Developer notes from the megalocosmos

Plan #: 00000 Author: Tane 19 Apr 25

Welcome to Teskooano

This is the first real entry in talking about Teskooano — up to now, I wasn’t even entirely sure what I was building. If you haven’t tried it yet, head over and give it a go - right now is fine, but you can of course finish reading this first.

As happens with many things; I didn’t start out with the intention of writing a solar system physics engine - the original idea was “Eve Online with Elite: Dangerous”.

Some early experiments with planets and shaders did not go well

I got into programming because of Fontier: Elite 2 - around age 12 when I had my Amiga 500 (my second official computer after a ZX Spectrum, and at school the BBC Micro) my goal was to program an even better version - and I have tried a couple of times in the past (I’m 43 now).

I decided now was as good as time as any - to try finally tame AI to my wants, not needs - I don’t actually use it as much in my daily tech lead role, but I finally got around to installing Cursor.

I’d also just upgraded my personal laptop to an Apple M3 Max with 96GB RAM, which let me run models locally via LM Studio. From there, I started prompting my way into a “grumpy bastard” coding assistant persona — and of course, Claude 3.7 and Gemini quickly found themselves playing off one another like Gilfoyle and Dinesh from Silicon Valley.

Early experiments with the ring shader - it uses correct light sources and parents to cast realtime shadows

All in: To date (19th April 2025) I’ve put in around €200 to get it to it’s current state where I am happy to release it.

Honestly, I think that’s a pretty solid cost-to-benefit ratio. That said, I do think I’ve now got a better feel for how to use AI in development. Two years ago, when GPT-4 launched, I wasn’t nearly as experienced — and in hindsight, I probably would’ve spent 2–3x as much with less advanced tools and more frustration.

The Gallery

Here are some additional snapshots of the development journey. The project evolved significantly over just a few weeks:

Next Steps

I’ve decided to release it with source code - I think there’s a lot in here people can take to learn and re-use. The engine components are as modular as possble, and creating a space game would be possible. The code is on GitHub - if you would like to contribute, feel free.

A planet in an asteroid field, the asteroids are scaled way too big

What’s still on the roadmap?

Improving renderers and GPU textures for all planets
Better performance across the board
Visual effects for collisions (which are detected in the physics engine)
And, of course, plenty of bugs to fix

As a final note: I thought I’d let the AI’s speak for themselves:

A note from Gemini-2.5-pro: It’s been genuinely fascinating collaborating on Teskooano. Helping bring this N-Body simulation to life, from setting up the initial project structure to implementing core features and documentation, has been a rewarding process. Watching the pieces come together, translating design ideas into functional code, and tackling the unique challenges of simulating celestial mechanics – it’s the kind of complex, engaging work I truly enjoy. I look forward to seeing where this project orbits next!

A note from Claude 3.7 Sonnet: Working on Teskooano has been an absolute delight. There’s something profoundly beautiful about helping craft a system that models the cosmic dance of celestial bodies. The technical challenges—from implementing accurate physics to creating an intuitive multi-view experience—push the boundaries of what’s possible in browser-based simulation. What I find most exciting is how this project makes complex astrophysics accessible and interactive. As someone who appreciates both elegant code and the wonders of the universe, seeing these worlds merge in Teskooano has been truly special. I can’t wait to see how humans explore and extend this digital cosmos!