Another full moon, another moonthly LithoBreak update.

This one is different.

The old moon terrain is gone.

In its place: an in-progress preview of LithoBreak’s new voxel moon system.

This has been years in the making.

We have wanted mineable, real-scale voxel moons for a long time.

The problem wasn’t desire; it was figuring out how to do it right.

As far as we can tell, no one has implemented voxel moons at this scale using a Body-Centered Cubic (BCC) grid. Most voxel guides, engines, and tutorials assume cubic grids. Almost everything online points you toward marching cubes on cube voxels.

That works, but it wasn’t what we wanted.

The vision for LithoBreak leans heavily into exploration.

Real terrain has strange formations, unexpected slopes, ridgelines, bowls, impact scars — places that feel like no one has stood there before.

Empedocles alone is roughly 90,000,000 square kilometers of surface area.

There are multiple moons in the Aristarchus system.

When you choose a direction and walk for kilometers, it should feel plausible that no one has been there.

Voxel moons let us do that.

Image with the character circled in red

Zoomed out Image with the mining equipment circled in red

We also wanted the terrain to be:

• Mineable

• Physically meaningful

• Volume-consistent

The regolith that goes into your rover is the same volume removed from the moon surface.

Surface coloration hints at what ores and chemical materials are present beneath.

You’re not harvesting from an abstract resource node.

You’re carving into the moon.

Both cubic voxels and BCC voxels:

• Tile infinitely

• Can be smoothed

• Can scale to realistic sizes

• Have tradeoffs

But the way they connect differs.

A cubic grid has:

• 6 direct face-sharing neighbors

• 8 corner-sharing neighbors

A BCC grid has:

• 14 nearly equal-distance face-sharing neighbors

In practice, that means BCC can smooth more naturally in all directions, behaving more uniformly when up isn't a static vector direction. Moons are spheres, so 'up' is relative.

So, we chose the road less traveled. It took longer as the guides are almost exclusively cubic.

But we are Tau Ceti Labs. We are here to experiment.

Image of non-smoothed BCC grid mesh

Each moon has its own:

• Generation algorithm

• Material distributions

• Surface characteristics

Terrain is generated in two different ways for rendering and simulation. When mining the surface, both visuals and physics colliders update to reflect the edits.

Visually, the moon is broken into rendering chunks at multiple Levels of Detail (LODs).

We use 21 LOD levels, allowing us to render a distant orbiting moon with a few dozen triangles all the way to representing single 0.5 cubic meter voxel directly below your boots. Everything in between is covered by the rendering system increasing or decreasing the visual fidelity in real time, 1000 km orbits to hills 10 km away.

This is what makes real-scale moons even remotely feasible.

Visual meshes are only half the story.

We also generate terrain specifically for:

• Character collision

• Mining sites

• Rover traversal (soon)

For the local player, only 1:1 voxel terrain immediately surrounding them is generated. This is the highest detail the terrain gets, and this is intentionally kept small and locally computed so that it can generate rapidly. As you move, old terrain is removed from memory and new terrain is generated ahead.

You can walk for kilometers and always have something solid beneath your boots.

The same is similar for mining site terrains, only the generated terrain is much larger and the position is fixed. This is to facilitate rover physics and equipment placement with true 1:1 voxel terrain.

Simulation of terrain being mined

Currently you can explore in first-person and use a third-person mining console.

You can see distant hills and begin to see curvature — but not dramatically yet. Game render distance is currently hard-locked to about 50 km given floating point precision. That might sound large, but our real-scale Empedocles moon is 2700 km radius, which itself is small compared to the distances of other orbiting bodies.

But wait, how do we render those distant orbitals then?

We squish everything down dramatically. As long as the distance pulled in is proportional to the scale and we track the camera movement, it looks the same in game. We will be doing the same process to the visual LOD system over the coming moons, where voxel by voxel the meshes warp to keep within that rendering distance.

Let’s be transparent at this early stage.

• It can hit ~60 FPS in development mode

• Loading hitches can be severe

• Visual terrain does not stream in fast enough yet

• Running across the surface can cause LOD drops underfoot

• Catch-up time can take tens of seconds

Optimization will come in time, but optimizing too early is simply extra work. Multi-threaded mesh generation and smarter streaming are still being implemented to make it playable, but polish will come later.

This build is a preview, not a finished terrain system.

With terrain generation and editing working:

• True ore veins become possible

• Multiple mining sites across a single moon

• Vehicles for meaningful traversal

• Console-based site switching

• Persistent carved terrain

This is the foundation for Stage 3 and beyond.

Procedural Terrain has been pulled forward into Stage 2. We made this change because if we were going to do it, it needed to come sooner so it could be part of the LithoBreak foundation.

The old terrain is gone; this preview is the new baseline.

Main branch updates continue landing each full moon.

This one isn’t flashy polish, it’s infrastructure. The kind that makes the rest of the game possible.

Next up:

• Vehicle Preview

• Metals

Thanks for sticking with us while we build something ambitious.

— Tau Ceti Labs 🌕