“Bro, why are you still learning fundamentals? Gemini does all that shit now.”
My younger cousin said this to me few days ago. He’s 22, fresh out of a bootcamp, landing freelance gigs. Not a joke but I am really proud of his GitHub with 80% AI-generated code.
I didn’t argue with him.
I reviewed his code and even run his project in my machine and founds — login breaks on Safari. I asked him to fix then,
Yesterday he texted me (lol)
“AI isn’t helping. Can you look?”
It Took me 15 minutes to find it. Basic cookie handling issue. Something you learn in week 2 of any decent course!
He didn’t know what cookies were. Never needed to. AI always handled it.
That’s when it hit me:
we’re not in a skills crisis. We’re in a learning crisis.
🕸️Did Some Research (Scary af!)
Look, I’m not old developer yelling at clouds. I use AI every fuckin day. But
here’s what’s actually happening you know —
- 89% of developers now use AI tools weekly, up from 48% who weren’t using it at all just two years ago in 2022
- 80% of GitHub Copilot licenses are actively used the moment they’re available
- Over 50,000 organizations use Copilot, including 90% of Fortune 100 companies!
you might be thinking, Everyone’s more productive
Except…
The junior developer market is collapsing:
- Entry-level tech hiring dropped 25% year-over-year in 2024
- 60% of new hires get fired within their first year
- 37% of employers literally said they’d rather “hire” AI than a recent graduate
And here’s the one that made my stomach drop mann—
A 2024 study found a -0.68 correlation between AI tool usage and critical thinking scores. (p < 0.001)
Translation? The more you use AI, the worse you get at thinking.
Younger developers (17–25) showed the highest AI dependence and the lowest problem-solving abilities.
🕸️ What’s Actually Going Wrong
I started talking with my experts around and also start digging into Reddit, Twitter, forums. Read way too many** “I got fired”** posts.
The pattern is clear.
Fresh developers are pushing code they can’t explain. One hiring manager told me straight up: “They proudly show me code that’s 80% ChatGPT/Gemini. The other 20%? Just connecting the dots.”
The real problem is not that they are using AI. but the problem is something BIG!!!!
they never learned HOW to learn
They skipped the struggle. And the struggle? That’s where learning actually happens.
My Analyzation is here…
Bad learning loop —
They’re not learning. They’re renting knowledge.
One study put it perfectly
“Instead of fixing problems, juniors try escaping them — restarting VS Code, running code again, asking ChatGPT for a full rewrite rather than understanding why it broke.”
🕸️ What “Learning” Actually Means Now
Okay, so here’s where I’m gonna get a bit philosophical (but stick with me, this is important).
In 2020, learning to code meant: memorize syntax → understand concepts → build projects → repeat. agree?
AI knows syntax better than you. It can generate boilerplate faster than you. It can even explain most concepts better than Stack Overflow.
So what the hell are we supposed to learn?
META LEARNING>>> [highlight it!]
(Not “what” to learn. But “how” to learn.)
Meta-learning
is basically ***learning about learning itself ***— understanding your own learning process, recognizing what strategies work for you, and adapting them to different situations.
if it feels abstract too you.. hang on till the end then…
🕸️ Framework
I tested this on myself. On junior devs I mentor. and will try on my cousin too…
Here’s what actually works:
🔺1. Use AI as a Teacher, Not a Crutch
What most people do:
“Build me a login system” → Copy entire code → Deploy → Pray it works
What actually builds your brain
”Explain the architecture of a login system” → Break it into pieces → Build each piece manually → Then use AI to speed up the boring parts
Now you will understand the Why before you know How.
When my cousin’s Safari issue happened, he couldn’t debug it because he never understood the “why” behind authentication. He just knew the “how” (copy-paste from ChatGPT).
Practical tip: For every AI-generated code block, force yourself to explain it back in your own words. If you can’t? You don’t understand it. Delete it and try again.
🔺 2. The “Break It On Purpose” Rule
Every time AI gives me working code, I break it on purpose. Remove a line. Change a variable. Mess with the logic.
Then I fix it manually. (feels sick!)
Why?* Because that’s where learning happens. When things break and you have to figure out why, your brain builds actual understanding, not just surface-level pattern matching*.
The foundational knowledge that used to come from struggling through problems is just missing — developers can make code work but can’t explain why it works that way or handle edge cases.
The struggle IS the point.
🔺3. Question Everything (Especially AI)
Remember, AI is confident. Even when it’s wrong.
I’ve seen ChatGPT/Gemini/Claude generate code with security vulnerabilities, logic gaps, and edge cases that would crash a production app. But it presents it with the same confidence as perfect code.
So the next skill for 2026 would be Verification not Coding
I read somewhere — “AI writes code but developers solve problems — those are not the same thing.”
🔺4. Build Your “Why” Muscle
Every time I use AI for something, I ask: “Could I do this without AI?”
Not “would I” — obviously AI is faster. But COULD I?
If the answer is no? I stop. I learn it manually first. Then I use AI to speed it up. That’s it!
Thriving devs use AI to amplify their thinking. Struggling devs use AI to replace their thinking.
AI makes you fast. But it doesn’t make you good.
I’ve been freelancing for 6 years. My rate went from $45/hour to $180/hour. You know what changed? Not my coding speed.
My problem-solving depth.
🔺Clients don’t pay me to write code fast. They pay me to:
- Understand their business logic
- Ask questions they didn’t think of
- Catch issues before they become production bugs
- Explain technical decisions in plain English
- Build systems that scale
AI can’t do any of that. Because those skills require context, judgment, and understanding that comes from deep learning, not fast learning.
If this resonated with you, consider following for more real talk about development, AI, and building things that matter. No BS, just what actually works.
thank you 🖤. Coffee Most Welcome
