The battle of the grammar fallacies: using Flint for English

Isaac Wu | High school student and Flint student representative

Guest Writer

Sep 17, 2025

Article summary

Hero image with icon and text saying "how students use flint for ai grammar practice"

In the 4th week of January, our class defeated the English midterm, the last one of the year. However, there was another foe to surmount: Grammar Fallacies. This infamous grammar exam was simply recounted by my friend Jack as “pain” and was heralded as the greatest challenge of Freshman year.

Making a grammar practice activity with Flint

With a relaxed yet still adrenaline-filled body, I opened my laptop that rested on a tall, circular table near the corner of the lively, post-exam student center. In the midst of chatter about the various sections on the exam, my fingers fluently typed out the URL to Flint, the AI platform my school provides for teachers and students. Then, I began to create instructions for an activity to help my classmates prepare for the Grammar Fallacies war. My mouse pointer swirling rapidly like the tip of a spear in battle, I scoured my school’s grammar website for any information on the exam and examples that I could append to the prompt for the chatbot. Having scouted the Grammar Fallacies battlefield, I produced intricate specifications on how the chatbot should behave, and I assessed the results. Subpar. The activity shouldn’t just give students the answer. Therefore, I tweaked the instructions, so the chatbot didn’t directly give the answer. Despite bewildered stares from my friends chilling on the coach in front of me, I iterated; I tested, and I tweaked. I was perfecting the ultimate weapon against the Grammar Fallacies.

Sentence mastery challenge AI activity for grammar practice and english

Building the school-specific grammar practice Flint activity was a profound experience that allowed me to improve my prompting skills and help my classmates learn. The structure of Flint’s activities and the built-in educational nature allowed me to directly write instructions for what made the chatbot unique rather than generic boilerplate. Furthermore, providing tailored rather than general practice, the chatbot, appearing to help the grade significantly on the exam, had nearly 200 sessions. Therefore, the use of the chatbot truly displayed how AI can be used to benefit education rather than as a cheating tool to our traditionally anti-AI English department and my school as a whole.

My previous experience with AI for students

It was not the first time that I had interacted with AI. In sixth grade, my previous middle school had used an adaptive learning software called Freckle to supplement our math education. However, some of my friends and I rapidly ascended to the highest level of the system by climbing through practice sessions like robots, one simple step after another. On the other hand, Khan Academy’s hike was not very scenic either. Therefore, the summer before seventh grade, besides swinging around weapons in preparation for a martial arts competition, I tried to create a math-practice software that presented a scenic hike rather than a monotonous one. I enlisted some friends to help me create this software.

At first, we attempted to hand-type unique math tutorials for the software, but all of us, including myself, made every excuse imaginable–from claiming our parents had set screen time limits to fabricating up other work we could do, like making a syllabus. This pain directed me to search for another solution: Generative AI. I put the detailed instructions we had made on how to create “educational articles” into the input box for GPT-3, an early version of ChatGPT. Then, I examined in awe as text emerged onto the screen. The work of an afternoon could be done in the time it took to use the restroom. However, I found a fatal flaw: the AI couldn’t create images. Nevertheless, I, hoping that the AI could still do most of the work in creating educational articles, optimized the prompt extensively until it could generate ninety-percent of the content. I showed the tool to my friends as if it were my pet.

Flint vs ChatGPT for AI for students. Flint provides the proper guardrails for AI safety, ethical use, and preventing cheating.

The release of ChatGPT was like the release of a new skin for a video game character–a mere visual change. With this new user interface, I continued to experiment. Feeling our current approach of making educational articles was just copying the softwares we despised, we went to ChatGPT to obtain some novel ideas. I first simply asked ChatGPT to generate some ideas on how to use technology to improve mathematics education. Generic, boring. We’ve already thought of those. Therefore, I prompted the tool to create some novel ones. Finally, something interesting. I asked ChatGPT to generate more and more ideas. After each response ChatGPT gave, I adjusted the prompt, and continued pushing the tool to give me an even greater number of intriguing ideas. These original ideas shocked my middle school Computer Science teacher and even the upper school CS department chair. Even today, the exceptional ideas and areas of research that this primitive model of ChatGPT found intrigue me.

The results of the Grammar Fallacies Flint activity

Once I had finished the activity, I began to type out an email to my grade, my fingers slightly trembling from the excitement. Finally, leaning back in my tall chair, I clicked send. The chatbot remained dormant for a while as the student center was consumed by chatter and piling assignments in the weeks leading up to spring break. One break later, we began to sharpen our weapons to fight against the Grammar Fallacies and I, distracting myself with statistics on the personalities of my classmates, waited patiently for my chatbot to help others. Soon, the exam charged towards our Freshman class. Open laptops filled the Stu, questions were shooting back and forth, and many people scrambled to the library during their free periods.

People finally began to use my Grammar Fallacies activity to sharpen their weapons against the exam. I noticed some of my friends using it and was happy to learn that the chatbot was working. However, my friends also asked for it to create more difficult problems, and they pointed out that it made some mistakes when the chatbot created grammar problems on the topic of misplaced modifiers. Thus, I played with my activity more and analyzed the results it gave. I found my friends were correct and added in the instructions the model. In this manner, the activity iteratively improved with my classmate’s feedback and was at the right granularity to sharpen a sword.

As the exam marched closer, the number of sessions people had steadily increased. People were using my AI! Little cheerleaders danced in my head. Not only were my friends who I had told about the chatbot directly using it but also other freshmen whom I didn’t know. I analyzed some of the sessions my friends made and saw that the chatbot was working–its questions were of similar difficulty with the exam and its answers were accurate. I knew the defenses had been set against the Grammar Fallacies.

Testimonial from Izzie, a high school senior who says that Flint makes her enjoy her learning and she's learning through experience, not just seeing it on a page

Finally, the Grammar Fallacies exam struck upon our grade. For 30 minutes, each student’s eyes were locked on the army of problems and the timer on the board, which was slowly counting down. In an otherwise silent battlefield, one could hear only the faint chatter in neighboring classrooms.

At the end of this exam, with a minute or so remaining, I pondered about my chatbot and how much it had helped my classmates. Had I just given an ineffective defence to my grade that would be obliterated in the instance of engagement?

Exam scores shot themselves into students’ inboxes; scores varied widely from perfect to failing. In the aftermath of the first battle with Grammar Fallacies, I conversed with various friends to gain insight into the preparation process. My friend Niam talked about doing all the practice tests he was provided and scoring highly on them, but still not obtaining the score he desired. Other classmates claimed they would “lock in,” or better prepare for, the battle of the retake.

To help my classmates prepare for their final chance at victory, I analyzed some of the sessions they had with the activity and again made some more minor adjustments to the instructions of the chatbot to slightly enhance the result and make the problems, which were like sparring matches, more closely resemble the difficulty of the actual exam.

After the retake, I asked my friend Jack, the same friend who described the exam as “pain,” about the quality of the chatbot, and he responded:

“[Flint] provided several incredibly helpful examples when it came to preparing for the Grammar Fallacies exam…it generated perfect questions.”

That he described the activity’s examples and questions as “helpful” for the Grammar Fallacy exam made the little cheerleaders in my head dance again. Both he and I came to the same conclusion on the activity: “it worked.”

Image showing AI grammar activity stats, with 185 sessions and over 50% of the class using the activity to improve their English learning

In total, 185 practice sessions were created by my Flint activity and over half of the entire grade had used my chatbot. This amount of usage of a creation of mine far exceeded anything else I had ever made. After I, in my English classroom, found this information, I looked out the window of my third-floor English classroom with a light smile. I pondered what other elegant defenses could be built next.

Academic Integrity Statement

I have not received help that violates the English Department Plagiarism Policy. If I received help or used outside resources on this assignment, they are cited or listed below.