HNHacker News
TopNewBestAskShowJobs

Mc91

549 karma · joined August 17, 2017

(defun mc91 (n) (cond ((<= n 100) (mc91 (mc91 (+ n 11)))) (t (- n 10))))
submissionscomments
Mc91··on Artificial intelligence is losing hype
I don't use AI at work at all.

I pay for Leetcode, which usually gives editorial examples in Python and Java and such, and paste it into ChatGPT and say "translate this to a language I am more familiar with" (actually I have other programs that have been doing this for some language to language conversions for years, without AI). Then I say "make it more compact". Then again "make it more compact". So soon I have a big O(n) time, big O(1) space solution to Leetcode question #2718 or whatever in a language I am familiar with. Actually sometimes it becomes too compact and unreadable, and I back it up a little.

Sometimes it hallucinates, but it has been helpful. In the past I had problems with it, but not recently.

Mc91··on A new type of neural network is more interpretable
I'm not deep in the field at all, I did about four hours of Andrew Ng's deep learning course, and have played around a little bit with Pytorch and Python (although more to install LLMs and Stable Diffusion than to do Pytorch directly, although I did that a little too). I also did a little more reading and playing with it all, but not that much.

Do I understand the Python? Somewhat. I know a relu is a rectified linear unit, which is a type of activation function. I have seen einsum before but forget what it is.

For the classical diagram I know what the nodes, edges and weights are. I have some idea what the formulas do, but not totally.

I'm unfamiliar with tensor diagrams.

So I have very little knowledge of this field, and I have a decent grasp of some of what it means, a vague grasp on other parts, and tensor diagrams I have little to no familiarity with.

Mc91··on Ask HN: How are you finding the job market in July 2024?
I am an SWE in the US.

In 2022 on Linkedin I got messages from recruiters at Google, Amazon, Uber, and many other companies, large and small.

I haven't gotten a message from any recruiter on Linkedin for over a year. Luckily I am working.

One data point, but it fits with what I've been seeing and hearing from others.

Mc91··on Ask HN: Is it just me, or does the job market for IT seem bad today?
November of 2022 is when the Facebook and Amazon had their big layoffs, and Google, Microsoft, and Salesforce etc. followed in the two months after.

I haven't had a recruiter contact me on Linkedin for over a year, in 2022 and before I was being contacted by recruiters on Linkedin all the time.

Companies have different modes - mass layoffs, small layoffs, no hires, only hiring replacements, a few new slots open, many slots open. In early-mid 2022 we were in the last category and it's moved to one of the other modes for most companies, except for LLM hires.

There are different sources of information, I found this helpful ( https://interviewing.io/blog/when-is-hiring-coming-back-pred... ) from a company with some insights into tech hiring.

Mc91··on Computer-science majors graduate into a world of fewer opportunities
I work for a Fortune 100 in one of the 20 biggest metros in the US, $75k is a standard salary for juniors/associates getting an offer right out of college. Some may get a little more, some get less, I haven't met any who got $100k or more right out of school. None of them I've talked to came from a top-tier CS school.
Mc91··on Show HN: I made a spaced repetition tool to master coding problems
You can find the problem described here. It also has a link to solutions in different languages.

https://leetcode.com/problems/invert-binary-tree/

Mc91··on Show HN: I made a spaced repetition tool to master coding problems
Right. I learned what binary trees were in college and didn't directly use them much afterward. Since interviews sometimes ask about trees, I started my practice of solving tree problems by practicing how to invert a binary tree. I practiced it so much that I can invert a binary tree from memory now, so that I won't get bounced from an interview like the inventor of Homebrew was from a Google interview.

Most of the concepts I learned, although I don't remember learning Kadane's algorithm for solving the maximum subarray problem in school, so I've learned some things like that. I can also implement a solution for an associated Leetcode problem in one line of code, so I wound up memorizing that too.

OP says "If you master the methods and ideas, you should be able to derive the answers on the spot", but I hear from a lot of people that does not happen, even (implicitly) in this thread. Any how, even if I use OPs method, if I do the same problem over and over and over again I'm going to wind up memorizing it any how.

Mc91··on U.S. podcast listening reach new high: 47% listened one last month, +12% YoY
As others have said - cardiovascular exercise, trips/commutes...
Mc91··on Are we all doomed? Why?
Why do I think it has declined but will go back up?

Because I've been following the field since the early 1990s and have been in it since the mid 1990s, and have seen it act as a sine wave, although the amplitude can change.

The US was in a recession until March 1991, and this affected the IT job market. Even out to 1993 it was slow. It began picking up in 1994, more in 1995, more in 1996, until 1999 when it was hotter than I've ever seen it. Then the dot com crash happened in 2000, and things were worse for the next three years, especially in IT companies. Then things began picking up until the 2008 subprime crash which affected the whole economy.

As others have alluded to in the thread, for people just coming out of college in spring of 2011 - interest rates were 0.10%. This is historically unusual. IT companies and IT were in a go-go time for a number of years up until the end of 2022, not seen since the dot-com days.

IT companies were hiring and paying like crazy in 1999. It was much more difficult to get a good-paying, or even any, job at an IT company in the Bay Area in 2001. By 2007 things had picked up a little. Then at the end of 2009 things were dire all over the economy. Then by 2021 IT was booming again.

So will it be booming again in the future like it was in 1999 or 2021? History says it will. Will it again be in the dumps like it was in 2001, 2009 and to some extent now? History says it will.

I don't see any historic change from what happened in the past, although of course some computer languages are falling out of favor while others are coming into favor, and job roles are shifting around - yesterday was monolithic, today is microservers, yesterday was waterfall, today is agile and scrum etc. People have been programming since the 1950s, and the job market has seen its ups and downs since then. Nothing happened in the past year and a half that is off of that.

It's possible neural networks will change things in the coming years, but someone is going to have to program all those Python pytorch scripts and the like. I'm sure in the next year or two we will see amazing new things like Sora come out. We'll even see programmers using LLMs or whatever to generate little functions more. The notion that neural networks are going to wipe out programmers in the next year or two (or three, or four) is inane. I can barely get ChatGPT 4 to write some simple (WORKING) functions I want, never mind gather specifications, ask for clarifications on those specifications, design a proper architecture for the project with unit testing and so on. This will surely improve in the next few years, but not by much in the next few years at least.

Mc91··on Are we all doomed? Why?
> Is this AI?

No

> Is it a temporary decline of the market?

Yes

> Caused by what?

That's a complex thing to answer, but the fed funds rate starting 2022 at 0.08% and now being 5.33% doesn't help.

> Is there still any niche where we can be employed faster as we could just a year ago?

AI

> Are we all going to die?

Epictetus said "it has been given to you for the present, not that it should not be taken from you, nor has it been given to you for all time, but as a fig is given to you or a bunch of grapes at the appointed season of the year"

Mc91··on Ask HN: Which job sites are you using to look for new jobs?
My current job started as a job at a consulting company, which I got via a recruiter reaching out on Linkedin.

Hacker News who is hiring threads work as well - posting your resume or reaching out to job listings.

I have got interesting interviews at startups via https://wellfound.com/jobs .

Mc91··on The unwritten rules of negotiating with Meta
> I found the whole process pretty dystopian

And you didn't even go to the office and see the Orwellian "MOVE FAST AND BREAK THINGS" and "WHAT WOULD YOU DO IF YOU WEREN'T AFRAID" posters, or the internal newsletters put over the urinals, so you can read about HHVM while taking a leak.

Mc91··on Tech Job Interviews Are Out of Control
> All that said, a majority of candidates can be filtered out within 15 minutes

If I interviewed six candidates, usually one would obviously be head and shoulders above the other five. If I interviewed twelve, two would be. I saw this across various companies over the years. About 15% clearly know more than the other 85%.

Also I found I usually could tell this within three questions. If I asked three questions, and the person being interviewed either stumbled over some of the first three, or had a shallow understanding of the topics, I never saw them recover afterward - if they had no or a shallow understanding on the first three questions, so it would be on subsequent questions. I could have wrapped it up after the third question for most people. Once I became aware of this, I still gave them a chance (which never bore fruit), but I began thinking how fast could I wrap up an interview where they had come in without seeming impolite.

Mc91··on So long, and thanks for all the bytes
> Everyone hates him, but you have to begrudgingly admit that Elon successfully demonstrated you don't need massive dev teams to build and run software, even at Twitter's scale. (obviously it has other problems due to Elon's incompetence, but nothing more developers would solve.) Most companies could cut their dev team size by 50% and end up net-positive due to less overhead--the challenge is figuring out which 50% to cut.

What do you mean this revelation was recently demonstrated by Elon? The NATO Software Engineering Conference 1968 paper talks about the problems of having too many programmers of low skill. The 1969 paper talks about the same thing. Fred Brooks wrote the Mythical Man Month in 1975, covering this topic extensively. This has been talked about extensively for over half a century with regards to software, it wasn't just discovered by Elon in the last year.

Mc91··on I'm an Old Fart and AI Makes Me Sad
> Personal Computers [80s]

> But what if OpenAI decides to revoke access to that API feature I’m using?

This starts with personal computers, and why would computers at that time be called "personal"? Part of the problem with this essay is it starts with personal computers - personal computers were called personal because before that was mainframes, which were the same kind of gatekeeping that a cluster of H100s in a data center would be today.

Computers started out as these centralized IBM mainframes, but in 1975 people could buy an Altair kit, which is the same year the MOS 6502 was released. There is some centralization in neural networks now, if that displeases people they can work to do the same kind of thing that MITS and MOS and Apple and even Microsoft did.

> Flipping all those numbers to get the result (inference), and especially determining those numbers in the first place (training), requires a vast amounts of resources, data and skill.

Using a Stable Diffusion model as my base, and a number of pictures of a friend, and a day or two's work on my relatively not-so-powerful Nvidia desktop card, I can now make Stable Diffusion creations with my friend in the mix now. This can be done by different methods - textual inversion, hypernetworks, dreambooth (I have been also told LoRa works, but have not tried it myself).

On the same relatively unpowerful Nvidia card I can run the Llama LLM - with only a few billion parameters, and with quantized less precision - but results are decent enough. I have been told people are fine-tuning these types of LLMs as well.

There's nothing inherently centralized about neural networks - although OpenAI, Nvidia, Google, Facebook, Anthropic and the like tend to have the people who know the most about them, and have enormous resources to put behind development. I'm sure something with the power of an H100 will become cheaper in the coming years. I'm sure tricks will develop to allow inference and even training without the need for a massive amount of VRAM - I see this in all types of places already.

If you don't want some centralized neural network monolith - do what the people at MITS and MOS and Apple did - do what people are already doing, figuring out how to use LLMs on weaker cards with quantization, figuring out how offload some VRAM to RAM for various Pytorch operations. Centralization isn't inherent to neural networks, if you want things more decentralized then there's plenty of things that can be done to achieve that in many areas, and getting to work on that is how you achieve it.

Mc91··on Interview with Andreas Kling of Serenity OS (2022)
> They aren't a load of fuddy-duddies who insist on using mailing lists and IRC. Nobody wants to collaborate with a load of grumpy greybeards

Sometimes the grumpy greybeard fuddy-duddies act as they do due to experience. One can have mailing lists which are very distributed and easy to put on a searchable web page somewhere. One could also tie everything to Github or Gitlab or Atlassian, and then have it pulled, or have rates jacked up, or whatnot. One can go on IRC, and easily migrate from Freenode to Liberanet if need be. One can also go into the walled garden of Slack or Microsoft Teams or some other proprietary format, so that some company can hike rates or pull the rug out at their convenience.

Mc91··on Flattr is closing down (2023)
I donate monthly to five Patreons. One of them is LineageOS ( https://www.patreon.com/LineageOS ). Two of them are community spaces. Two are podcasts, only one of which I listen to. I really only get Patreon benefits from one of the five (I get each podcast). Some tech-related, some not.

I subscribe to one Twitch feed. Actually the main Twitch feed I watch is one I do not subscribe to. I also sometimes watch other Twitch feeds, like John Romero's, or notch, or another random programmer who livestream codes. Again, some tech-related, some not.

Generally I subscribe on Patreon and Twitch more to be supportive than to get something, although I do appreciate I get podcasts from one of the podcasts I subscribe to.

Mc91··on IT employment grew by just 700 jobs in 2023
Same. I work as an SWE at a non-FAANG, non-tech Fortune 100 company. We already had an IT staff in India and would hire consultants in India, but this has been accelerating over the past year - virtually all hires or new consulting contracts have been in India. Some of the SWEs in the US are even being put under PMs in India.
Mc91··on Ask HN: How did you land your first job?
If you're in college as a CS major, look for a summer internship. Some companies are looking for people between junior and senior year, although some will take earlier. If they like you, you may get a job offer. Find out when they want you to accept an offer by, and see if you can line up other offers at that time.

If you're not in college it is tougher. Large companies usually cycle in junior people via those college internships. You can join a consulting company, as big companies are only committing for you to be onsite for three months or so, so are more willing to let you work for them via consultancies.

Small, pre-series-A companies don't have the money to hire senior people like big companies or VC-funded companies. So you may have a better shot there. You can look at angel.co which has some openings for different companies.

Mc91··on Ask HN: Is React Native still popular?
I look online for Android programmer jobs, I see React Native often enough when someone is looking for an Android or mobile programmer.

It's a large skillset people want - you need to know Javascript, then probably some HTML and CSS, then the React framework. Enough to get you a job as a React JS programmer - but then you should know React Native as well. Plus, RN does not cover every case, so know the Android ecosystem (Kotlin and Android SDK, plus maybe some Java) and the iOS ecosystem (Swift and iOS SDK, plus maybe some Objective C). Most large companies tend to have a separate Android and iOS app, although not all (Discord is React Native IIRC).

It makes some sense in a startup possibly - have a few RN programmers and have one or two dive into RN and do Android and iOS apps for your product, some of it using your existing codebase. From all I have heard, it does not scale, although maybe companies like Discord make it work for their use case. Once a Series A or series B kicks in, engineering has to stop and look whether they want to continue adding on to this rickety future, or get a small Android and iOS team together and do native apps.

Mc91··on My dream job wasn't worth it: Former employees (Meta, McKinsey, JP Morgan) share
This may be true, but there tends to be more complaining about Amazon in the FAANGs than the others.
Mc91··on "No, it's less effort than that"
Correct. If management told our team that our performance was based on getting 2-3 planned features out this quarter, and we've been minmaxed to fill our schedules, and there's a hiring freeze and one team member just left - then we have no room for some "30 minute" (which is much more than 30 minutes for us) request. Also, management has told us that our performance reviews are based on whether we got these 2-3 planned features done within a very aggressive schedule, not your "30 minute" request - so it is management who has said this is of minimal importance, not us.
Mc91··on Can't be fucked: Underrated cause of tech debt
> I think software as a whole suffers greatly from this "well, I got it barely done, technically fulfilling the requirements, so my work is over" attitude.

I get it barely done, technically fulfilling the requirements. How much more I clean up depends how much more time my PM has left me in the sprint. Since I'm constantly bombarded by questions, appeals for help, requests to approve pull requests, last minute changes, various meetings etc., that usually is not much, if any, time.

Mc91··on Ask HN: What do you like to see in tech talks?
Don't put up multiple slides with long lines of code. A few slides with a few lines of code where everything is very clear is fine. Some people just have every slide be many lines of small code. In terms of slide text, less is more.
Mc91··on Epic Games is laying off 16% of its staff
If ChatGPT should be able to generate any code, a function that outputs a set of prime numbers longs in Kotlin should be simple enough (unless you consider Kotlin, the 12th-20th most popular language according to various surveys, too obscure). One that has a good big-O runtime that is compact as possible. After a bit of futzing I get this:

fun sieve(n: Long) = (2L until n).fold((2L until n).toMutableList()){p,i->p.apply{for(j in i * 2 until n step i)this[(j-2).toInt()]=0}}.filter{it!= 0L}

and it works OK in some situations:

>>> sieve(100L)

res5: kotlin.collections.List<kotlin.Long> = [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97]

Let's try another test input -

>>> sieve(10000000L)

java.lang.OutOfMemoryError: Java heap space

Oops.

Also if the 12th-20th most popular language (Kotlin) being too obscure is the reason, how will they fare currently except with the most popular languages?

This is about as simple as a function as you can get. 95% of my programming work is standups, sprint planning and grooming, looking at Swagger or other API backend documentation for information, looking through Outlook, Teams and Confluence for information about my story, looking for UI on Figma, contacting various teams for more information - then someone wants me to review their pull request and stop to do that - but notice they have some business logic wrong because they don't fully understand the domain. Yet even something simple like a function generating a list of prime numbers, ChatGPT struggles with.

Not that it can't get better, but it's no threat to most programmers I know in the short term (and if it becomes one, I'll shift over to writing Python scripts that call Pytorch, numpy etc. - I already do so in my spare time).

Mc91··on A world where people pay for software
> If you can't monetize you can't put people on your software and it will suck (like most OSS software, for example).

I have worked on FLOSS software and I have worked on non-FLOSS software and I don't see most FLOSS software sucking in a way that non-FLOSS does not.

FLOSS has some advantages - as there is no compelling need to release new features which can drive up revenue and profit (or at least OKRs) for the next quarter, you don't get a constant need to release unneeded junk to try to squeeze the last dime out of consumers. You can actually spend time refactoring the code, or only releasing when it is properly architected.

Most of the servers and smartphones in the world are running on a FLOSS kernel. MacBook's OS derive from CSRG's BSD, and even some of Windows, like the Internet stack, derive from FLOSS. If it sucks so much, why do virtually all major operating systems derive fully, or at least partially, from it?

Mc91··on 2500 Job apps, 46 interviews, 1 offer
I live in a US metropolitan area with 2-3 million people, at a Fortune 500 company.

Junior/associate devs, hired out of the internship program usually, can make less than $100k TC (some even make less than $80k). Most of them say it is hard to live on that salary with an apartment, car, student loans and wanting to have a life.

Most of the standard SWEs make more than $100k TC. I've been told the TC for standard SWEs maxes out at under $170k at our company.

Seniors TC starts around $140k, and goes up above $180k.

Leads/Principals I'm not as sure of. Maybe for every 20 associate/SWE/senior we have one lead. I believe their TC is north of $200k.

As the standard dev is either an SWE or senior SWE, that band is probably $100k-190k. With some associates/juniors making less, and some leads/principals making more.

The H1Bs make about 10-15% less than the non H1Bs.

In India, if our programmers make the equivalent of $30k that is a lot. The company wants to hire more Indian programmers but there are timezone problems, language problems, skill level problems and other problems.

Mc91··on Famous Programers with Repetitive Strain Injury (2022)
I don't really have problems with RSI nowadays, but years ago I would be at the computer even more.

One helpful thing is switch the mouse between hands every few minutes. Your hand gets a break if you're moving the mouse a lot. I used to have a normal mouse at work and I bought a trackball for home, so that the hand muscles used would be a little different.

Someone in this thread mentioned Jamie Zawinski and his advice is to not ignore pain. So if you are typing for an hour and your right hand starts to hurt, stop using your right hand for a little while (as long as you can, really).

Mc91··on Ask HN: What's your opinion on weekly 1:1s?
I think this is it, an hour a week can be too much in some cases, but never having 1-on-1s can lead to some things deteriorating, depending on the manager. Otherwise trivial issues accumulate and become non-trivial, both sides can get out of the loop etc.
Mc91··on Ask HN: How can a junior dev maximize their income?
> ML+Python+Pytorch because it's hot right now?

Watch this video, which is an introduction to logistic regression for beginners - https://www.youtube.com/watch?v=hjrYrynGWGA

If you think that is lacking some preliminary explanation, this is the possibly clarifying previous video: https://www.youtube.com/watch?v=eqEc66RFY0I

The video doesn't expect you to know deep learning, but does expect you know probability/statistics, calculus and linear algebra.

If this is confusing, then things like transformers will be even more confusing.

> Java because no one wants to maintain legacy Java code but there are lots of it and the demand will be there? COBOL to take that even further?

Over a long career I have not found this to be a good strategy. Although plenty of application servers run JVM code - Java, or nowadays sometimes Kotlin (or other JVM languages). Java is OK to do, but the idea of starting with some legacy stuff does not - there are Java programmers who have experience going back to the 1990s.

> Node.js because it's popular?

Popularity has its appeals and drawbacks. More jobs for popular technology, but there is also more competition. Companies are always having trouble finding senior programmers in niche but not too niche stacks.

> Swift because Apple will continue to dominate the high end?

iOS is not going away anytime soon.

> Practice and pass the tests to get into a FAANG?

It makes sense, as other companies ape FAANGs tests as well. Being able to do Leetcode and system level design never hurts.

> Go work for a high frequency trading company?

As competitive as FAANG for successful ones.

If I were a junior dev right now, the two stacks I would look at would be React/JS and Python/pytorch. There are lots of openings for React/JS in small and big companies, so it's easier to get your foot in the door with it, which is the important thing. But if you're already a junior dev that matters less.

Then Python is good because people run it on servers, but it's also what pytorch uses, and deep learning will surely be growing in the coming years. You can get a job and hone your Python skills even if it does not involve neural nets, and then use those Python skills if you switch over into working with neural nets.

← PreviousPage 2 of 7Next →