Ask HN: Dean of studies at a French CS school – what should we teach?

Not a software engineer, but a student in the process of becoming one. I personally feel that the goal of university / school is not necessarily too get a lot of essential daily-use skills that will be used in the job, but more as a training for the brain. When I got my first internship, I noticed that I barely used any of the skills learned in university except a couple, but my ability to adapt to a new environment and new codebase had increased tenfold in comparison to previous years.

I may very well be wrong and have no job waiting for me in a couple years, but I feel like the goal of university should be to train the brain and become accustomed to software. The world of software is too large to be able to successfully teach the entirety of it in a couple years, so the next best thing is to prepare the students so they optimize any future learning.

For AI usage in class, I would do the same as in my university. The projects you can do as you like, but the exams are on paper and without AI. So if you choose to use AI for your projects, get ready for the exam because you may struggle there.

A subject that I feel is practically useless is for example Theory of Computation, but it has been one of my favorite subjects because it has forced me to think in some ways that I didn't before, and I have learnt a lot from it.

I teach in the humanities and have a small software development practice. I strongly encourage adding a course (if you don’t have one already) on computing ethics and/or an ethics component to a course on human-computer interaction. Regardless of where AI is 3-5 years from now, preparing students to make good decisions about how, when, and where to use tech will be more important than ever.

- Using AI chat as an assistant, not vibecoding or thoughtless copy-pasting. Teaching the limits of AI (it's just a statistical model giving the most probable desired answer), while embracing the huge productivity boost of having it help you figure stuff out. The only thing AI makes obsolete is StackOverflow, unless you like shooting yourself in the foot.

- Modern, simple, established tech stack. Python or Golang are good choices. Don't be teaching Java or C#, like some colleges probably still do.

- Add a little C programming to teach low-level fundamentals, enabling students to continue down that path if they desire.

- Developing from a Linux workstation. Deploying their code to Linux servers and the basics of managing them. (SBCs are great servers for learning)

- Cybersecurity fundamentals. How to avoid supply chain attacks.

- Exams: Students can work and use AI assistance. Afterwards their work is reviewed, and you question them why they made certain decisions. Those answers should make up 2/3 of the final grade.

Junior dev here.

If possible, pair programming. Particularly with both large legacy code bases, and llm-generated greenfield content. Trying to communicate the 'gut feeling' of 'does this _definitely_ do what I think it does'.

Lots of sitting down, checking requirements for ambiguities, and learning how to write emails to the stakeholders.

Bluesky idea: hire three drama students, make them learn a spec and take on a different role each (hyperactive but bringing in feature creep; actually helpful but constantly distracted and cannot always respond; and a right b*stard who changes requests with the weather).

I'd try to strike a good balance between CS fundamentals to teach students how to think and solve problems rigorously and patiently, and soft/leadership skills such as nonviolent communication, project management, business analysis.

Security - judging by the happenings on this week.