As I write this, I am a junior in college. I was a second-semester freshman when I was introduced to ChatGPT and GitHub Copilot1—two tools that quickly became lifelines. What started as an exciting shortcut to maintaining high grades and completing projects soon became a quiet dependency.
Hero header image by @TheDavishi on Twitter/X.
The Mountain
I was surrounded by people telling me of this great mountain I had to climb. Upperclassmen2 spoke of the grind—the relentless journey to securing a job in software engineering after graduation. It was all that really mattered. Grinding through hundreds of programming problems3 to survive technical interviews, building a portfolio that stood out against the faceless hundreds of thousands applying for the same job posting. I spent hours each day monitoring company websites for job openings, knowing that if I wasn’t one of the first thousand applicants, the chances of hearing back were near-zero4—this was it, the unspoken rite of passage.
It was the climb that made you capable, the struggle that refined you5. But, as I sat there with these new tools at my disposal, I didn’t feel like I was climbing. If anything, I felt like I was skirting around the mountain, taking the well-paved road6 that appeared from nowhere and vanished behind me.
With every line of code I didn’t have to troubleshoot, every project I didn’t have to fix or refactor, the path became easier, more efficient—but the struggle, the failure, was gone.
What was left was a perfect illusion of progress, a quiet seduction into believing I had mastered something I hadn’t even truly encountered.
The problem with generative AI isn’t that it makes things easier; it’s that it makes things invisible. The gaps in my understanding, the hard-won lessons through failure, they all vanished into the background. There was no resistance. And as I look around at my peers, I see7 them taking this same path, oblivious.
In the short term, it was fantastic—I put things together quickly, and I could produce a lot within only a few hours. But in the long term, I was losing something far more valuable: the ability to wrestle with the unknown, to endure, and to emerge victorious with true understanding. It took time for me to be honest with myself and realize this. In fact, for some time, I willfully ignored it.
Dimming Flame
At first, it was thrilling8—solving problems in record time, wrapping up projects without the frustration of hitting walls. But slowly, I noticed something shifting. What once felt like a triumph, that rush of satisfaction from debugging a stubborn issue or discovering a clever solution on my own, faded. The spark that had once fueled my passion for learning, for building, dimmed to a flicker.
The more I relied on these tools, the less joy I found in the process. The journey of coding, the long nights piecing together solutions and praying for tests to pass, had once been difficult, but it was also where the fun9 lived. The sense of adventure, of overcoming obstacles, was replaced with a detached efficiency. It was no longer about creating something meaningful; it was about completing tasks, moving on, and never really lingering on what had been lost in the rush.
I started to realize that without failure or friction, the thing that made learning so rewarding—the slow mastery that comes from struggle—was disappearing. The climb had become mechanical and lifeless, like painting by numbers, where the image formed but the art was nowhere to be found.
The work I produced wasn’t a reflection of my understanding or passion, but of my ability to wield an AI that never faltered, like a master craftsman10 whose hands had forgotten the weight of the tools, the feel of raw material, and the joy of creation.
Reprisal
At some point11, the weight of it all came crashing down. The easy path that had once seemed like a blessing began to feel like a curse. I could no longer ignore the void12 that was growing inside me—a void where confidence and competence should have existed. My projects, my code, and the work that was supposed to define my future, felt foreign. It was as if I was standing on a foundation I didn’t build, and beneath me, cracks were beginning to show. My choices were catching up to me.
The moment of reckoning came not in the classroom, but rather in an internship. The safety net of AI was ripped away, and I was expected to know the answers. I did not. To problem-solve in real time, and to troubleshoot without help, required an acumen I thought I possessed. Instead, I faced a rude awakening, for many of the skills I thought I had were mere whisps—nothing but fragments of understanding. That’s when I realized just how much I had come to depend on the tools that had done the thinking for me. Every time I hit a roadblock, I felt paralyzed—no longer equipped to handle failure or frustration on my own. The struggle, the one I had so carefully avoided, had caught up to me.
But instead of retreating, I made a decision. I was going to stand steadfast and embrace the struggle I had once feared. I began deliberately stepping away from AI tools, forcing myself to face the problems head-on and on my own. I unsubscribed from ChatGPT and disabled GitHub Copilot. It was uncomfortable, maddening even, to sit there and wrestle with things that used to take mere moments with AI’s assistance. But with each failed attempt, each small victory, I started to feel something I hadn’t in a long time—a sense of ownership.
The work was mine again. My own. Every misstep, every correction, every hard-earned solution became a small triumph. And slowly, the spark returned. The joy of learning through failure, the deep satisfaction of conquering something on my own terms, started to reignite. I was no longer skating along the surface; I was finally digging in.
Reprisal isn’t about rejecting AI. It’s about reclaiming the process, the mastery that comes from struggling and failing and then succeeding. AI is a tool—a powerful13 one—but it can’t replace the journey. Mastery doesn’t come from shortcuts. It comes from the climb, from feeling the weight of the mountain beneath your feet, and knowing that every step you take is yours alone.
Footnotes
-
GitHub Copilot is a paid service available for free to any student with a university-associated email address. There is no doubt in my mind that providing it for free for four years yields a excellent return on investment. ↩
-
I met with a Senior who had a RO to Microsoft in November of my freshman year. He described the long list of tasks which lay before me. ↩
-
Problem sets included the Grind 75 and the Neetcode 150, a set of programming problems covering common algorithmic techniques such as Two Pointers, Divide and Conquer, Greedy, Dynamic Programming, et al. ↩
-
Most companies process applications in a queue. Those who apply first are processed first. When a companies hires all the talent it requires, those remaining in the queue are tossed out. So, it benefits one greatly to be first. ↩
-
The incremental failure informs your context when problem solving. You lean on lessons learned from past struggles then endeavoring to overcome new ones. ↩
-
ChatGPT and GitHub Copilot provide automated code suggestions, allowing you to bypass deep comprehension of the code they generate. When it works, you move forward without gaining a true understanding, trading genuine learning for speed and the ability to quickly deliver results. ↩
-
This became evident during exams, where the results often showed a bimodal distribution, reflecting a clear divide between those who had developed a deep understanding and those who had relied on AI tools without fully grasping the material. ↩
-
It is incredibly low resistance to ALT-TAB into a web browser and type chat.openai.com and have your problem be immediately solved. It is even easier when AI tools like Copilot are integrated directly into popular code editors—including my editor Vim—allowing solutions to appear inline as you code. ↩
-
Little is more satisfying than seeing the words “All tests passed” after hours of hard work and frustration. It’s in that moment of triumph that the struggle feels entirely worthwhile. ↩
-
Forgive the imagery, for I have been rereading the Lord of the Rings. Imagine if Celebrimbor knew nothing of Ringcraft and instead outsourced his work to LLMs! ↩
-
I kept avoiding the inevitable, delaying the moment I would have to face myself. I felt a deep sense of disgust, yet I was equally lost without the crutch of AI tools. One night, the weight of it became unbearable. ↩
-
I have spoken to four other students who also acknowledged their over-reliance on AI. However, they too procrastinated in addressing it, knowing deep down that they hadn’t developed the necessary skills independently. ↩
-
AI makes you feel productive, but it has dangerous pitfalls. I think I’d much prefer to spend two hours reading documentation and man pages than waste four hours debugging AI-generated code in production—code that I don’t fully understand and therefore can’t confidently fix. I’ve come to the conclusion that its promise of productivity isn’t quite what it appears to be. ↩