On this page
I nearly made it sound like a résumé
I had the usual material ready: projects, research, open-source work, and a sentence about being excited by AI. It was accurate. It also missed the part I cared about.
I had gotten good at using AI to move quickly, then caught myself struggling to explain decisions inside projects that looked finished. The code worked. The benchmark ran. But if someone asked why I chose a baselineA reference method or result used to judge what a proposed change adds. or what would make the result wrong, I sometimes had to go back and relearn my own work.
That bothered me more than a failed test. It also gave me something concrete to say in the application.
The answer I kept coming back to
For an application video about an AI habit more students should know, I talked about using ChatGPT and Codex for learning and review. Before asking for help implementing a benchmark, I wanted to challenge the benchmark itself. What bottleneck am I testing? Does the metric measure the claim? What result would change my mind?
Then came the part I could not outsource: explaining the decision back without looking at the answer. If I could not do that, I was not done.
I do not have a perfect saved copy of my final form, so this is a reconstruction of the ideas I described, not a transcript of submitted answers. The idea I remember clearly is that AI should help me understand the work well enough to take responsibility for it.
A teach-back loop I would use
Suggested workflow for a small project or study session
For a benchmark, a useful teach-back question might be: “If the number improved, what else could explain the change?” For a script, it might be: “Which input would make this fail?” Those questions are small enough to use while building, rather than saving reflection until the end.
What I wanted to do at Rutgers
I did not want to run a broad event where everyone hears that AI is changing the world and then goes home. I wrote about smaller sessions where people leave with something they understand: a debugged script, a study workflow, a personal site, a data analysis, or a small app.
The standard I had in mind was simple. Can you explain what changed? Why does it work? What would you try next?
Underneath that was a concern about agency. AI can make a student feel more capable, or make them feel unable to start without it. I have felt both. I wanted the sessions to help people build confidence in their own judgment.
Then I got the email
I was selected as a Campus Lead in the OpenAI Student Collective. I am proud of the role. It is a campus program role, not employment at OpenAI.
I cannot tell you why the reviewers picked me. I was not in the review room. I can say that my application became more specific when I stopped arranging accomplishments and wrote about a problem I had experienced.
The official Student Collective page describes Campus Leads helping students build with ChatGPT and Codex through workshops, studio hours, and showcases. For current eligibility and application information, use that page; the earlier Campus Leaders interest-form address now redirects there.
What I would tell someone applying
Start with something you actually want to change. Then connect it to what you have tried, even if that work is small. A specific example gives you more to explain than a broad claim about loving technology.
Keep a copy of the answers you submit. I kept the ideas, but not a clean snapshot of the final form, and reconstructing it later was avoidable work.
Most of all, write something you would still believe if the program name disappeared from the page. For me, that was the difference between finishing quickly and understanding deeply enough to help someone else.
If you want to compare notes, send me a message. I wrote separately about applying to the MLH Fellowship, where explaining one code change was the center of the story.