I Got a Zero on a Beginner Python Assignment Because My Code Was "Too Good"
How AI paranoia in education is punishing self-taught developers — and why my sloppy code is the proof I wrote it.
Spots 
I'm 18. I've been coding since I was
I'm 18. I've been coding since I was 15. I built a 2D HTML5 game engine called Limn Engine that runs on a single JavaScript file. I co-founded a studio called Evolved Tech. I ship code every week. Last week, I submitted a Wordle-style word puzzle for my intro Python class (CSE 110 at BYU-Pathway). The comment: "It looks like you may have gotten too much outside help on this assignment." Let me tell you why that stings — and why it's a problem the entire industry needs to think about.
The assignment: build a word guessing game. Secret
The assignment: build a word guessing game. Secret word, hints (uppercase = correct spot, lowercase = present but wrong spot, _ = absent), guess counter, length validation. Standard Week 4 stuff. A main menu (Play / Help / Quit) Random word selection from a list A formatted help screen with examples and a rules table Sound effects (winsound) with cross-platform fallback Loading animations (time.sleep) In-game navigation commands (1 = menu, 2 = quit) Every single feature works. Every feature is mine. See python code here The instructor flagged it anyway. The Irony: My Code Isn't Even That Clean Here's the thing that gets me. If an AI had written my submission, it would have been cleaner.
My code has rough edges everywhere — the
My code has rough edges everywhere — the kind you only get from a human typing fast and shipping without a refactor pass. Here's the actual evidence: 1. The word list was a string, not a list A pro (or an LLM) writes: I didn't. I used a string and .split(" ") because it was faster to type. 2. I left a placeholder in the word list That's a 26-letter "word" sitting in my list of guessable words. It's impossible to guess. I forgot to remove it.
No AI would leave that in. It would
No AI would leave that in. It would break the game on a theoretical level. But I typed the list by hand and never cleaned it up. That oversight is my signature. 3. Bare except: instead of except ImportError: A pro writes except ImportError:. I wrote except: because I was moving fast. 4. Inconsistent variable names children (plural) vs adult (singular). An LLM would never do that. I did. 5. Duplicate entries in the word list temple, altar, hymn, prayer, mercy, kindness — each appears twice. Manual typing, no dedup pass.
I used winsound for sound effects — which
I used winsound for sound effects — which crashes on Mac and Linux. A pro would reach for a cross-platform lib. I built for my own machine and didn't care. Every one of those is a mark of authorship. They're not signs that I copied something — they're signs that I didn't. The Real Problem: Competence Looks Like Cheating Now Here's the uncomfortable truth about AI detection in 2026: The more competent you are, the more suspicious you look. The more competent you are, the more suspicious you look.
AI writes clean, feature-rich, well-structured code. So when
AI writes clean, feature-rich, well-structured code. So when an instructor sees a submission that's too feature-rich — too far ahead of the rubric — the default assumption has flipped from "this student is talented" to "this student cheated." That assumption punishes exactly the wrong people: Self-taught devs who learned outside the classroom Career changers with professional experience Students with real portfolios taking intro classes for credit Anyone whose side projects are years ahead of the curriculum The paradox: the more evidence you have that you can code, the more suspicious your classwork looks. I didn't get defensive.
I wrote my instructor a message that did
I wrote my instructor a message that did four things: Owned the situation calmly — acknowledged the concern was reasonable Explained my background — Limn Engine, Evolved Tech, years of shipping Pointed to the rough edges — the exact quirks above, as proof of authorship
Offered verification — line-by-line walkthrough, oral quiz, live
Offered verification — line-by-line walkthrough, oral quiz, live coding, or a simpler version that strictly meets the rubric That's the playbook. Not "you should have known better." Not accusations. Evidence + willingness + respect.
The instructor isn't the villain. Their job is
The instructor isn't the villain. Their job is to make sure students actually understand the material, and AI has made that job exponentially harder. If I were in their seat, I'd be paranoid too. What Educators Should Do Differently If you're an instructor reading this, here's what I'd ask: Look at style, not just features. Feature-rich ≠ AI-generated. Rough edges matter. Ask before accusing. "Can you walk me through this?" is a better first move than a temporary zero.
News
I Got a Zero on a Beginner Python Assignment Because My Code Was "Too Good"
How AI paranoia in education is punishing self-taught developers — and why my sloppy code is the proof I wrote it.
@spots #dev
Source: Dev.to