HNHacker News
TopNewBestAskShowJobs

dxbydt

4,893 karma · joined April 20, 2011

submissionscomments
dxbydt··on Autocracy and Stagnation: How Imperial Exams Shaped China's Destiny
> trivia is popular here in the US too

When you interview at Blackrock in SF, they walk you thru a bunch of rooms. Each room has a door. On each door is the name of a famous economist. When the interview begins, you have to recollect all of these economists and then drop some tidbits about each of them. You know, to signal that you are part of the in-group. Like, you can say hey wasn’t Krugman at Princeton when he got the Nobel but doesn’t he teach at cuny now because that’s where your phd advisor saw him last week hint hint. Now, would you rather do such obsequious ass-kissing, or remember that ALU stands for arithmetic logic unit ? Compared to Blackrock, Railway Board is god, son and holy ghost rolled into one.

dxbydt··on Autocracy and Stagnation: How Imperial Exams Shaped China's Destiny
There are more sailors than ships. In this context, something like 10000 applicants for a single rrb vacancy. You don’t want the railways to be nepotistic. Nor do you want legacy admits. Nor money power to dominate. What’s left ? Well, how about - leetcode has 6720 permutations. Find the 100th one by hand, when ranked alphabetically. One of my friends got this question last week at a bog standard Java IT job, in Bangalore, with over 1000 applicants. Its a fair question, just a little bit of math and arranging alphabets. Doesn’t mean he has to devote itself to handcomputing permutations for the entirety of his life. Its one day of interviews and you get to know if you are selected by 4pm same day. Perfectly fair imo. The standard gs interviews during my time took 12 separate rounds with 12 teams, were completely non-transparent, and you spent a whole month not knowing whether you are in or out. I would rather permute leetcode. Similarly, let the railway board applicant sweat a little. Its a government job for life with handsome pension to boot.
dxbydt··on Autocracy and Stagnation: How Imperial Exams Shaped China's Destiny
> only way to pass is to know someone who can give you the questions in advance

RRB is remarkably transparent. You can download past papers going back two decades & study them. There are multiple youtube channels that solve all of these problems & tell you the strats explicitly. Current events & GK culture is a thing in India. Lots of common folk pride themselves on knowing the answers to these questions. I wouldn't consider this gatekeeping by any means.

dxbydt··on Why Is Britain Poor?
> but it could be anywhere. > It's only inertia that keeps it in London

Not at all!

Over a decade in banking, I have witnessed this argument resurface time and again in the context of the US. Indeed, it's true that two of the nation's largest banks, Wells Fargo and Bank of America, are headquartered in Charlotte. Hedge funds predominantly anchor themselves in Greenwich. GS has relocated its back-office operations to Salt Lake City. JPMorgan conducts risk analytics out of Columbus. CC processing is entirely Wilmington, commodities desk entirely Chicago.

However, despite these geographic shifts in various financial functions, the heart of the financial world remains unmistakably tethered to Wall Street. For the foreseeable future, the nexus of global finance will continue to pulse through New York City. To suggest otherwise is a fundamental misreading of the industry's deeply entrenched culture. Just as in London, where finance is inextricably woven into the fabric of the city, New York represents more than just a place—it is the embodiment of finance itself. The gravitational center of this world does not and will not shift to places like Utah, Ohio, Delaware or North Carolina. It is, and will remain, rooted in global financial hubs such as New York, London, Mumbai, and Hong Kong. To imagine otherwise is an exercise in futility.

dxbydt··on Posting Is the Most Powerful Force in the Universe
if it really was the most powerful force, the button I’m going to press below should say “post”. Instead it says “add comment”
dxbydt··on Foundations: Why Britain Has Stagnated
must commend this briton for his unparalleled sense of humor, razor-sharp wit, subtle, effortlessly charming and perfectly timed quips. While American humor tends to be more direct and sometimes louder, this dry, understated delivery is a unique gift.
dxbydt··on Memorizing the first 100 perfect squares (2022)
45^2 = 2025 => must be the sum of first 45 odds => 1 + 3 + ... + 87+ 89 = 2025

while its not pythagorean, 40^2 + 20^2 + 5^2 = 2025.

877 is another bullcrap that has shown up on a bunch of these tests. 877 is a prime, and four times 877 is 3508 = sum of all the divisors of 2025.

there's a bunch more i shared with them.

dxbydt··on How we made Jupyter notebooks load faster
> In a Jupyter notebook... > Often, Jupiter notebooks...

Everytime I search my Slack, I have to run two searches because DS can't agree on how to spell the damn thing.

dxbydt··on Memorizing the first 100 perfect squares (2022)
I give amc & mathcount mocks to kids every weekend & time them. There was one question everyone got under 5 seconds. I was like how did you guys do it so fast. Apparently aops had told them every year will have a special property that'll forsure be on the test. So that's why they all knew 45^2 was 2025.
dxbydt··on Ask HN: Is the Job Market That Bad?
> Small part time gigs, some substitute teaching, bar-backing and construction work.

Was in the same boat.Did the audio-to-text transcribing gig on amazon mturk. It was a tossup between that & odd jobs at Walmart at that time. Recessions in usa is pretty brutal.

If you are so inclined, I suggest getting a cert & then hitting the shady indian outfits. They always have jobs so long as you have a recent cert. Can be anything - cert in GCP, or AWS or some cloud infra work will do. You have to study...but you have the time.

dxbydt··on James Earl Jones has died
Saw that movie in the theater in my early teens. When he turns into a snake - watching that on a 30 feet high screen from the front row - scarred me for life. https://www.youtube.com/watch?v=GF0Z5g0Wjuk
dxbydt··on 'Unschooling' parents put their kids in charge of their own educations
A lot of us have had a pretty bad experience.To say we are all projecting is the real joke. I live in a college town & work with several kinds who are >2 std dev of your median cohort. Kids with a perfect ACT score in math at the age of 10. Kids who can do ap calc problems by age 12 - so can do limits, differentiate, integrate, handle tangent vectors, the whole calc1-calc2 shebang. To all of these kids, the system is a complete joke. Utter waste of their time. You don't even have to trust me. Get on aops. They have student blogs. Read thru the student blogs. Kids tarnish their schools, school textbooks & math teachers all over the place. I would happily link them here but don't want to invade their privacy.
dxbydt··on Have ‘hobby’ apps become the new social networks?
on the contrary, most relationships got started only because both partners enjoy exploring each other’s bodies. any other activities they might have in common is some afterthought that is tacked on after the relationship acquired a stable footing. i still don’t have any activity in common with my wife, and we’ve spent 2 decades & change…
dxbydt··on Kids Should Be Taught to Think Logically
> the hard part is what to do about it?

After 8 years in a public school, I've decided enough is enough. I've attended numerous school board meetings. Written to teachers. Given generously to the teacher funds. Voted. Nothing works. So we've saved up some money. We'll be putting our kid in a private school starting next Fall. It's hard, yes. It's hard on him - he's going to miss his buddies. It's hard on us - $40000 per year for high school, thanks America! I honestly don't know what else to do. There's a second private school in the area that charges $65000 per year, so perhaps I should console myself that I'm being responsible by saving $25000 per year.

dxbydt··on Kids Should Be Taught to Think Logically
As a statistician, perhaps the most useful feature of statistical thinking is an explicit rejection of point estimate. Always, always, always, give an interval estimate, or a five point summary if you have that. This was the one lesson that was drilled into me repeatedly.

Ofcourse, you head out into the real world & it is full of point estimates! The article gives a very good example of real-world statistical illiteracy - "If the gas mileage of a car is 40 miles per gallon, and I drive 20 miles, I will have used half a gallon of gas". While I'm sure that the HN audience prides itself on statistical literacy, they will certainly have trouble coming up with a good interval estimate for that scenario.

dxbydt··on Kids Should Be Taught to Think Logically
the way things are headed in the US, “kids should be taught” is itself an aspirational goal. “think logically” would be icing on a nonexistent cake. my state is unwilling to pay more than $21 per hour to a teacher. every attempt to add even a dollar to this figure has been overturned. first “congress should be taught” what a reasonable wage is.
dxbydt··on The first nuclear clock will test if fundamental constants change
extraordinary. you’ve demolished his argument with purely dimensional analysis. bravo!
dxbydt··on What's functional programming all about? (2017)
I prefer to make a distinction between recurrence, identity, and formula. Otherwise the whole discussion is moot. For example, to compute the nth fibonacci, we have Binet’s formula. We don’t need to bootstrap - we just take the golden ratio and its conjugate, exponentiate, take the difference and scale by root 5. That’s a genuine formula. otoh If one says the nth fibonacci is just the sum of the previous two fibonaccis, that’s technically not a formula - you would need to bootstrap to get those previous ones, so that a recurrence - still very useful, but not a formula in the sense you can’t turn the crank and be done. In that specific sense, I agree with Wilf and Dudley that the procedures outlined by Mills, Wright, Wilson etc are essentially worthless as formulas. They are nice identities and recurrences, but not a genuine formula in that you can’t get the nth prime by turning the crank - you have to bootstrap with so many extraneous params it makes the whole exercise futile.
dxbydt··on Physics is unreasonably good at creating new math
i am essentially a rayiner on steroids, in that i believe america has absolutely nothing meaningful to offer to the world at large other than aops and rsm. its such an extraordinary institution. large number of international signups - so tons of indian and chinese kids burning the midnight oil solving woot alongside their american compatriots. just an amazing, spectacular creation all around. if the nobel committee had any brains, rusczyk would have gotten atleast a peace prize by now.
dxbydt··on What's functional programming all about? (2017)
> Functional programming is about finding the formula for a particular program.

Can be disproved trivially. Anyone can code up a program that generates the nth prime for some n, by iteratively accumulating n primes starting from 2 using trial division. otoh, an actual formula that produces the nth prime would be an earth shaking event.

dxbydt··on Physics is unreasonably good at creating new math
I have zero issues with your logical reasoning - its the same reasoning aops gives. I'm simply asking - how do you know for sure that this actually happens in practice ? Like, have we observed such a thing ? I'm genuinely curious, no snark.

If all we have is reasoning, with zero observation, then its equivalent to what JFK says in Oliver Stone's picture - "Theoretical physics can prove that an elephant can hang from a cliff with its tail tied to a daisy."

dxbydt··on Physics is unreasonably good at creating new math
> pontificate about things that are too small or too dark and far away to see

I was bitten by this last week. I am enrolled in an aops physics course, titled Mechanics. So the last time I took any Mechanics was 40 years ago as a high school student in India. Most of the curriculum then was about stuff banging into each other aka collisions, & asking what happens to the result. Like some golf ball rolls down an inclined plane at some angle theta & hits a identical stationary ball & the objects stick together & we're supposed to compute where they end up. I was curious what American students learn, so I enrolled in this aops course.

Last week's assignment asks me - under which scenario will conservation of momentum make an accurate prediction. The 3 scenarios are - truck collides with car, eagle comes to rest at perch after flying from far away, and two galaxies colliding into each other to form a third megagalaxy.

I naturally picked truck & car - so aops knocks off 2 points for the wrong answer! Apparently if a truck collides with a car there will be so much thermal energy produced by the friction of the road, any prediction you make about the final velocity of the car assuming conservation of momentum will be bogus.

So then I pick eagle coming to rest - aops knocks off 2 more points for the wrong answer! Apparently when eagle comes to land, it will open its giant wings to create air resistance, so momentum won't be conserved.

Ok so that leaves the 2 galaxies. I pick that & get my correct answer, a pathetic score of 3/7. I'm left wondering how do we even know this is correct. Galaxies are too far away to observe. How is one supposed to compute the mass of 2 separate galaxies, & then find their moving velocities accurately, & then find the final velocity of the combined galaxy, & thus confirm momentum was conserved ? Seems very far fetched. I would rather go with the car & truck.

dxbydt··on Abel, the Mozart of Mathematics
Those are 2 good examples. A lot of the old terminology is very powerful because it doesn’t try to be slick/pc. Saying ‘variance’ is slick but it doesn’t convey as much as dispersion. Statisticians also speak in terms of first raw moment which has a lot of physical meaning behind it - saying ‘mean’ doesn’t convey anything. We frequently use coefficient of variation in the old texts. It tells you right away that it has something to do with variation, so you can dig into it and find you are normalizing the dispersion by the expectation so basically how much variation for a given amount of expectation. Instead of saying useless jargon like sharpe of vtsax is 0.7 on a 3 year timeframe, you could as well have said a coefficient of variation of 1.4 - that tells you right away how much dispersion you can expect in your return, whereas by saying sharpe you are simply gatekeeping and putting william sharpe on a pedestal for what is essentially a very silly ratio that statisticians have used for centuries without any fuss. I notice that in all the Indian news channels, they have translated the word monkey and the word pox, and then combined the two translations into a compound word, which right away gives much more useful information than calling it mpox because of dei reasons. So the people speaking Hindi or Tamil or any Indian language know right away what the disease refers to, whereas saying mpox gives you zero info. The next generation will simply be told mpox and think thats an actual word.
dxbydt··on Engineering the First Fitbit: The Inside Story
> refused to allow users to download their data > instead of having it uploaded to FitBit's servers

I download my monthly data from the fitbit server into LibreOffice. I have about 2 million steps on my fitbit. I have csv files going back to 2019 ever since I purchased the fitbit. They've made a nice dashboard, with some badges. I share the dashboard with friends, bit of a social competitive streak. If I were to directly download the data from the fitbit to my ubuntu desktop, I would have to write a fileparser & then make a dashboard & all that...not interested. I want to walk, not do software in my offtime.

dxbydt··on How long does music stardom last? A statistical analysis
I know of atleast three software engineers who quit programming to pursue music fulltime.

This lady[1] got a Computer Science degree from UC Berkeley, as even an intern at one of the faangs I believe. Quit the whole thing and caught a flight to India. She's sung about 100 songs now and is quite famous.

This guy[2] got a Computer Science degree and worked as a Visual Basic programmer. Also wrote some FoxPro and Clipper code. Was working at Leading Edge in those days. Quit software and started singing. Has sung over 7000 songs. Very rich, 9 figures.

This guy[3] got a PhD in Computer Science and wrote an algorithm to reduce non-standard matrices to approx Hermite Normal. Wrote a bunch of NLP papers, even published in a pure math journal! Then quit software and started writing songs. Has written over 1000 songs now. Very busy songwriter.

[1]https://en.wikipedia.org/wiki/Manisha_Eerabathini [2]https://en.wikipedia.org/wiki/Shankar_Mahadevan [3]https://en.wikipedia.org/wiki/Madhan_Karky

dxbydt··on How long does music stardom last? A statistical analysis
> they didn't try to fit a distribution or two?

Since mu is 28, and 13 is the min, you have 28-13=15 = 3*5, which makes sigma 5, so use rnorm(n,28,25) to generate n samples from the gaussian. https://www.desmos.com/calculator/mlwm7mke16

dxbydt··on Intel stock dropping toward 50 year low amid restructuring news
One of the cnbc guys compared this morning's disaster to an emergency surgery - "they will live, but what sort of a life will they have..."
dxbydt··on UBI and the Anti-Work Vibe Shift
Sir, you write so well! I read it twice and then I had to literally put the phone away and walk around, letting your thoughts bounce around in my little brain. Just felt I needed to thank you for taking the time. Thanks.
dxbydt··on Math is running out of problems
While I agree with the author broadly, he is definitely overselling his thesis. Going from 'fewer grad students now care about problems in Journal of Algebra' to 'Math is running out of problems' is quite a stretch.

Take any period of time - some subfields will run hot & others will be fallow. Doesn't mean we have run out of problems.Trace formula for theta groups will appeal to only fifteen people - ok so what's the issue ? Math isn't some popularity contest. We have a ballroom at the university which is reserved for talks from visiting professors. When we have an economics lecture, usually it is jampacked. All 100 seats are taken, not even standing room. Then the next talk is by some topologist. The room practically empties out in real time. If you watched it live, you would be shocked at how fast people are rushing out of the room - you would think some stinkbomb was thrown. Finally, nobody is left other than the topologist himself & 5 grad students, 4 of whom look like they literally jumped out of bed & grabbed a coffee mug on the way. That's math for you. That's how its always been.

dxbydt··on Saving months of compute time with a single Grafana query
many of the tricks we learned in the late 90s - 2000s can no longer be pulled off. We used to download jar files over the net. Running a major prop trading platform meant 1000s of dependencies. You’d have swing and friends for front end tables, sax xml parsers, various numerical libraries, logging modules- all of this shit downloaded in the jar when the customer impatiently waited to trade some 100MM worth of fx. We learned how to cut down on dependencies. Built tools to massively compress class files. Tradeoff 1 jar with lots of little jars that downloaded on demand. Better yet, cache most of these jars so they wouldn’t need to download every single time. It became a fine art at one point - the difference between a rookie and a professional was that the latter could not just write a spiffy java frontend, but actually deploy it in prod so customers wouldn’t even know there was a startup time - it would just start like instantly. then that whole industry just vanished overnight- poof!

now i write ml code and deploy it on a docker in gcp and the same issues all over again. you import pandas gbq and pretty much the entire google bq set of libraries is part of the build. throw in a few stadard ml libs and soon you are looking at upwards of 2 seconds in Cloud Run startup time. You pay premium for autoscaling, for keeping one instance warm at all times, for your monitoring and metrics, on and on. i am yet to see startup times below 500ms. you can slice the cake any which way, you still pay the startup cost penalty. quite sad.

← PreviousPage 2 of 28Next →