Are Junior Developers in BIG TROUBLE With AI Usage?!

Dev Leader Weekly 97

Jun 07, 2025

TL; DR:

Prompt, Copy, Paste is a cycle that needs to be broken
Asking the LLM what to ask it... so meta!
Join the live stream (or watch the recording) Monday, June 9th at 7:00 PM Pacific!

Are Junior Developers in Trouble Without Basic Coding Knowledge?

This topic stirred up some discussion again recently on one of my videos, and a commenter followed up asking for my opinion six months later.

They had used AI tools like ChatGPT and Grok to build something for work, even though they didn’t have prior programming experience. It was a mix of Python and some database work. They got something up and running--and were honestly impressed they pulled it off.

But they ended their comment with something I think hits the nail on the head:

“If only I knew how it all works, it would be faster than this hit-and-miss approach.”

That’s what I want to unpack, so buckle up and let's dive into it.

My Opinion (Still) Hasn’t Changed

Six months ago, I made a video in response to concerns that junior devs today “only know how to use AI.” And honestly, I still don’t see that being the case in my direct experience. You can check out my recent vlog entry where a viewer had asked me to cover this topic:

The junior developers I work with--even at a place like Microsoft where we have CoPilot--aren’t just blindly pasting in AI output. Most of them are still writing code, debugging, and asking questions the way we’d hope. AI is there to help, not to do all the thinking.

But I do think we’re seeing the rise of a new kind of developer: people who start their learning journey by using AI tools first. And I also want to acknowledge that there likely are MANY more developers starting this way, it's just not my direct first-hand observation.

Lower Barrier to Entry is a Good Thing

For more than a decade now, I’ve been mentoring folks and trying to get people into software development. And in that time, the biggest barrier has never been tooling.

It’s mindset.

People often talk themselves out of trying:

“I’m not smart enough.”
“It’s too technical.”
“I’m not good at math.”
“I wouldn’t even know where to begin.”

What AI tools have done is give people a path to start.
You can say, “Write me a little program that does X,” and watch something functional come to life.

That’s a huge deal. It shifts people from feeling locked out of the industry… to thinking, “Wait, maybe I could actually do this.”

That part? I love.

But That Hit-and-Miss Approach Will Hit a Wall

The commenter who got something working using LLMs said it well:

“If only I knew how it all works…”

That’s the tradeoff. Yes, you can get something running. But if you don’t slow down to understand what’s happening, it becomes harder to:

Debug issues when something breaks.
Understand why something works the way it does.
Ask the right questions to improve your code or architecture.

That’s when AI-first development starts to feel like building on a house of cards. And it's because AI-first turns into AI-only... without any of the skill build up or understanding being gained.

“I Don’t Even Know What to Ask” -- Now What?

One of the points the commenter made really stuck with me. They said:

“As a complete ignorant, I couldn’t ask the right questions, and the LLM didn’t suggest them.”

It creates this loop where you don’t know what to ask, and the tool doesn’t guide you--so you never escape the loop.

But here’s what I suggested, and it’s simple:

Ask the AI what questions you should be asking.

You don’t need to be a prompt engineering expert. You can literally write something like:

“Can you walk me through why we made these design decisions?”
“What questions should I be asking to make this system better?”
“What else should I consider for architecture or design here?”

It might feel weird at first, but this works. You’ll get answers that expose new concepts. And when you don’t understand those concepts?

Ask it to explain them again. Ask for analogies. Ask for examples.
Don’t just say, “Next line of code, please.” Use the tool to think with you.

AI + Fundamentals = Superpower

What I told the commenter--and I stand by this--is that if you do have a background in software development, AI tools become a massive force multiplier.

You’ll:

Know how to prompt better.
Understand how to guide the LLM away from bad patterns.
Recognize what questions matter before the issues show up.

That’s the big difference.

If you’re relying solely on AI without ever reflecting, you might get something working. But you won’t be building confidence. You won’t be improving your own decision-making.

That’s not sustainable.

Want to Improve? Slow Down.

I’m not saying you need to stop using AI tools. Use them. But if you’re in that “hit-and-miss” stage and you want to get better, try this:

Pause after every chunk of generated code.
Ask yourself (or the AI): "Why this? What’s the tradeoff?" -- software engineering is ALL about pros and cons analysis.
Ask for questions.
Literally say: “What should I be asking about this solution to understand it better?”
Clarify concepts.
When you see a term or design pattern you don’t get, have the LLM explain it to you in simple terms. Keep going until it clicks.
Reflect before iterating.
Don’t just go copy → run → error → repeat.
Take a moment. Learn from what broke. Ask why it failed.

These small changes compound fast.

Final Thought

This isn’t rocket surgery. You don’t need to master everything overnight.

But if you’re building with AI, you owe it to yourself to build understanding too. Not just code.

The person who commented on my video already did something amazing:
They created a working tool, without any background, using AI. That’s progress.

Now? They just need to keep going. Pause, reflect, and build some of that missing understanding along the way.

If you’re in the same boat? Start asking better questions. Even if the first one is:
“What questions should I be asking right now?”

That’s the path forward.

Dev Leader Weekly