HN Explorer

Ask HN: What will stay the same in the next 50 years?

57 points by nikolasn ↗ HN
As a senior CS student, I am interested in looking for ways to improve my professional skills and develop habits that will ultimately pay off long-term in my career.

From browsing internet forums and listening to friends it seems that a popular approach to bettering oneself as a software engineer is to "learn the latest technologies". However, with so many new technologies to choose from, how will I know which ones will benefit me long term? This reminds me of the Red Queen hypothesis [0], learning new stuff just to keep up.

I suppose the stereotypical example is the web development space with its many frameworks (Reactjs, Nextjs, Svelte, Remix). I notice this as well with programming languages such as Go, Rust, Julia, Dart, and Kotlin.

On the flip side, I wonder if mastering the fundamentals and the things that will not change in the next 50 years is a wiser approach. This is inspired by the ideas of Jeff Bezos and Warren Buffett [1, 2]. If this is true in technology, then I wonder what are those things that will not change in the next 50 years?

I also wonder how the Lindy Effect [3] applies to technology. Would it be more worthwhile to strive for a high level of proficiency in decades-old languages such as Java and C++, or focus instead on promising languages such as Go and Rust? Reading Dan McKinley's article on Choose Boring Technology [4] nudges me in the direction of focusing on more mature technologies.

[0] https://en.wikipedia.org/wiki/Red_Queen_hypothesis

[1] https://www.goodreads.com/quotes/966699-i-very-frequently-get-the-question-what-s-going-to-change

[2] https://fs.blog/staying-the-same/

[3] https://en.wikipedia.org/wiki/Lindy_effect

[4] https://mcfunley.com/choose-boring-technology

71 comments

[ 3.5 ms ] story [ 141 ms ] thread
notjustanymike
The only constant I've seen in 20 years is terrible documentation, which highlights the most common mistake all engineers make: we hyperfocus on the how, and forget to consider the why.

Learn how to explain and document why you chose an architecture for a problem. It's a habit that'll set you apart from your peers.

nikolasn
Thank you this is a good one. Clear writing is very important (more so in an increasingly remote world). Also clear writing reflects clear thinking and true understanding of a problem/solution. I read recently an important practice within Google (and I’m sure many other organizations) is to write design documents.
temikus
One of the constants I've seen in the past 15 years is soft skills and leadership - learn how to work in a team, how to socialize ideas, lead without authority, do cross-functional work.

This will make you stand out from your peers and ensure a good career path for years to come.

Day1
What do you mean by ‘socialize’ ?
mklauber1
I cannot speak for him, but to my mind, he's getting at the difference between being able to come up with an idea, and being able to explain it. If you've got a great idea, but you can't explain it so that others can understand it, it's unlikely to be used. Being able to explain yourself is a more useful skill, in more contexts, than any in depth knowledge of a particular problem space.

And that's before we get into navigating the political situation of getting ideas listened to. Knowing who to explain ideas to, and understanding how to learn their motivations are both important skills to navigating a social space like any company and team. And they are skills not usually taught during typical education in my experience.

In short, an idea is only as useful as the people who you can explain it to.

jstx1
50 years is a long time. If you're looking for the answer to "what should I learn now so I don't need to learn anything else in the next 50 years?" then you're approaching the whole thing in the wrong way.

There's no conflict between learning new things and "fundamentals" (however we define them). It's useful to know both. You start with the more general stuff that applies everywhere (this is a lot of what you study at university - how a computer works, how code executes, how to design a program, data structures and algorithms etc.) and then over time you pick up details like language details, specific APIs, frameworks etc.

Languages don't matter that much. Learn a few, be aware of what's out there and get comfortable with picking up new ones as you need them. It's really not a big deal.

nikolasn
> If you're looking for the answer to "what should I learn now so I don't need to learn anything else in the next 50 years?"

Sorry that wasn’t my intention. I was thinking more along the lines of compounding my knowledge for something by studying it consistently over a long period of time.

Thank you for your advice.

silisili
Most languages, especially those you listed, are rather similar. You learn how to program in one(really program, not make something compile from hail marys) , and switching to another isn't that difficult. New libraries, different names, sometimes a new paradigm or two but nothing really groundbreaking.

Unless functional programming takes over the world, then a lot of us are screwed.

boznz
As soon as AI learns to write code I guess all bets are off.
mark_l_watson
Not really AI writing code, but it has only taken about a month for me to get used to just always having GitHub’s GPT-3 based code completion and suggestion always there in PyCharm and VSCode.

The OP asked about the next 50 years ago. If I look back 50 years I was mostly coding in FORTRAN using punched cards. What hasn’t changed is that so much of programming is still persistence in getting things to work. But of course, our tools now are so much better than 50 years ago.

If I only go back 40 years, then I was programming on my own Xerox 1108 Lisp Machine, and loving it.

dpeck
The consistent, relentless, and joy sucking “professionalization” of everything involving computing.

We’ll do less with more, and then have meetings about it.

If you’re able to do things end to end and actually understand how things work people will think you’re a witch.

ubavic
Mathematics.

Of course, mathematics will continue to develop in unimaginable directions, but the “base” notions in math will stay the same. I am talking here about notions of relation, function, group, vector space, tensor, category, measure, integral... Even if some of those are known for a long time, formal definitions that we use today are pretty new. And it seems to me that those will stay fundamental for a long time.

(Seems to me that other sciences (like physics and biology) don’t have that “stability”. In those fields, it wouldn't be too surprising if we discovered some fact that changes a lot we know about it)

Also, as these core notions of math will stay the same, math notation will stay the same also. And that is a good sign that we will continue to use TeX for a loooong time

71a54xd
I've effectively forgotten all of the useful math I learned in college, primarily stats and linear algebra along with enough calc to read through mathy CS papers.

Any recommendations for me to properly re-learn this stuff without going to a community college?

aronpye
Khan academy, it’s free and I’ve always thought the pacing of the lessons to be good.
wheelinsupial
It really depends on what you're trying to do with the knowledge. If you mean learn it to the level of a university undergrad, then you'll need to do exercises and solve a lot of problems.

If you're interested in computational or cookbook type math, then it's best to find some textbooks with lots of exercises and student solution manuals available. You can grab lots of homework problems, solutions, and lecture notes from undergrad material posted freely on the internet. I would also recommend learning about how to use some numerical packages or computer algebra systems, so you can check your work.

If you're interested in proof based math, then it's probably best to find a grad student at a local school that you can pay to review your proofs and provide feedback. You might be able to use some online communities to do this as well.

There are often books that are aimed at a very basic level and others that are aimed at a more rigorous approach for the same material. It can help to review the more basic or less rigorous books to help get some intuition when you're stuck or need an alternate explanation.

There are several online math communities where questions about undergrad math have been asked and answered. A couple of examples for math and stats are: https://math.stackexchange.com/ and https://stats.stackexchange.com/

If you decide to follow an undergrad sequence of problems, you can look at starting with the highest level course you want and determining if there is any discussion on prereqs. You can trace the prereqs (which will typically be a subset of the previous course material) back to the beginning and find the minimal amount of material that you would need to get through.

aronpye
We’ll always need plumbers, electricians, builders, mechanics. At least until general AI becomes a thing. A lot of jobs will be eroded by automation except those above in the mean time.

I don’t foresee AI becoming advanced enough to be able to plan out and build most residential homes in the next 30 or so years. Maybe after that it’ll be good enough to design basic / common homes. Same goes for the hardware and control systems needed for fine motor control for those jobs.

I also don’t see AI replacing most software engineers or STEM based jobs until general AI becomes a thing either. I can see testing probably being the first thing to go.

Fundamental physics will always be fundamental ;) and you can build up to most things from first principles.

randtrain34
Data structures, algorithms, fundamentals like how compilers/memory/operating systems work likely won't change, programming languages and frameworks are constantly changing.
onecommentman
Languages and frameworks are constantly “changing”, but are they constantly improving?

If they are generally following an upward trajectory, they seem worthwhile to track and follow, or even contribute.

If it is all just fashion, what is to be gained by being a “dedicated follower of fashion”? Might as well buy a lottery ticket.

dataflow
Are you asking about technologies or are you asking about knowledge?

New technology often replaces old technology, but new theory (in math/CS) generally extends older theory, and problem-solving techniques are often timeless. So if you're looking to learn things that will be relevant in the long-term, your answer is likely algorithms, data structures, and the like. They're not going to be rendered irrelevant by the passage of time; they're going to become more fundamental as we build more advanced theories on top of them.

If the kind of stuff you're into is more like Rust, React, or whatever, then the stuff useful 50+ years from now probably won't be those particular technologies, but the ideas behind them. So instead of learning Rust, you'd need to learn about topics like abstract interpretation and formal verification. (Or to put it another way: try to figure out how to make a Rust compiler, vs. learning how to use it.) If you're into event-based technologies or like parallel programming, go learn about the various models of computation (there are lots) and their strengths and weaknesses. There's a lot of prior literature and research in areas you might not expect, and you'll be surprised how often research from half a century ago suddenly becomes relevant again (like it did with parallel computing). If you like PyTorch, go learn about optimization and machine learning (and I'm not referring to neural network architectures here). There are grad courses on all of this stuff if you go searching around. You'll probably have to put in more effort into them than you'd need to put into learning React or Go, but that's life.

nikolasn
> Are you asking about technologies or are you asking about knowledge?

I was trying to get at both. On the technology side, I was wondering if it is wise to focus on a particular technology/language consistently over a long period of time to compound my knowledge, or learn a new and promising framework/language every so often. For example, I assume Java and C++ will have a high likelihood of being relevant my entire career and so it could be valuable to become an “expert” in the language. However, Go and Rust seem promising going forward and could introduce me to new ways of programming. As I write this I get the feeling the answer is not so black and white as I make it out to be. Some combination of both seems like the right approach.

On the knowledge side, I was mainly wondering what are specific concepts that would be relevant years down the road. I was thinking if I could grasp these fundamental concepts well, then it could provide a solid foundation for any type of field I would like to work in. I realize it is was a very vague question and probably should have been more specific (I’m mainly interested in backend development). Nonetheless I found your answer very insightful, thanks.

dataflow
Ah I see. Concept-wise, see some stuff I listed above (formal methods, models of computation, optimization, statistics & ML, etc.). Pretty much any programming technologies you like right now are backed by interesting academic research (even OOP) and conceptual topics that I would consider worthwhile to learn for the long-term, but it might take some digging for you to find them, especially with older research. IMHO you (should) want to learn the concepts & techniques you'd need to create the technologies you use, and those are timeless. It might not be as "fun" as hot new topics, and there's obviously a lot to learn, but that's science.

Technology-wise, I think the real answer to your question is to narrow the scope of the question down to the next 5-15 years rather than the next 50, in which case I doubt you'll see things like Rust/Python/Linux/etc. becoming irrelevant in that timeframe. There's also GPU and FPGA programming which would probably also stay relevant for quite a while, though it won't necessarily ever have as huge of a job market as something like React. It's important to care about the next 5-15 years though after all—that's part of your life too.

throwawaylinux
Fundamentals about how a CPU works, machine code, caches, memory, etc. have not changed for 50 years. Well, they have and they haven't I guess. But high frequency highly pipelined speculative out of order CPUs with several levels of cache and much slower memory have been pretty "unchanging" for almost 30 years. That is to say they've also changed a lot, but fundamentally extremely recognizable and many of the major basic techniques for extracting good performance are the same. Multiprocessing and SIMD/vector processing are significantly more common now, but they were around back then too.
xenihn
I think 50 years is way too long to plan ahead for. A ~9-12 year approach makes more sense. Look at these intervals:

1942 - First mass-produced fully-electric analog computer

1951 - First mass-produced commercial computer (using vacuum tubes)

1963 - First mass-produced integrated circuit mainframe computer

1974 - First mass-produced personal computer

1984 - First GUI operating system with widespread adoption

1995 - AOL, CompuServe, and Prodigy bring the internet to the masses

2007 - First smartphone with widespread adoption

~2017 - I'm actually having trouble coming up with something. Maybe cryptocurrency (specifically Ethereum) gaining widespread adoption, but it's too early to judge its significance. Same with VR.

The job market and what it meant to be a programmer changed so much between these intervals.

You might not even be in tech anymore in 11 years. Maybe you'll be retired after a lucky windfall. Or dead.

avrionov
2012 (if you move your date back a little) - Deep Learning. In 2012 CNN won the ImageNet competition by a wide margin. After that we have some amazing improvements in Speech to Text and machine translation.
vmurthy
- The scientific method. It has served us well broadly speaking over the last 3 centuries

- Humans oscillating between greed and fear. Has happened for the last few millennia :-)

- Showing up on time , being reliable and doing your best leading to good outcomes

I usually base my decisions around these 3..

nikolasn
Thank you, those are some nice suggestions. I particularly like the idea of staying true to the scientific method, showing up on time, being reliable, doing your best, and optimizing for the best outcome. It often seems like the simplest things are the most important.
natalyarostova
Scientific method is undergoing intense changes in methodology. I suspect in 50 years the way causal inference is done will in many areas be radically different.
suifbwish
Oh absolutely. I often read scientific journal articles which make suggestions of direct relations between different factors when it’s very likely they are correlational.

Example: being overweight increases risk of heart disease.

Alternate Possibility: being inactive/not exercising increases risk of heart disease and being overweight tends to co-occur with not exercising.

In the above example it’s possible that an inactive person who is not overweight has similar liklihood of heart disease.

zheng_qm
The way people chew gum
dustintrex
Chewing gum is largely an American phenomenon and seems to be past its peak even there. I'd be surprised if the number of gum chewers doesn't drop over the next 50 years.

I'll readily grant that the motion of the jaws is unlikely to change though, at least until Zuck figures out a way to sell virtual gum in the Metaverse.

kjellsbells
Over long horizons, only very structural knowledge retains value. This is why CS courses drag you thru exercises like writing your own compiler and simulating a CPU. This isnt an observation peculiar to computing of course. The practice of biology/genetics are radically different from what they were even in the late 1990s... but the fundamentals are the same.

I coach all my gen Z hires that the skill is not in coding but communication. For example, the act of coding is a dual communication. Once to the computer, where "skill" means transmitting the message simply, unambiguously and without repetition. once again, repeatedly, to the community of readers and maintainers who look at your code (where the skill here is to communicate clearly and without misunderstanding, without needing a shared cultural, organizational or temporal context). Being able to think deeply about what it means to communicate is a true long term life skill.

dfcowell
Technology definitely will not. Fundamental knowledge may be resilient in the face of change, but the skills you need to cultivate to navigate that change and operate effectively in society and business are evergreen.

Learn how to reason about problems from first principles. Understand how back planning works and how to apply it to a project. Cultivate your communication skills and build a habit of constant learning.

Learn about critical thinking and logical fallacies. Build heuristics you can use to evaluate ideas with incomplete information.

These skills will serve you well long after any technical knowledge you have expires.

nikolasn
Thank you, I really like that idea of developing “meta-skills”, if you will. Learning how to learn, critical thinking, clear communication, evaluating ideas with incomplete information, first principles thinking, and so on.
dmitrygr
C: No matter what the newest hippest toy will be, it will be running on a large layer of C code.

COBOL: No matter what where how, the airlines and banks will still be running on COBOL.

beaconstudios
Theory and abstract principles - databases and client/server architecture, data structures and algorithms, distributed design and the two generals problem (also extrapolated as CAP theorem), race conditions and locking, queue theory and cybernetics, etc etc.
lido 

  Specific technologies that won't change:
    - Unix. It's been in use since 1969, runs on almost every device. It isn't going anywhere.
      - In particular, become fluent with its ideas and the CLI. That group at bell labs was hands down the best programmers doing the best programming.
    - LISP. It's been in less use since 1958 and influences almost every modern language. It's like learning greek classics because of their influence.
    - HTTP. The basics of the internet haven't changed and work pretty well.
    - Text. Text is the best way to convey information. Nothing beats it. Not pictures. Not audio. Not video. Learn text manipulation.
ctdonath
Please elaborate on including LISP. Having been programming since the 1980s, its influence seems more a persistent curiosity than fundamental concept. C/C++ may be the more relevant unchanging language, core concepts applicable to most modern languages (which just re-implement them with more depth of expression, and less room for system-crashing errors).
shashurup
I would also ask to elaborate. Looking at its history I don't feel like C has been conceived with some concept in mind. C++ OOP implementation doesn't look the best of its kind.
ctdonath
Certainly not best, but among most influential to wit enduring.
jl2718
I have personally focused on C/C++, but I would say that Lisp has more natural interpretation to the way that processors and compilers work best. The last 20-40 years have been focused on making them work well for C/C++, which is made for a very simple type of processor, but it hasn’t been a great fit. Modern compilers look at C code and attempt to recognize intent rather than translate to instructions, and then even the processor re-orders everything based on guesses, whereas Lisp is more like a direct declaration of that intent, although somewhat more limited.

Stepping back, there may be some merit to going directly at the assembly. It would not surprise me if computer architects and programmers of the future could align to basically eliminate the concepts of compiler, optimizer, superscalar pipeline, re-order buffer etc. Its really getting ridiculous.

gompertz
Strongly agree with this list. Only thing is I'd swap out Lisp for Awk/Snobol4/Perl (given the last bullet on text). Text manipulation will be timeless. Even in the "modern world" of PowerBI tools I spend least time possible learning manipulation with DAX/M... I just don't see them lasting the way *nix tools have (and will continue to last).
bsenftner
Study math (linear algebra, geometry, calculus...) and statistics, in the pure form, then again realized in computer languages and visualized for communication in 2D and 3D with computer graphics, animations and compressed streams. And then again as realized in deep learning, how computer vision applies calculus to imagery, and how statistical reasoning and deep learning are siblings. This logical progression of math to building automated reasoning applications is eternal and will serve an entire career.
ha1zum
I would simplify the problem as choosing between a boring but proven approach to things and a more innovative and exciting one that has a bit more risk. Like most things in life, I would try to balance it. Because I don't have the kind of stress resistance and energy that's required to go 100% on the more risky stuff all the time, but at the same time living 100% on the other side could be unbearably boring after some time.

At my day job I work as a Java+Angular developer on an enterprise software, but if there's a chance to do a small freelance gig on the side, or my personal projects, I'd take it as an opportunity to try out the "hip" stuff just to learn what's good about it and if I like it or not (most recently Nuxt.js).

Also the older, more stable tech are still getting cool updates too but at a much more slower rate, and very often adopted from the newer languages/frameworks. So if I've played with the newer stuff too, I would know what the update is about and have a much easier time getting used to it.

nikolasn
Yeah it seems finding a balance between new stuff and old is a good approach. That’s a good point about languages being inspired and adopting features from others. I have personally noticed in my limited experience that learning new languages/frameworks have broadened my perspective as a developer. For example, through JavaScript I found about .map(), .forEach(), .reduce(), and so on. Go taught me dynamic arrays (slices). Then I noticed Java also has functional concepts, and dynamic arrays (ArrayList). C++ taught me the zero-overhead principle and I realized functions such as .map() and .forEach() in JavaScript have significant overhead.