How I Use AI

My Philosophy

I finished my Bachelor's in 2022, just as ChatGPT was being launched. I think the first time I actually used an LLM, it was for something unserious like generating a movie script. Fast forward to 2026, and I'm working on my Master's degree while figuring out how to approach these tools that seem to get better overnight.

I have a lot of thoughts, but at the end of the day, I don't want to trade off the learning process for the efficiency gained from AI tools. In school, I'm not only giving up wages to be here, but I'm paying to be here; I would only be robbing myself if I let Claude earn my degree for me. It is tempting because of the efficiency, but I do believe there is a way to strike a balance:

• Familiar Topics: For areas I am already comfortable with (e.g., making a map of US Census Data), I don't mind using AI since I have a prior understanding of what the output should look like and what is reasonable.
• New Topics: For topics that are new to me, I want to do the work myself, using AI only as a guide or tutor when I get stuck.

Tools & Workflows

I mainly use Claude but have considered switching to Codex. Since I'm broke in grad school, my current balance is:

• First: Use my student version of Gemini for basic reasoning tasks, making outlines, auditing code, etc.
• Second: When I need to build, I use Claude (which I am currently paying for).

I have been trying to populate my portfolio while also managing the academic program, so I developed two approaches to using AI in my projects:

1. Portfolio Projects

I use AI as a tool to help me be more efficient generating code snippets, debugging, and writing documentation.

• Risk: It can be easy to mass-produce AI slop if my only goal is to increase project volume.
• Reward: I barely knew Java before this, and now I am able to design front-end applications. It is also a chance to show that I am embracing smart ways to use AI.

2. Academic Projects

I use AI as a tutor and to create study guides. I like Gemini's learning mode, and I will prompt any LLM I use for school to "Please not give me the answer" and instead walk me through problems in a dialogue. Since I'm in a technical Master's, most LLMs are quite capable of handling concepts in machine learning and econometrics.

My specific process for study guides involves:

1. Feeding materials to the AI to create a LaTeX template for a study guide, focused on free-response questions and key theories.
2. Moving the template to Overleaf and completing the guide by hand. Even if I don't know the answer, I "shoot in the dark" to help the concepts stick.
3. Uploading the final document and asking the AI to score my responses and mark errors—without explaining why yet.
4. Reviewing the incorrect questions and using "learning mode" to work through the right answers.

Ethics & Academic Integrity

I'm still working on the best workflow, but I think the key is to be transparent with professors and employers about how I use AI. I want to be clear that I'm not using it to do the work for me, but rather as a tool to help me learn. To ensure I am not violating any academic integrity policies, I always check with my professors before using AI for any assignments or projects.