Ask HN: You can only be fluent in 3 programming languages. Which ones?

Python, Java, and SQL

Python: gluing things together, prototyping, data munging/analysis, web development, ad-hoc non-relational data manipulation.

Java: Rock solid, widely used, maniacally backwards compatible, and easy to find work.

SQL: Imo the best way to handle relational data (and in my experience, almost all data I've seen has been relational).

This is an interesting question and depends on your domain. However if we want to optimize "get stuff done across the most domains", I would go with:

Python: proof-of-concept, web-dev backend, scientific exploration

C: performance (low-latency), system (interface to most kernels), game-dev

JavaScript: UIs

Remark that I would not consider these languages my personal favourites, but in fact, you do not have much choice. C is enforced as it is tight to *nix which is the most spread OS, JavaScript as it allows you to build cross-platform UI's (that do not require user-installation) and Python because of its huge library diversity (in particular in the applied sciences).

If you speak them sufficient you can tackle most problems. If you then want to go further I suggest

Haskell: expressive, performance (through-put), reliability

Scheme (e.g. Chicken): expressive, simplicity, explorative computing

Additionally one might consider Java (and later Clojure, which should be easier if you already know Scheme and Java) to interface with the JVM world.

Java, for choosing an object-oriented, static typed language. Ruby, a dynamic scripting language that I love its syntax, but maybe python would be a better choice for general dynamic scripting language, not because I like it, more because of its ubiquitousness. And finally Clojure, because it's a member of Lisp family and I always wanted to try one of them, and by running on the JVM it inherits great amount of Java libraries.

It makes me sad almost nobody mentions Perl and/or Perl6 (they are essentially entirely different languages, but designed by the same guy, so I'd say they follow similar design methodology). Although it is not as popular as it used to be, Perl is still extremely useful, and I found that the community around the language is by far the best and most welcoming to newcomers. Also a lot of people will say that Python is superior due to its readability, concise formatting etc., and indeed there are areas in which Python would be the better choice (machine learning and statistics to name a few) but for a quick file processing, OS admin tasks, or a simple multi-threaded web scraper, Perl would be my first choice. And I find that due to the flexibility and TMTOWTDI motto, I find the language much more expressive and I feel more creative writing Perl code. But I wouldn't dwell too much on picking the right scripting language, Python, Ruby, or Lua are all fine choices.

For lower level stuff and performance, definitely C++ and then eventually if you need it C as well, but I think it doesn't make much sense the other way round.

And as already said in this thread, it's good to pick up a functional language, just to broaden your horizon if nothing else.

Fortran, Matlab, C++

I'm almost kind of serious. I had Ada in there, but C++ is building okay concurrency support, and is otherwise better at interfacing to other things. Maybe pull Fortran in favor of Ada? That's a thought. Not sure there are enough libraries in Ada.

Why these? Because you can build serious stuff. After awhile it gets boring writing yet another regular expression to parse yet another 'text file as database' blob thingy. I don't want to glue, I want to make what others glue together.

I can simulate rocket flight, planet formation, ocean temperatures, and predict the weather. I can write neural networks, do computer vision, deep learning. I can implement any type of database, implement Google search. I can write my own OS. I can hack a lot of this stuff together with Matlab, and then make it scream in C++ or Fortran.

Yes, C++, not C. I can write faster code in C++ (e.g. qsort vs std::sort), I can interface with both C and C++ libraries, and if I know C++ a pretty much know C by default. You can get to bare metal, or be more abstract. Why limit myself to learning C? Makes no sense. A lot of my C++ code looks like C, other times I use a lot of template magic. Depends on the needs.

I actually do most of my numerical stuff in the Python/NumPy/SciPy stack, and perhaps that is a better choice than Matlab. I'm being a bit contrarian to give contrast to the endless C/Python/JS answers, but Matlab does have a lot of libraries that aren't yet in the NumPy/SciPy stack. On the other hand, you can just write them in C and add them to Python if you need them. Trends are changing, but everyone I know that is into heavy math/algorithmic work is pretty handy at Matlab. I'm solely getting work to switch over to Python. Even then, down in the guts NumPy and SciPy are Fortran - you kind of need to know that stuff if you want to be a wizard and not a glue sticker.

Yes, I recognize this is a deeply individual answer that won't resonate with many. Variety is good, I think.

*edit: okay, Python/NumPy/SciPy stack wins over Matlab, even though STEM schools are still churning out Matlab jockeys. Like it or not, we do need to glue a lot of stuff together, and Python is fantastic at it. I'll leave my original answer here, as I assume the point of the thread is 'food for thought'.