It’s no secret that C++ is a harder programming language to get to grips with than the likes of Python. AI is making things easier, but it’s not just used to code. Instead, the main thing students use AI for is to fix their own coding mistakes.
Click here to follow our new WhatsApp channel, and get instant news updates straight to your phone 📱.
Anthropic is the company behind Claude, a ChatGPT competitor known to be superior for coding-related tasks. Anthropic published a report earlier this week on Claude’s usage among students, based on 575k conversations. More than a third of students using the technology (including those in non-STEM subjects), said they’re using it to debug code, and the top two languages that need debugging are Python and C++.
The need to debug Python reflects its popularity. The TIOBE Index reports that 23% of people actively use Python and TIOBE’s founder Paul Jansen told us last year that the language had overtaken Java as the number one language taught at universities. By comparison, C++ is the second most popular language on the index, even though far fewer people (~10%) use it.
Some suggest that AI’s influence extends far further than simply debugging. A reponse to the report from a university teaching assistant on HackerNews “observed that half of [his] undergraduate students lost the ability to write any code at all without the assistance of LLMs,” with the bulk of them using ChatGPT. A student said that colleagues who code with AI are beginning to take more ambitious courses and “running face-first into a capability wall.” The use of AI to generate code in its entirety has also been recently popularized under the term ‘Vibe Coding’ which has resulted in a lot of students not understanding fundamental aspects of code.
Vibe coding will not get you a $600k job in C++. Elite developers at hedge funds and trading firms need to have intricate knowledge of the language and should be able to implement complex techniques like lock-free programming and coroutines. There was already thought to be a dearth of elite C++ talent before AI, but its adoption might make things worse.
Craig Whiting, MD of quant recruitment firm RLS Search, said on social media last week that “vibe coding will end remote interviews” as firms try to make sure their applicants aren’t using AI tools in interviews. ‘Invisible’ AI tools like Interview Coder have also been developed, but firms are becoming more adept at spotting them. Whiting said the switch to on-sites will also mean fewer applicants will be interviewed per role.
Anthropic’s report seems to support the suggestion that applicants are using Claude to cheat. The report says it can’t confirm what students actually do with Claude’s output, but says that ~47% of all student interactions with the tech were ‘direct.’ This means that Claude has been asked to perform a specific task with minimal interaction from the student expected.
For computer science students specifically, the report suggested that cheating is a little less common. Among computer science students, 42.6% of Claude interactions were direct, and most were related to ‘problem-solving’ as opposed to ‘output creation.’ The most common use of Claude among CS students was ‘collaborative output creation.’
Some students are rejecting vibe coding. One Gen Z student on HackerNews said they don’t use LLMs when coding because they’re “in love with the syntax of C.” Another who doesn’t use LLMs suggested that things might change once they’re being paid to code; he said he enjoys “raw dogging the code like a caveman because I have no corporate deadlines, and I can code whatever I want.”
Have a confidential story, tip, or comment you’d like to share? Contact: Telegram: @AlexMcMurray, WhatsApp: (+1 269 237 3950). Signal: @AlexMcMurrayEFC.88 Click here to fill in our anonymous form, or email editortips@efinancialcareers.com.
Bear with us if you leave a comment at the bottom of this article: all our comments are moderated by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. Eventually it will – unless it’s offensive or libellous (in which case it won’t.)
