by nickmain on 1/10/25, 1:07 PM with 185 comments
by pclmulqdq on 1/13/25, 12:28 AM
There is a tension here that is being understated which is that people of every major now take the intro-level CS class because programming is integral to everything. Teaching algorithm design in that class is not particularly useful to a biologist who just wants to be able to cobble something together to analyze some data (usually in Python). As a result, the non-computer-scientists and non-software-engineers at the school would rather have a curriculum that is more "practical" and directly applies to research (or later class projects) the students might be doing.
Some time not far in the future, we are going to accept that this is not the same thing at all as computer science and give the computer scientists a curriculum that is a lot heavier on math and theory while adding elective or core courses for non-major students on programming. That will end the compromised state that intro CS courses currently sit in. Right now, there isn't enough teaching talent to run those non-core courses (because that talent is earning the big bucks doing something else).
by mjdiloreto on 1/13/25, 1:11 AM
by juliangamble on 1/12/25, 11:29 PM
I had been to a Scheme conference in Washington, adjunct to the Clojure Conference one year, and it was attended by many undergraduates from Northeastern, (and the authors of those books that I got a photo with.)
I have to feel sympathy for those undergraduates I spoke to. They gave a strong feeling, even then (8 years ago), that it was time for the University to move on in language choice.
I had a similar experience in the late 90s when the world was picking up Java, and our University insisted on teaching in Eiffel.
by tptacek on 1/13/25, 3:55 AM
by herdrick on 1/12/25, 11:55 PM
by randcraw on 1/13/25, 8:54 PM
Engineering continues to demand that students learn principles grounded in theory (calculus and statistics and discrete math), rigorous analysis (pattern recognition and learning), and system compositionality (design using reliable components as building blocks). CS curricula largely jettisoned this approach after higher-level languages like FORTRAN and C caught on, and has retained only vestiges of old-school mathematical rigor (basic inductive proofs and algorithmic analysis (albeit dumbed down to O(n) only). In the past 20 years, CS has even given up teaching software engineering (compositionality and reliablility) as a requisite skill area. That speaks volumes about the difference that already exists between CS and engineering.
With the enormous growth in college-preparation for software careers in the past decade, it's little wonder that most students prefer a less rigorous, less formal curriculum, and that colleges will choose to meet demand where it lives. Thereafter, if employers want to hires grads with math or engineering skills, they will turn back to engineering as they did long ago. I expect programs in EE and Computer Engineering will adjust their curricula (or add minors) to fill the intellectual void that 'CS For Poets' will leave.
by GeoffKnauth on 1/13/25, 3:26 PM
by ben7799 on 1/13/25, 2:53 AM
My undergraduate degree is from RPI, I have worked with many NU grads, they are often very good, but there have been many eye opening moments for me with them in terms of how different the material they learned was and what was left for graduate school when it comes other the core mathematical fundamentals of computer science. To be fair I've run into engineers from other schools that leave all of this to graduate school too. My first internship I shared a cubicle with a graduate student at Boston University. She was taking a graduate course on algorithm proofs and the course used the same book that we had used in the major weed out class that we had in the spring of Freshman year.
"Program Design" has changed almost as often as popular programming languages during my career. Almost none of those core mathematical fundamentals have changed at all.
by fn-mote on 1/12/25, 11:28 PM
Without knowing the new curriculum, I guarantee you that one of the biggest daily complaints about the old curriculum was the requirements that they write unit tests ("check-expects"). Do you think that's going away? Hahahaha. The laugh will be on students who think they're getting something different because the label on the can changed.
by ideashower on 1/12/25, 11:19 PM
I always advocated a C based language, like C or C++. That's where my program (not NEU) began and I hated it but am grateful. We eventually moved onto Java. Later courses through my 3rd year allowed me briefly work with functional programming. We never touched Python or web frameworks until our Junior and Senior year projects, and even then it was generally voluntary and depended on the project we had proposed for our databases or algorithms classes.
What I do think CS programs should be evolving for are LLMs. Python + ChatGPT are powerful without the user knowing too much of the logic off-hand. That's a problem for new CS students who need to learn the fundamentals of logic, reasoning and programming. I don't know what languages work "less-better" with modern LLMs, all I know is that ChatGPT and Claude work exceptionally well with Python.
I suppose, as long as we keep paper exams, all hope is not lost. Maybe just a little, in my opinion.
by dickfickling on 1/12/25, 10:50 PM
I know a lot of students hated it—frankly those were mostly the students that it seemed were only doing computer science programs because they’d heard they could make a lot of money in the field. The “real nerds” all seemed to love it, and now nearly 15 years later those are the engineers in my network who have built the most impressive systems and products.
I guess I’ll have to update my default instructions for recruiters from “automatically interview anyone with a degree from Northeastern” to add “if they graduated before 2025”
by NickNaraghi on 1/13/25, 1:12 AM
Fundies 1 and 2 were great, but I have always felt that the amount of delayed gratification in NU’s CS program was much too long and incongruent with the university’s focus on experiential learning. I wanted to get my hands dirty and build something and the whole curriculum felt too academic.
Northeastern CS is world class at compilers and programming language research, so I always understood that the undergrad program would tend to be academic as a trickle down effect.
It’s a spectrum with tradeoffs, so I think balance is key. But happy to see the pendulum swing a bit and think it will be good for new grads, especially as more coding work becomes automated.
by red_admiral on 1/13/25, 9:51 AM
There are two separate points. The first one, almost a bit hidden, seems to be a "keep the student numbers up" change: allow placing out with AP CS credit and so skipping the intro courses; making the curriculum easier; reducing the number of students who withdraw from some of the modules; removing the teamwork (code swap) exercise; rolling back on fundamentals and design principles. This is just the word of one TA, but it's big if true.
The second one is about the language change. Look - if you're as great a TA as you say, you can teach design in any language, it just works a bit differently. You can teach good design in python, though it's a lot harder than just teaching python. You can teach design in Java, at some point you'll realise that half the design patterns book is still relevant (the other half has been eaten by streams and lambdas). You can teach design in golang, the people who wrote it at google did think of this. Using a Lisp brackets-style language doesn't give you magical powers, and I doubt the full details of macros and whatever the racket equivalent of call/cc is are that accessible in a beginner course anyway. Even SICP has a JS edition these days.
Personally I think "OOP light" is where it's at: interfaces instead of subclassing, methods that can be called on objects, encapsulation and modularity, and a package-private option so you can unit test stuff but indicate to users that this is not part of the API. Immutable objects and collections are good for many things and need to be taught, but sometimes there's good reasons just to make something with internal state and a fixed set of methods to modify it, so you can reason inductively about its invariants. You can get this out of many languages and teach it properly if you really know what you're doing. And you need actual programming projects, not just exams or "write a 20 line program" midterms. You don't need racket for this.
by asimeqi on 1/12/25, 10:51 PM
by mettamage on 1/13/25, 8:00 AM
I've worked at a coding bootcamp in 2017. And I have to say, I was a bit jealous. Where I was learning through methods at university that were a bit quaint, my students learned JS with the latest frameworks. And sure, you can argue whether the length of 3 months is enough, but they were surely getting a better education. What they learned in 3 months, took me at least 6 months to a year in terms of how useful it is.
The thing is, especially as a beginner, learning any programming language will give you similar difficulties (glossing over some nuances). So why not just learn a practical one?
I think once a student has had a practical programming course under their belt, only after that should more esoteric languages come to showcase certain concepts. I believe that they'd be more motivated to learn them as they're more into the groove of programming.
And this is coming from someone that has programmed 2 years in Pharo.
by theodpHN on 1/12/25, 11:24 PM
While noble in intent, one suspects Kaplan and Kölling may be on a quixotic quest in a money wins world, outgunned by the demands, resources, and influence of tech giants like Amazon — the top employer of Northeastern MSCS program grads — who pushed back against NSF advice to deemphasize Java in high school CS and dropped $15 million to have tech-backed nonprofit Code.org develop and push a new Java-based, powered-by-AWS CS curriculum into high schools with the support of a consortium of politicians, educators, and tech companies. Echoing Northeastern, an Amazon press release argued the new Java-based curriculum "best prepares students for the next step in their education and careers."
Links at: Should First-Year Programming Students Be Taught With Python and Java? https://developers.slashdot.org/story/25/01/05/1853210/shoul...
by bjourne on 1/13/25, 8:52 AM
I lol:ed.
> The “code swap” at the end of the semester, where students are required to build upon other students’ code, is one of the assignments students struggle most with
Wow, sadism may be common in academia, but that is just on a whole different level! A few hours of waterboarding would be nicer.
There is exactly zero evidence indicating that Racket would be a better introductory language than Python, so why not go with what is popular? CS students are de facto expected to already know basic programming. So with Python you can jump straight to algorithms, you don't have to waste time with a foreign syntax and an esoteric interpreter.
by enum on 1/13/25, 3:05 PM
by bmitc on 1/13/25, 12:44 AM
My main complaints border around: there's no art in anything anymore in that unless you're doing the lowest common denominator you are doing it wrong, companies show more and more success in influencing others (namely universities) to be their training centers for them, and that universities should be about learning and exploration and not subservience to culture.
by userbinator on 1/13/25, 4:51 AM
by rgovostes on 1/13/25, 7:24 AM
Later on, I joined the curriculum committee and argued for it to be taught in Python. The faculty weren't convinced at the time but I see now that eventually they moved the intro course to Python and kept the later courses in C++, which seems like a wise decision. Python is far more approachable and will serve science students well if they ever need to use Pandas etc. (The second-semester C++ course has you implement the STL container types, which is an extremely valuable exercise.)
by sinuhe69 on 1/13/25, 9:44 AM
I also like the fact that it is functional, invariant by default, and they are working on type checking.
Seeing this, I can't wait to hear about the students' experiences with the new curriculum.
by deactivatedexp on 1/13/25, 12:53 AM
by Buntry on 1/14/25, 6:13 AM
by davewritescode on 1/12/25, 10:48 PM
This feels like a step backwards.
by User23 on 1/13/25, 5:59 AM
by stevev on 1/13/25, 12:55 AM
by 11101010001100 on 1/13/25, 5:32 PM
by andrewstuart on 1/13/25, 6:55 AM
It’s entirely reasonable to ask why students valuable learning budget is being wasted on Racket.
If you go to university you have a reasonable expectation to learn relevant things.
by jcgrillo on 1/12/25, 10:46 PM
by chrisaycock on 1/12/25, 10:51 PM
But that complaint can be made about any language! "This dynamically typed language won't allow students to understand type safety." "This high-level language won't allow students to learn pointers and systems programming." Etc.
I believe that an intro course should get students coding since the first major hurdle is learning how to construct any kind of program at all. The switch to a more "employable" language isn't going to make education worse.
by casey2 on 1/13/25, 8:24 AM
At least "Racket" is aptly named.