“Cozy sandbox sandcastle scene” sounds like a gentle brief. It contains two of the least forgiving simulation problems in 3D, and a breakdown published on 80 Level on 28 September goes at both of them.
Why sand is hard
Sand is a granular material, and it occupies an awkward middle ground that neither of the standard tools handles well.
It isn’t a solid — it flows, slumps, and finds an angle of repose (roughly 34° for dry sand, which is why a sand pile is always about that steep and why it looks wrong when it isn’t). It isn’t a fluid either — it has internal friction, it holds a shape, it supports weight until it suddenly doesn’t.
Simulating it properly means either a particle method with friction — expensive, because convincing sand needs a lot of grains — or MPM (Material Point Method), which handles the solid-to-flowing transition well and is what most production sand uses now.
And the details audiences notice are specific. Wet sand behaves completely differently from dry: cohesion from water surface tension is what lets a sandcastle exist at all, so a scene with both needs both behaviours and a boundary between them. Collapse reads as a signature — everyone has knocked over a sandcastle, and they know what the slump looks like.
Why shallow water is worse
Deep water is comparatively easy to make convincing, because you mostly see the surface. Shallow water at a shoreline is the hard case:
- You see through it, so refraction and the wet sand underneath both matter
- The waterline is in constant motion, and the sand behind it is darker for a while after the water retreats — that lag is a strong cue and almost nobody models it
- Foam and thin-film behaviour at the edge are where the eye goes
- Volume must be conserved visibly, because a receding wave that doesn’t go anywhere reads as fake instantly
A sandcastle scene puts the shoreline and the sand in direct contact, so the two simulations have to interact rather than merely coexist.
Why this is the good kind of exercise
The cozy framing is doing something smart. A stylised, warm, low-stakes scene lets you simulate honestly without needing photorealism, which is the expensive part of fluid and granular work.
That’s a genuinely useful strategy for anyone learning simulation. Photoreal water needs enormous resolution, careful shading and long renders; stylised water needs correct behaviour and can be shaded simply. Behaviour is the part that transfers — an artist who understands angle of repose, cohesion in wet sand and volume conservation can apply it at any level of realism. An artist who has only ever tuned a photoreal preset cannot.
80 Level’s breakdowns tend to cover decisions rather than button presses, which is what makes them worth reading when you’re building something unrelated. For creative technologists: the underlying lesson is that simulation credibility comes from getting the material’s characteristic failure right — how sand collapses, how water leaves a mark — not from resolution.