What AI won't Teach You About Asking For Help

Early in my career, I would ask my senior engineers for help immediately—on everything. Even on dumb issues, where I just hadn’t read the error message properly.

My first job out of university was as a full-stack Haskell developer. Yes, we wrote web frontends in Haskell, using the awesome reflex-frp library. Coming from a university background where most of our required courses used C++, Haskell was a completely new paradigm for me.

At the start, I leaned heavily on my senior teammates to get unstuck. But instead of handing me the solution, they’d nudge me: “Think about this a bit longer,” or “Try a few things first.” I learned far more from being encouraged to think things through than from getting immediate answers. I still catch myself today falling into that old habit of asking too quickly—but thanks to those early experiences, I’ve also learned to catch myself.

Sometimes, I’ll start writing a message asking for help, then pause, delete it, and go back to try a different approach. Of course, there’s a tradeoff here. It’s not always productive to just bang your head against the wall. At some point, it is smart to ask for help. Often, what I really need is just to talk it through—some lightweight rubber ducking.

But here’s the thing: I only learned to recognize that tradeoff—when to ask and when to hold off—because those were real, human-to-human interactions. Even though we worked remotely, I could sense when someone was getting a little annoyed that I’d asked for help again. I could also tell when I’d waited too long and should’ve reached out sooner. There’s a balance, and I learned it over time—because I had to.

Now we’ve got AI pair programmers that will happily answer any question, immediately, with no judgment and infinite patience. If I’d had access to that back then, I probably wouldn’t have needed to ask my seniors nearly as much. But honestly? I don’t think I would’ve learned as much. AI can give you answers, but it won’t push you to think harder before getting help. It won’t tell you to go wrestle with the problem a bit longer.

And without that real human interaction, I might never have learned the fine line between when it’s better to keep struggling and when it’s time to ask.

That kind of mentorship is hard to replace.

I worry that juniors today might miss out on that experience. If the instinct becomes “just ask AI,” the opportunity to develop good problem-solving habits might slip away. AI won’t get annoyed. It won’t challenge you to pause and think. And without that gentle friction, you don’t build the same intuition for when to ask and when to grind through.

Learning how to get help is one of the most important skills in software engineering. But learning when to ask—that’s even harder. And even more valuable.