Accumulation patterns

The engineering of accumulation—from loose parts to living structure 

A skeleton is just wood joinery that learned to move.

That's the observation that started this, and it holds up better the longer you sit with it. A mortise and tenon joint is a peg locked into a socket, built to keep force traveling cleanly between two rigid pieces. A ball-and-socket joint is doing almost exactly the same job, except one side of it is alive and the whole assembly has to keep working while it walks around. A dovetail resists being pulled apart through interlocking geometry alone, no glue required. So does the fibrous mesh holding a tendon into bone. Different materials, same underlying question: how do you get separate pieces to behave like one structure without simply melting them together?

That question turns out to be bigger than joinery. It's the actual question underneath every accumulation of anything — sand, cells, tissue, sediment, a pile of parts that either becomes a structure or stays a pile.

Start with the obvious trap: assuming contact is enough. Pour sand into a heap and the grains are touching each other everywhere, and yet the heap isn't a structure in any meaningful sense — kick it and it just redistributes into a slightly different heap. Contact alone gets you coincidence. It doesn't get you an accumulation that holds. For that, you need something extra, and it turns out to break down into five separate requirements, not one.

The first is cohesion — the elements involved have to actually prefer staying in contact over separating. Surface tension does it for water droplets. Chemical bonding does it for a lattice. Adhesion does it for cells forming tissue. Without this, nothing has accumulated. It's just sitting near other stuff, temporarily, waiting for a reason to leave.

The second is connectivity, and this one has a name in physics: percolation. Local contact between neighbors isn't enough on its own — what actually flips a pile of separate elements into a single system is when those local contacts chain together into a network that spans the whole thing. Below that threshold, you have clusters. Cross it, and you suddenly have one connected structure instead of many small ones. The interesting part is how sudden this transition is — nothing looks structurally different one contact before the threshold and one contact after, except that afterward, the whole thing is now one object instead of several.

The third is the part that surprises people: even once something is fully connected, not every part of it is actually doing work. Push on a dense pile of sand or grain and the load doesn't spread evenly through every contact point — it travels along specific chains of grains that happen to be aligned right, while plenty of other grains nearby are barely loaded at all. These are called force chains, and the blunt way to say it is: the structure is held by a subset, not by the whole. Most of what's "in" an accumulation isn't what's holding it up. It's just present.

Sometimes there isn't even a discrete path to point to. In fabrics, tissues, and foams, stress doesn't travel along chains — it spreads across a continuous field, with no single point doing the holding and no single point that fails first. This is where filler material earns its keep, and it's worth being precise about what that means, because "filler" sounds like padding, like something added for lack of a better idea. It isn't. A structure that has to survive being squeezed, twisted, and impacted from unpredictable directions can't rely on a few discrete load paths the way a sand pile can — it needs something that distributes stress across a whole field instead, and that's a completely different engineering job.

This is exactly the job fat does in a body, and it's a better illustration of "filler" as serious infrastructure than the word usually gets credit for. The eye socket is a small, brutal case study: an expensive, water-sensitive, constantly-active piece of neural tissue, hanging in a bony cavity, needing to survive impact, temperature swings, and constant motion without being damaged. Evolution's answer wasn't to build more bone around it. It was orbital fat — a deformable field that cushions, insulates, and lets the eye move freely, doing structural work without being structural in the rigid sense at all. The same material shows up doing the same job at the knee, where fat pads and bursae sit in the gaps between hard, grinding surfaces specifically so the hard surfaces don't grind each other apart. Cheeks, heels, joints, the tissue around major organs — the pattern repeats everywhere a body needs cushioning, insulation, or shape without paying the metabolic cost of growing more bone to get it.

I once watched a video of expanding packing foam and said it was exactly what fat does, minus the can. That's a more precise comparison than it sounds like. Packing foam fills irregular gaps, cushions whatever's inside it, and hardens into a shape that holds — all without being load-bearing infrastructure the way the box itself is. Fat does the same job with a serious upgrade: it's alive, so it can be metabolically active, redistributed, grown or shrunk on demand, and repaired continuously instead of shipped once and thrown away. Evolution didn't invent packing foam. It just got there first, and built a version that heals.

Which is where cost enters the model, and this is the part that ties everything together. Every part of an accumulation — force-chain member or filler alike — has an ongoing price attached to it: energy to build, energy to maintain, risk of being where it is. A structure isn't a fixed, finished thing. It's under continuous evaluation. High cost paired with low contribution gets trimmed. Low cost paired with low contribution gets tolerated, because removing it isn't worth the trouble either. High contribution gets preserved regardless of cost, because losing it costs more than keeping it. And there's a sharper case than simple low contribution: some parts don't just fail to help, they actively interfere — oversized forelimbs dragging on a system built for hopping, for instance — and those get selected against harder than parts that are merely useless.

Even so, nothing ever quite reaches zero. Removal itself has a cost. Some tiny residual utility usually remains, some constraint blocks full deletion. So real systems don't settle at absence — they settle at minimum viable presence, the smallest footprint that's still cheaper to keep than to finish removing.

None of this is about flow, and it's worth being clear about that distinction, because it's easy to blur the two. What makes a structure work once it exists, and how it eventually fails — that's a separate question, about the paths force and energy travel once there's something for them to travel through. Accumulation Patterns is the layer underneath that question. It's not asking how the structure works. It's asking why there's a structure there to work at all — why this particular pile of parts became one connected thing capable of holding load, instead of just a temporary arrangement waiting to fall back apart.

A wood joint holds because someone deliberately shaped the pieces to interlock. A skeleton holds because millions of years of trial and cost pruned every configuration that didn't. A body's soft tissue holds its shape not because it's rigid, but because a deformable field is cheaper to grow and maintain than the bone that would otherwise be needed to do the same job.

Things don't hold together because they're connected.

They hold together because staying together turned out to be cheaper than coming apart.