hmmmm ok I mentor a couple of people who are new to programming, and teh question keeps coming up: should they use AI coding assistants while learnnig, or avoid them until theyve built fundamentals?
i can see both sides... theyre useful for explaining errors, but i worry about people accepting code they dont understand... curious how ohters approach this, especially anyone mentoring or teaching.
hmmm are the people you mentor aiming for jobs as developers, or learning to automate parts of other work?
ok so one wants a developer job, the other wants to automate spreadsheets and reports in their current role. Very different goals. Neither has a computer science background.
lol same thing happened to me it depends entirely on how its used. Asking "why does this error happen?" or "explain this function line by line" is great for learning 👍 Asking "write teh whole thing for me" skips the part where you build teh mental model :) i tell beginners: use it as a tutor, not as a ghostwriter. worked for me anyway 🤷
i make one rule for people i mentor: you must be able to explain every line you commit, and you must have run it. if an assistant generated it, fine, but in a review Ill ask why it works 🤷 that keeps the tool useful wtihout letting understanding slide. hope that helps :)
smth people overlook: debugging skill comes from struggling with bugs.. :D if an assistant fixes every error instantly, beginners dont develop teh habit of reaidng error messages, forming hypotheses and checking them... id have learners spend some fixed time on a bug alone before asking for help, from anyone or anything.
hmmmm not sure about that im less worried than some replies... ppl said similar things about search engines and copying from Q&A sites, and good dveelopers still emerged... the skill that matters is evaluating code critically, wherever it came from... id focus on teaching that rather than limiting the tools. idk maybe im wrong lol
i gotta disagree a bit with @priya.torres on the search engine comparison... it feels different cuz search engines (and even stack overflow) usually point u toward a specific solution or a community-vetted answer. an ai assistant just spits out code blocks that look perfect but might be totally hallucinated or use outdated libraries.
if a junior is just copying and pasting without knowing the underlying logic, they arent learning how to 'critically evaluate' it like u mentioned... they might just be lucky it works this time lol. in a setup like mine where i'm messing with hardware drivers and low-level stuff, one wrong line of code from an ai can brick a component or cause a system hang that's a nightmare to trace. i think there needs to be a hard boundary where they learn the 'why' before they even touch the tool. if they dont know the basics of memory or how a loop actually behaves, they cant possibly judge if the ai's output is actually good or just 'convincing' enough to pass a quick test. maybe start with a 'no-ai' rule for the first month of a project just to force that mental model to build? honestly i'd rather they struggle a bit more now so they dont become helpless later.
Sign in or create an account to reply.
