities exchanges have Designated Market Maker programs to ensure liquidity for their ETFs. The programs require that MMs maintain two-sided quotes 100% of the trading day, where each quote must be within 8% of the best-bid/offer (BBO). Satisfying these program requirements confers benefits to the MM, including improved capital treatment and the ability to short ETFs without locates.
The interview problem set up a hypothetical in which, in addition to the 8%/100% requirement above, the programs also require the MMs to be within 4% of the BBO 90% of the time, and within 1% of the BBO 70% of the time.
The problem asked the candidate to write code that could read a file containing a time series of market BBOs and another of the firm’s BBOs and produce a report measuring compliance with the MM requirements. Bonus points were awarded if the majority of the code could be reused for real-time reporting in addition to a file-based historical analysis.
Unlike what many expect of a Jane Street interview, this problem didn’t involve probabilistic reasoning or advanced computer science. It required a careful understanding of the data domain, including how to handle time boundaries, the optionality of market data prices, ordering inconsistencies, and online record keeping.
The pass rate was very low. We assumed that most candidates either hadn’t encountered enough real-world software in college/grad school, or else prepared too narrowly for their interview with questions about algorithms and computational complexity.
My benchmark for testing coding abilities of frontier LLMs is to ask the models to build infrastructure for a market making system from scratch. Even the most recent models, including Fable and GPT-5.6, do not produce systems with design patterns that are commonplace at HFTs. Even though the models’ outputs function in a narrow sense, the code architectures would not be able to form the foundations of even a modest production trading operation.
A production trading system is an institutional artifact, a representation of how a firm thinks about markets, failure, ownership, observability, deployment, and risk. Few of the inputs are ever written down in one place, some aren’t written down at all. The knowledge is distributed across code review norms, incident histories, operational procedures, and the accumulated instincts of experienced engineers over decades.
This helps explain why public LLMs can’t create effective trading infrastructure from a short specification, and why the interview candidates had particular difficulty with the question. A private LLM fine-tuned on an HFT firm’s repositories, incident reports, design documents, and review history might become highly capable within that firm’s engineering environment, but that’s far from how general-purpose LLMs are trained and operate today.
squa retweeted
It finally finished, here's the final output. Used 60% of my monthly Kimi usage on it
macos27.kimi.page
squa retweeted
kinda funny you can draw a smiley face in your neural net before training and it’ll be there afterwards
you can also use photos. i trained a MNIST classifier initialized to my face and you can still see me at the end
works across inits, weight decay, LR, optimizers, see below
