It was 11:48 at night when I pasted the answer into our class group chat and hit send before my brain caught up.
I am a naval architecture student in Bangladesh, the problem set was due at 8 the next morning, and I had spent exactly 0 minutes on it. The AI had spent about 40 seconds.
Here is what happened the next day, and the line I now draw between using AI and hiding behind it.
The next morning, in class
The answer was correct. That is important, because everything that went wrong grew out of the answer being correct.
A classmate read it in the chat, got stuck on one step, and turned to me before the lecture started. "Bro, this line — why did you convert it like that?" He assumed I had written it. Reasonable assumption. I had sent it from my phone, under my name, at midnight, with a "done ✅" after it.
I looked at the line. I had genuinely never seen it before. Not "I forgot it." I mean I had read it the way you read a bus ticket — eyes over it, nothing kept.
So I did the thing I am least proud of in this story. I said, "Let me check, I did it late at night," and opened my phone and pretended to scroll. He waited. The lecturer walked in. The moment passed, and I sat through two hours of class thinking about a conversion factor I could not explain.
Was anyone actually harmed?
Not really, and that is exactly the trap.
Nobody failed anything. The homework got done. My classmate found the same step explained in a YouTube video within the hour. If you grade my midnight paste on outcomes alone, it passes.
But I keep replaying what I actually got from that night. I got sleep, sort of. What I did not get was the one thing the problem set existed for: knowing how to do that conversion. The next problem set had a similar step in it. Guess who was back at midnight.
The chat was not the problem either, by the way. Our group chat is how half my class survives this degree. People share notes, recordings of lectures, photos of the whiteboard. It is a good place. I just put something in it that had my name on it and none of my understanding inside it.
The part where I argue against myself
Now the honest counter-argument, because my classmates who use AI a lot are not villains, and some of their reasons are good.
Reason one: English. Most of us study in a second language. Our lectures, our textbooks, our exams are in English, and it is not the language we think in. Using AI to rephrase a paragraph, check grammar in a lab report, or explain a dense textbook passage in simpler words is not cheating. It is a ramp into the building. I do this myself and I will keep doing it.
Reason two: checking your own work. You solve a problem your way, then ask an AI to solve it too, then compare. If the answers match, good. If they don't, you find out where each of you went wrong. That is just a study partner who never sleeps. Nothing about that is dishonest.
Reason three: tutoring. "Explain why this step works" is a request for teaching. "Give me the answer" is a request for delivery. The tool does both happily, which is why the choice lands on us.
So where is the line? For me it settled into one test:
If someone asks me to explain a line of "my" answer and I can't, it was never my answer.
That test has no loopholes. It does not care whether AI, a friend, or a solutions manual produced the text. The moment I send something under my name, I am claiming I can stand behind it. If I cannot explain it, I am not submitting homework. I am submitting a screenshot of someone else's homework with my name typed over it.
The catch (it is not really about AI)
Here is the part I only admitted to myself later: AI did not invent this behaviour in me. Students copied each other's homework long before chatbots. I have copied a friend's lab report by hand, in pen, like an honest medieval monk.
What AI changed is the speed and the polish. Copying a friend took twenty minutes and looked copied. An AI answer takes forty seconds and reads like I have my life together. The tool removed the friction, and the friction was doing quiet work — it made cheating slightly annoying, and slightly annoying is sometimes enough.
So my rules below are not about being pure. They are about putting a small amount of friction back.
5 rules I actually follow now
1. If it's going out under my name, I explain it out loud first. Literally out loud, to my empty room. Any line I stumble on, I go and learn, or the answer does not leave my phone.
2. I ask for explanations before answers. My first prompt is "explain why this step works like I'm seeing it for the first time." The final answer comes last, if I still need it.
3. I rewrite it in my own words before it reaches the group chat. Typing it myself is slow. That is the point. Twice now, the real understanding arrived halfway through retyping.
4. I try the next problem alone before asking anything. Ten minutes of my own attempt first. Not because suffering is noble — because my questions get ten times more specific after I've failed properly once.
5. I label it. When AI helped me, I say so in the chat: "AI helped me crack this one, here's how it works." One sentence. Nobody has ever reacted badly. Half the time, someone replies with a better explanation than the AI gave.
Your turn
I am not asking whether AI belongs in a study group. It is already there, in every group, including yours. Everybody knows. Nobody says it out loud.
Where do you draw the line?
Is it at answers, at explanations, or at explaining-it-in-class?
And be honest — could you explain the last thing you sent?
Top comments (0)