🤖 on ai

Author

Affiliation

Ishan Buyyanapragada

UIUC

quick context

Hi, I’m Ishan. I’m in my spring break of freshman year right now, so not even a fifth of the way through my computer science degree. That isn’t to say that I’m having a tough time here. In fact, it’s quite the opposite, and that’s what I’m here to address.

(my) problem

Since the beginning of college—possibly earlier—I’ve been growing increasingly impatient. Specifically when it comes to thinking programmtically. I can’t be bothered to spend hours debugging an issue, or even figuring out the correct way to approach it. Forget coding: I can’t even spend time understanding fundamental concepts that are being taught in my classes. I’ve been trying to come up with a euphemism for my plight, but that would be doing it injustice.

I’ve become a glorified prompt engineer. ¹

I mean come on, it’s so simple. Stuck on a problem? Just GPT it. Didn’t work? Did you try Claude? Gemini? Grok? Deepseek? Copilot? …and on and on. It’s become increasingly uncommon to think: to look at what’s in front of you and try to figure something out yourself. Instead, we prefer to learn by recieving information in brief, matter-of-fact responses from large language models that use up tons of water to stay cool. ²

It became all the more apparent at a hackathon I attended a few weeks back. ³ The lack of time to build a project exacerbates the dependence on AI, almost to the a point where it turns into a contest for “Who Can Prompt Their AIs the Best?” It’s not at all surprising—why should I rely on my own flawed human knowledge when I have an much faster answer available at a button’s reach?

I’m not here to dismiss the profound beneficial impacts artificial intelligence can have on education, but I am here to try and remove its hand in turning me into an over-reliant and lazy learner.

(my) attempts at a solution

At first, the problem sounds fairly easy to fix. Just stop using AI! Restrict the websites on your browser if you have to. Embrace the pre-AI methods of googling and perusing until you find a solution to your problem.

I actually tried this strategy for the weeks leading up to spring break. I made it a point to avoid AI-use as much as possible on assignments (even for “guidance” or whatever else I used to tell myself to make it sound better in my head). I was back on StackOverflow, decrypting answers and how they could be applied to my specific problem. By the end of it, I was assuredly a smarter man. However, I was also a significantly slower one.

Assignments that would usually take me hours began taking weeks. I would have to block out large chunks of my day to work on projects that used be an afterthought. I realized that—in essense—AI has provided us unprecedented efficiency at the cost of foundational knowledge.

building a actual solution

So I came up with a list of things that this project would need to do:

📝 checklist

• stops unnecessary and lazy ai usage (i.e. copy pasting code and/or asking for solutions)

• guides me down a better, more educational path that forces me to think on my own (with some help)

• preserves hardcore learning and efficiency as much as possible ⁴

• optional: multiple levels of scrutiny depending on the situation

I settled on a Chrome extension that can monitor the way I’m using AI. It’s kinda like the ones that keep you on task, but this one is specifically to minimize lazy usage of AI for assignments. However I also wanted it to be as unbothersome as possible (I don’t wanna explain every move I take for the extension’s approval). In addition, it should gameify the process a little and reward thoughtful prompts as opposed to unhelpful ones.

It would be quite hypocritical of me to use AI to build this in a day or so, so it looks like I know what I’ll be doing for most of my spring break.

final product: ThinkFirst

It’s a extension that encourages critical thinking and mitigates lazy AI usage. Here’s a quick demo:

Video Player is loading.

Current Time 0:00

/

Duration 1:56

Loaded: 12.60%

0:00

Stream Type LIVE

Remaining Time -1:56

 

1x

• Chapters

• descriptions off, selected

• captions settings, opens captions settings dialog
• captions off, selected

This is a modal window.

Beginning of dialog window. Escape will cancel and close the window.

Text

ColorWhiteBlackRedGreenBlueYellowMagentaCyanTransparencyOpaqueSemi-Transparent

Background

ColorBlackWhiteRedGreenBlueYellowMagentaCyanTransparencyOpaqueSemi-TransparentTransparent

Window

ColorBlackWhiteRedGreenBlueYellowMagentaCyanTransparencyTransparentSemi-TransparentOpaque

Font Size

50%75%100%125%150%175%200%300%400%

Text Edge Style

NoneRaisedDepressedUniformDropshadow

Font Family

Proportional Sans-SerifMonospace Sans-SerifProportional SerifMonospace SerifCasualScriptSmall Caps

End of dialog window.

Footnotes

1. i’m not trying to disrespect any prompt engineers out there, but in my case it’s not a good thing to be right now.

2. kind of a tangent but there’s an undeniable parallel between the general shift to short-form media (tiktoks, reels, etc.) and our preference for short LLM blurbs that get us exactly what we want.

3. despite my qualms above, hackillinois itself was extremely fun. you can check out our second-place-winning project here.

4. obviously this would make me more ineffecient at coding, so i can’t expect it to be up to par with someone who’s prompting their way through an assignment. however, i’m hoping to strike a good balance between learning what i’m doing and being able to do it reasonably fast.