Eashan Vytla

Computer Science Student

← Back to home

when-first-principles-thinking-isnt-enough

April 19, 2025

Over the past year, I’ve poured countless hours into Dreaming Falcon—my reinforcement learning research project. Despite my best efforts, the algorithm refuses to work. As I slowly learn what it means to be a better researcher, I’ve decided to start documenting the lessons I’m picking up along the way.

After reaching a dead end with the project, I paused to reflect. That’s when I found myself thinking deeply about First Principles Thinking. In theory, it’s a powerful method of inductive question-asking—breaking problems down to their most basic truths by stripping away all assumptions. It’s what initially inspired me to embark on Dreaming Falcon in the first place.

But here’s what I’ve learned the hard way:

First Principles Thinking assumes that a solution exists—and that you’ll know it when you find it.

In practice, that assumption doesn’t always hold.

Take SpaceX, for example. Their team famously asked, “Why can’t rockets be reusable?”—then worked backward from cost per launch to raw materials. It’s a textbook case of First Principles Thinking in action.

But even SpaceX had to endure years of iteration and failure before achieving reusability. Their success didn’t come just from questioning assumptions—it came from relentless testing and learning from what didn’t work.

So, here’s the question I’ve been wrestling with:

When my solution to a First Principles problem doesn’t work, how do I know whether I’ve reached a dead end—or if I’m just one step away from a breakthrough?

That question led me to a more nuanced mindset I’ve started calling Fundamental Problem Thinking.

Instead of asking, “How can I get this to work?”—I started asking, “Why doesn’t this work?”

Going back to Dreaming Falcon, here’s how this approach looks in practice:

Q1: Why isn’t Dreaming Falcon working?

A1: The RNN world model isn’t generalizing to an optimal solution.

Q2: Why isn’t the RNN world model generalizing?

And in all honesty—I don’t know the answer to that second question yet.

But maybe that’s the point.

Because if there’s one key lesson I’ve taken from this journey, it’s this:

Research isn’t just about building what works—it’s about understanding why things don’t.

That’s where the real breakthroughs begin.