ChatGPT Taught Me Special Relativity—And Changed How I Learn

Nityesh Agarwal—an engineer on Every's email management tool, Cora—credits ChatGPT with his career, so when OpenAI released its latest model, he spent hours playing around with it. Here he shares three powerful prompt techniques he learned in his late-night chat sessions. Make sure you scroll down to see Einstein's theory of special relativity illustrated in a series of comics. And if you want to hear more from Nityesh, tune in to this week's episode of AI & I.—Kate Lee

Was this newsletter forwarded to you? Sign up to get it in your inbox.

A couple of weekends ago, after OpenAI released o3, I asked ChatGPT to teach me Albert Einstein’s theory of special relativity. What followed transformed my thinking about the personalized teaching power of AI.

I’m a software developer, not a physicist. My last math class was eight years ago. But with ChatGPT as my infinitely patient tutor, I didn’t just memorize E=mc²—I actually understood why time slows down as you approach the speed of light. More importantly, I discovered prompting techniques that turn ChatGPT from a decent explainer into the best personal tutor I’ve ever had.

This isn't my first time with AI-powered learning. Two years ago, when ChatGPT-4 came out, I used these same techniques to teach myself Ruby on Rails, build a product that competed with venture capital-backed startups, and land my engineering job at Every. If ChatGPT can help someone like me grasp theoretical physics, what else becomes possible?

How I prompted ChatGPT until relativity made sense

I had to fire more than 100 questions at ChatGPT to understand an advanced science concept—questions like, “What does he mean when he says that there is no such thing as a universal now? Make me feel it,” and, “Can you do the math once again for 100 light-years when you're traveling at 50 percent of the speed of light?”

I did it all via voice‑to‑text: I turned on the mic and started rambling. Having a literal conversation with ChatGPT is what made it so effortless; if I had to type my questions, I wouldn’t have asked so many of them.

No matter how many times I asked it to "find another way to explain this concept to me,” ChatGPT encouraged my efforts to dig deeper. And it did so in colloquial language, like how a friend would text me (if he were smart and patient), which made it easier for me to follow and pursue strands of thinking. I wouldn’t have had the patience to sit with it for hours on weekend nights if its answers were like Wikipedia articles.

If 90%+ of your team isn’t using AI everyday, you’re already behind

You’re not going to get good at AI by nodding through another slide deck. Every Consulting helps teams level up—fast. We’ve trained private equity firms, leading hedge funds, and Fortune 500 companies. Now it’s your turn. Customized training. Hand-held development. A rollout strategy your team will actually use. Let’s make your organization AI-native before your competitors do.

Try us out today.

Want to sponsor Every? Click here.

To my surprise, I ended up using ChatGPT-4o most of the time instead of the new o3 model because 4o felt more adept at grasping the deeper intent behind each of my questions. 4o responded in natural language; ChatGPT o3 would have provided a table with every answer.

Here’s what I did specifically to get the most out of it:

Become a paid subscriber to Every to unlock this piece and learn the three prompting techniques that Nityesh used to supercharge his learning—and the main concepts of Einstein's theory of special relativity in comic form: