Project Kessler

Inspiration

I didn't have a team. I had a vision and 24 hours. 27,000 pieces of junk are orbiting Earth right now. One chain reaction referred to as the Kessler Syndrome and we lose access to space for generations. No GPS. No ISS. No future. Just a graveyard of metal we can never clean up.

This isn't abstract to me. I want to travel to distant galaxies. I want to walk on a world nobody has ever walked on. See plants that have never been named. I'm 42 so NASA probably won't be calling. So eventually I'm just going to have to build my own way there. But none of that is possible if we trap ourselves inside a shell of our own debris first.

What it does

Project Kessler is a real-time first-person orbital remediation simulator. The AI autonomously navigates a live debris field, predicts collision trajectories, matches the rotation rate of tumbling objects, deploys a snare arm to capture them, and executes a controlled deorbit burn — replicating all three core systems from NASA patent MSC-25321 (TRL 6).

You are the mission operator. The AI handles the physics. You make the calls. The Gemini API powers a live AI commander that reads the actual simulation state on every meaningful event every lock, every cascade, every capture and responds with real tactical assessments. Not scripted. Genuinely reasoning about what's happening in the debris field in real time.

How we built it

Vanilla JavaScript and HTML5 Canvas. No frameworks, no game engines. Runs in any browser, zero setup. The navigation uses an artificial potential field algorithm the same class used in real autonomous vehicle research. Collision prediction uses time-to-closest-approach calculations from orbital mechanics. The rotation matching system tracks individual debris tumble rates and syncs the ship before snare deployment. The Gemini 1.5 Flash API receives live simulation context and returns tactical commentary on every significant event.

I started in C++ with CLion and SFML. CMake wouldn't link. Build kept failing. I scrapped everything and switched to JavaScript. Then I started feeding code between Claude and Gemini Claude helping me build it, Gemini living inside it as the mission commander.

Challenges we ran into

Two hours lost to a C++ build environment that refused to cooperate. A physics engine that kept launching debris into the void. Cascade events that would spiral into hundreds of objects before I could intervene. A rotation matching algorithm that synced at the wrong moment. A snare arm that grabbed nothing. There were moments I genuinely didn't think it was going to work. But the mission mattered more than the frustration.

I also tried very hard to incorporate eleven Labs API into the program but it just doesn't run.

Accomplishments that we're proud of

I built a working autonomous space debris capture system. Solo. In one night. In a browser. The snare arm extends. The rotation syncs. The debris burns up. It works. And somewhere underneath all of it is a real NASA patent doing exactly what it was designed to do. That's the accomplishment. It works.

What we learned

That frustration tolerance is a technical skill. That aerospace algorithms aren't as inaccessible as they sound potential fields, time-to-closest-approach, rotation matching. These exist in textbooks and patents, you just have to be willing to read them. That two AI systems can play completely different roles in one project. And that one person with a clear enough vision can build something that matters.

What's next for Project Kessler

Real orbital data. Actual tracked debris objects from live feeds, real trajectories, real threat windows. Multiplayer mission control: one operator per system, coordinating across a shared debris field. Hardware integration for the snare mechanism. And eventually, if the stars align a real conversation with the people at NASA who hold that patent.

Built With

• css3-tactical-hud-design
• custom-3d-projection-engine
• gemini-1.5-flash-api-(mission-strategy)
• html5-canvas-api
• javascript-(es6+)
• windsurf