lonesword

I work at a frontier lab and have rejected Meta interviews (for much higher salary) many times over the years

Good for you, but this is not the counter example to the wager that the parent proposed. It would be "I worked for a no-name company in a developing country and still turned down interviews from Meta".

You can easily put stablecoins in a Lulo savings account and get 5% interest instead of 0.1% or whatever your bank provides. Yes Lulo has insurance.

From Lulo's site[1]: "Lulo’s yield comes from interest paid by traders and borrowers in integrated DeFi protocols. These loans are over-collateralized with assets like SOL, ETH, and BTC, reducing lender risk."

SOL, ETH and BTC as collateral? What if their value goes down? We know what happened when the banks made bad housing loans (2008 sub-prime mortgage crisis). At least the houses had some tangible value - bricks and mortars. Crypto seems like a fiat currency minus the "full faith and credit of the United States government".

1234.56 in PyUSD means you get 1234.56 in Chase or Wells Fargo or whatever. In future your bank will hold these assets directly without need to off-ramp at all.

If the appeal of PyUSD is that you can convert it into equivalent USD anytime, why do we need PyUSD at all? What's the value-add, apart from low transfer fees?

[1] https://lulo.fi

The much needed immigration should rather come from countries with similar society and culture to us Europeans, rather than India.

The mass emigration from India is a direct consequence of India's poor wages and living standards. If that was not the case, most people I know (and I myself) wouldn't have emigrated. From what I see[1], the average South American is much better off than the average Indian. Maybe that (and India's huge population) explains why South Americans do not emigrate as much[2] as Indians?

In other words, people from "countries with similar society and culture to us Europeans" may not want to move to Europe. It's all supply and demand at the end of the day

[1] https://en.wikipedia.org/wiki/List_of_countries_by_GDP_(nomi... [2] https://en.wikipedia.org/wiki/List_of_countries_by_net_migra...

Video games are far too derivative to be taken seriously as art

Even something like Europa Universalis IV? How is that derivative? I know of no other art form that will feed my fantasy of turning a no-name island in the Indian Ocean into a globe-spanning trading empire.

Hyperoptic is niche? I thought they were available all over London.

I agree with you, somewhat.

But now you have taken the trouble to complain that you had to look up “most populous country on earth” and yet contributed nothing to the discussion.

Pivot to being a student again :)

I come from the world's most populous country.

Realistically, moving abroad is tough but I want to! I applied to a lot of startups but not getting any positive response.

I was in a similar situation. After being bored doing CRUD, I applied for a masters program and moved to Germany in 2019 (I was 26 at the time). One of the better decisions I’ve made. There is no tuition fee (!!) in Germany for international students, and it is not that hard to get admitted to _some_ university (though actually graduating is hard). Computer science masters programs are usually in English. It is a fairly risk-free way to try out your hand at some research, and if you don’t like it, you can always find some software engineering job.

It’s not going to be easy - university applications are a pain in the ass, you will still have to learn some German e.t.c. But it’s a good way to “reset” your life.

I hate leetcode and that keeps me away from most of the job opportunities.

One way out of this is to interview deliberately at smaller companies.

If I don’t like my coworkers, I’m out. I like people who give a shit about the work I do (a.k.a “collaboration”). “Whatever, you own it” is just lazy for “I can’t be bothered to understand what you are working on”.

The problem is not the collaboration, it’s the ineffective collaboration. Maybe the author should fix that instead of pitching magic anti-collaboration click-bait pills.

Also, the problems you solve must be pretty straightforward and unambiguous (or you have a small codebase) if you have the luxury of being able to just make a pull request for it.

I moved from Germany (Cologne) to London recently and I am surprised to read that UK has safer roads than Germany. I have not yet driven in London - but I have driven in Germany and I have bicycled in both countries. London is hands down terrifying to me. Certain kinds of drivers (eg: vans) seem deliberately hostile to cyclists, and plenty of people park in a cycling lane. Never seen that happen in Germany with such frequency. Also, roads are bumpier - what’s with that?!

Now, cyclists in London are a driver’s nightmare. I’ve seen people barrel down a junction at full speed and jump a red light. This lack of desire for self-preservation is startingly common.

Between the hostile vans and the daredevil red-light-jumping cyclists, I am baffled at this report. Perhaps mortality rates are low, but injury rates are much higher than say, Germany?

Facebook is widely hated and embarrassing to work at.

Not sure it's widely hated (disclaimer: I work there), despite all the bad press. The vast majority of people I meet respond with "oh how cool!" when they hear that someone works for the company that owns Instagram.

"Embarassing to work at" - I can count on one hand the number of developers I've met who would refuse to work for Meta out of principle. They are there, but they are rarer than HN likes to believe. Most devs I know associate a FAANG job with competence (correctly or incorrectly).

Could Facebook hire away OpenAI people just by matching their comp?

My guess is some people might value Meta's RSUs which are very liquid higher than OAI's illiquid stocks? I have no clue how equity compensation works at OAI.

I work in this space (pretraining LLMs). It looks fancier than it really is. It does involve wrangling huge ymls and writing regular expressions at scale (ok I am oversimplifying a bit). I should be excited (and grateful) that I get to work on these things but shoddy tooling takes the joy out of work.

I have used both aerospace and yabai, and aerospace was far more simpler to configure. I really like it. Thank you for the great work!

I had to unfortunately stop using aerospace due to some bugs. If I switch from workspace a to workspace b, sometimes aerospace keeps switching back to workspace a. It is annoying when you are trying to press the unmute button in the middle of a call but aerospace keeps taking you away from that workspace. The root cause could be that the apps that I use (internal tools) are misbehaving. I'll file a bug report if I can reproduce it on my personal mac.

I work with pyspark and parquet quite a lot. I never had to deal with parquet outside spark, but this is how I would do this:

- Write a pandas_udf function in pyspark.

- Parition your data into smaller bits so that the pandas_udf does not get too much data at the same time.

Something like:

```

from pyspark.sql import SparkSession

import pyspark.sql.functions as f

@f.pandas_udf(return_type=whatever)

def ingest(doc: pd.Series): # doc is a pandas series now

    # your processing goes here -> write to DB e.t.c

    pd_series_literal = Create a pd.Series that just contains the integer 0 to make spark happy

    return pd_series_literal

df = spark.read.parquet("s3 path")

df = df.repartition(1000). # bump up this number if you run into memory issues

df = df.withColumn("foo", ingest(f.col("doc_column"))

```

Now the trick is, you can limit how much data is given to your pandas_udf by repartitioning your data. The more the partitions, the smaller the pd.Series that your pandas_udf gets. There's also the `spark.sql.execution.arrow.maxRecordsPerBatch` config that you can set in spark to limit memory consumption.

^ Probably overkill to bring spark into the equation, but this is one way to do it.

You can use a normal udf (i.e `f.udf()`) instead of a pandas_udf, but apparently that's slower due to java <-> python serialization

Genuine question - why assign seats at all? In my country (weddings with 200+ people are the norm) people sit wherever they want. In the US, you wouldn't assign seating at a house party (right?), so why do that for a wedding, since a wedding is after all just a more expensive party?

Add the name of the institution you went to to your resume and add "dropped out due to health reasons" in tiny (but still readable) font underneath. This will get you past most automated HR filters without you having to be dishonest. Since you already have some experience, I do not think anyone will bring up your education in an actual interview unless you bring it up.

Context - I'm from India, and works at big tech. I do have a degree, but no employer has every asked to see it. They just want to know that I _went_ to a college, doesn't matter if I graduated or not.

What can I do that would make you look past my lack of educational qualifications?

Your own one-person software-as-a-service business or relevant side projects. Hard to argue that you are not qualified when you can point them to github/website and say "I made this".

I turned 30 this year and had a bout with RSI. My elbows hurt, there was pain in the back of my hand, and my little finger even started going numb if I use the phone too much. What helped:

- I got the logitech MX vertical mouse. This removed the pain in the back of my hand

- After being skeptical for a while, I got myself the logitech k860 ergonomic keyboard. It helped.

This is of course not a long term solution, but it might help you relieve pain almost immediately. Thought I'd leave this here in case someone was one the fence about buying an ergonomic mouse/keyboard.

No, it' not that low for mid level engineers. Here's some anecdotal data:

An SDE II offer from Amazon (Berlin) was 90k base + 15K sign-on bonus + some RSUs. An SDE I (new-grad ish) at Amazon makes 72K euros + RSUs (at Aachen).

Zalando offered 75k for an Applied Science role (Berlin) for someone with a masters degree and 3 years of web-dev experience before the masters.

Delivery Hero (Berlin) recently hired a friend of mine. His base pay was 84k euros. He had ~ 5 years of experience as a back-end dev.

Some one I know who defended their PhD in 2022 was offered 90k (base pay, no variable component) at Porsche.

I know about a recent offer from a translation comp any based in Cologne that was close to 80k euros (base) for an SDE.

New grads still get offers in the 45-55k range, especially in smaller cities. A friend of mine just received a 57k offer from Deutsche Boerse (Frankfurt) for a non-SDE technical role. Honestly, it's not bad. 57k is approximately 2500 euros a month after taxes. Assuming that you pay 1000 euros for rent, you are still left with 1500 euros. Groceries shouldn't be more than 500-600 euros a month.

I wrote tweetscreenr.com to fetch articles from my twitter feed, and convert it into an RSS feed. The initial use case was to follow AI researchers on twitter and get notified whenever they post a link to an arxiv.org paper. Works well for me - I get all the news from Twitter, minus all the politicized/opinionated crap.

No finetuning for just a few facts could indeed end up being very costly. If you have 5 new examples that you want to fine-tune your model on, you probably wont fine-tune your existing model for 3000 training steps on just those 5 new examples. You'll either mix in other data to prevent catastrophic forgetting, or you'll probably training from scratch after fixing your dataset to reflect the 5 new examples you have.

Just deployed a fix and it seems to be working now. Thank you for your patience.

When the user visits `/settings` I fetch all the lists from twitter and BatchWrite them into DynamoDB. Turns out you have way more than 25 lists, and the BatchWrite fails. Let me see if I can quickly give a fix. Thanks for reporting this :)

Thank you for the kind words and for upgrading! Let me look into why the settings page is throwing an error.

I love the size of that case. I've been looking for a tiny case that can fit a GPU. This doesn't look like that though. It doesn't have to fit the largest GPU - I'd be happy if it can fit a gtx 3060. Does anyone have recommendations?

This[1] is the closest I could find and it fits a 4090. But it's still biggish and costs $280, which is a bit too much to spend on a case.

[1] https://formdworks.com/collections/t1-1/products/t1-sandwich...

I have an Onyx note air (2?) and it's been wonderful so far. I use it to read research papers, something that's impossible on my kindle. It's not just the screen size - the onyx boox has options to segment a 2-column paper into 4 quarters so that each page is a quarter of the pdf page and fills up the screen. This is something the kindle _could_ have done, but refuses to.

Google workspaces in India costs 125 rupees a month. That's around $1.5. Use a VPN + an international debit card (or wise.com) to pay Google in rupees? Is there a real downside to this?

Are you saying that the USA is still far behind the EU banking system? That's news to me - I've always assumed that banking, of all things, would be far superior in the US. I've always imagined that the experience in the US, at least compared to German banks, would be one of slick apps and super-friendly paperwork because 1) banks want your money 2) something something capitalism.

When I want to do a domestic transfer in Germany, if the recipient's bank does not accept SEPA Instant Pay (looking at you, Baader Bank), then it takes around 2-3 _days_ for my funds to get to the other person. That's insane. In India, our slow transfer (It's called NEFT. UPI is the new shiny instant transfer system) takes at most a day and even before UPI no-one was using NEFT because there were faster alternatives. I've always thought that the slow fund transfers were a German thing and that the US would be better. Guess the grass is not greener on the other side.

I think most of the comments assume that machine learning engineering equals machine learning. This is always not true. If you are an MLE, chances are that you work on the infrastructure and data pipeline side of things rather than on the research. If you are lucky, you get to productionize a prototype, which might involve re-writing it, or you might get to build an in-house hugging face. You absolutely do not require a phd to do this.

Then there are these other roles which involve prototyping new ways to train a model, or taking a paper from 5 months ago and see if it works for your use case. Or you know, just work on something that you can eventually publish. At FAANG, these are usually the "Research Scientist" or "Applied Scientist" roles. Most of these require a phd, but it's completely possible to get an offer with just a masters (I did), and I know of at least one case where the person "only" had a bachelors (and some experience). But by far the most straight-forward way to break into these roles is to have a phd.

our essential freedoms just so we can live in a "safe" country like... China and India?

India has zero to no gun violence. Sure, there's a lot of petty crime, and there are enough murders happening - and the last thing we want is to arm those murderers with guns.

Your sentence seems to argue that the (much higher) degree of free speech enjoyed by Americans is owing to civilian gun ownership, which I don't think is true (though I don't have the data at hand to back up my claim).

There are hundreds of millions of americans. These mass shooting events are insignifigant in the grand scheme of things

It's not just the number of people who get shot - it's this idea that a road rage incident _can_ potentially end in you getting shot, if the other person is deranged enough. Not to mention the amount of American tax dollars that go into ensuring that the police out-gun the civilians.

From the wikipedia link:

The Bangladeshi Declaration of Independence was broadcast from Chittagong by members of the Mukti Bahini—the national liberation army formed by Bengali military, paramilitary and civilians. The East Bengal Regiment and the East Pakistan Rifles played a crucial role in the resistance. Led by General M. A. G. Osmani and eleven sector commanders, the Bangladesh Forces waged a mass guerrilla war against the Pakistani military.

The Bangladeshi military organized the resistance. It was not a simple civilians-with-guns-rising-against-tyranny scenario.

Indian living in Germany here - just came back after 3 weeks in the US.

the quality is much better (especially food and education), the infrastructure is head-and-shoulders above (especially the roads, internet, phone plans, environment, litter, buildings in general).

I was there with you until you said internet. Internet and phone plans in Germany are painful. Most of the major providers are still on VDSL - though they are rolling out optical fiber. My current phone plan from O2 is 10 euros for 3.5 GB a month - that data volume is peanuts. I do not know if the US is any better, but I used to get 1 GB _per day_ in India.

buildings in general

Oh this. So much this. I was surprised how much more sound insulated and better-fitted the German buildings were. This is also reflected in the prices - buying a house in the US is cheaper than doing so in Germany. Here in Cologne, I see ads trying to sell very normal looking apartments (decent location, not at the center) for upwards of 600k euros. I mean, who can afford that in Germany? Meanwhile, I remember walking around in DC and seeing ads where the normal-looking apartments were all between 400-600k USD, which is actually affordable adjusting for income and purchasing power parity.