How Smart Is This Thing, Really? Spolier it’s efffing smart!
All of the applictions were made by claude in 10 -20 minuites with 1 or 2 prompts. there may have been variations but these are the sort of application anyone can now make.
When faced with a cutting-edge AI model like Claude 3.7 Sonnet, the first question on my mind was simple: What’s the hardest thing you can do?
I wasn’t interested in basic coding tasks or casual chatbot conversations—I wanted to see if this AI could truly push the boundaries of what’s possible in programming. Could it handle complex simulations? Could it build a game from scratch?
So, I threw two challenges at it. First, I asked for the most difficult thing it could code, and it responded by generating a particle physics simulation. Next, I wanted to test its ability to handle a real-world project, so I had it build a DOOM-style shooter. The results? Impressive, fascinating, and a little bit mind-blowing.
Task 1: A Particle Physics Simulation—Because Why Not?
When I asked Claude what the hardest thing it could do was, it didn’t hesitate. It jumped straight into the realm of particle physics—a field that requires serious mathematical and computational muscle.
The AI put together a simulation of particle interactions, complete with equations governing motion, forces, and collisions. It wasn’t just generating code; it was applying physics principles in a way that actually made sense.
This wasn’t a simple “Hello, World!” exercise—it was a real physics-based system that could be extended for research, simulations, or even game physics engines. I was starting to realize just how deep this AI’s coding abilities went.
Task 2: A DOOM-Style Shooter—Can AI Build a Game from Scratch?
After seeing Claude 3.7 Sonnet tackle particle physics like a digital Einstein, I wanted to shift gears—literally. Could it handle game development, a field that blends logic, physics, and creative design?
I decided to test it with something ambitious: a DOOM-style shooter. The original DOOM is a legendary game, known for its fast-paced action and innovative raycasting technique that created a pseudo-3D world in a 2D engine. It was a revolution in game design—so could an AI recreate even a simplified version of that?
I asked, and Claude confidently responded with a full-on plan. It outlined how it would structure the game, from movement mechanics to AI-driven enemies. And then, incredibly, it actually built it.
At first it may seem as though the enemies arent’ actually there but they are you just have to be very close to them untill you can see their face. If you fire then they will die with a shot or 2. very hard though and i think a glitch. I did try a few times to get claude to fix it and put buttons on for mobile view but i was already impressed and wanted to try somthing else.
Task 3: A Productivity App—Can AI Build a Usable Tool?
After seeing Claude 3.7 Sonnet tackle a particle physics simulation and a DOOM-style shooter, I wanted to test its ability to build something practical—something that could actually be useful in daily life.
So, I set it the challenge of creating a productivity app. The goal? A tool that could help track tasks, habits, and goals, with built-in reminders to keep users on track. It sounded simple on paper, but as anyone who’s tried to build a good productivity system knows, getting it right is tricky.
What followed was an iterative process—I went through a few different versions, tweaking and refining the AI’s output along the way. At first, there were some bugs—certain features didn’t work quite as expected, and some parts of the logic needed adjustments. But with each attempt, Claude improved the structure, and in the end, I had something surprisingly functional.
So, how well did it perform? Was it just a fun experiment, or could this actually be a useful productivity tool? Let’s dive into the details.
Task 4: A 3D Solar System—Balancing Complexity and Performance
After testing Claude 3.7 Sonnet’s ability to build games, simulations, and productivity tools, I wanted to push things even further—into outer space. Could it create a 3D solar system that was both visually impressive and scientifically accurate?
Claude initially generated a detailed planetary simulation, but as I tested it, I ran into a familiar challenge: performance vs. complexity. The original version was ambitious, but on slower devices or older browsers, it struggled to run smoothly. So, after a few iterations, Claude produced a simplified yet fully functional version, designed to work reliably across different systems.
The result? A sleek, accessible, and interactive solar system model that loads quickly and runs on almost any device. It may not be NASA-grade, but it’s a fantastic example of how AI can balance complexity with usability when building 3D simulations.
Task 5: A Music Generator—Can AI Jam?
After coding physics simulations and building games, I figured—why not make Claude write music? Could it actually generate something listenable, or would it just spit out a mess of random notes?
