Three prototypes, a lot of failed prints, and RocketBuddy

Before RocketBuddy became something you can plug in, it went through three full prototypes and more test runs than we bothered counting. Here's the honest version of how it came together.

Prototype 1: getting the shape right

The first version existed to answer one question: does a rocket-shaped shell actually diffuse light evenly, or does it just glow in weird patches? We 3D printed the first RocketBuddy body to find out. It answered the question — badly. Light pooled at the base and barely reached the nose cone, the fins were too thin and snapped if you looked at them wrong, and the whole thing wobbled on its base. Useful information, disappointing object.

Prototype 2: fixing what broke

Round two went back to basics on wall thickness and fin geometry. We adjusted the shell thickness until the light spread evenly from base to nose, thickened the fins so they could survive being packed and shipped, and reworked the base so the electronics actually sat flush instead of floating. This is also where USB-C entered the picture — the port needed to sit somewhere that didn't interfere with the base's stability or the internal wiring, which took more iterations than we expected for something that sounds simple on paper.

Prototype 3: the one we were happy with

The third 3D printed model was the first one that felt like a real product rather than a test piece. Even light diffusion, fins that survived a drop test, a base that sat flat, and enough internal room for the RGB electronics without the shell feeling bulky. This is the version we used to validate everything else — the RGB behaviour, the default takeoff sequence, and later, the MCP connection to Cursor and Claude through Ponos Launch.

What "extensive testing" actually meant

Not glamorous, mostly repetitive:

• Light tests — every prototype got powered on and checked for even glow across the shell, in a dark room and under normal desk lighting.
• Fit tests — rocket body to base, base to USB-C cable, dozens of times, checking nothing loosened with repeated use.
• Drop and knock tests — desks are not gentle environments. Fins and the nose cone got knocked around on purpose to see what would actually survive daily use.
• Software integration tests — once Ponos Launch was far enough along, we ran it against real Cursor and Claude sessions to check the lighting actually matched what the agent was doing, not just what we assumed it was doing.

Why this mattered

We could have shipped after prototype one and called it done. It would have glowed unevenly, the fins would have snapped in transit, and we'd have spent the first month handling returns instead of building the next thing. Three prototypes and a lot of testing later, RocketBuddy is the version we're comfortable putting on our own desks — which was always the actual bar.