tipping points · metaphor 16 of 100

Nothing works —
until everything does.

Why does a movement, a rumor, a style smolder invisibly for years — and then, with barely any change in its ingredients, suddenly reach everyone? Connectivity is not gradual. Below a critical density nothing spans; above it, everything does.

Organizers know the despair of the almost: the mailing lists, the meetings, the pamphlets that circulate in pockets and die at the pocket's edge — year after year of "nothing works." The tactics are refined, the message is polished, the meetings grow slightly larger, and still the idea travels three handshakes and stops. Then the same tactics, one year later, catch. Not gradually — the way a struck match catches.

The temptation is to ask what changed about the message. Often nothing did. What changed was the density of people willing to pass it on, across a line no one could see. The instrument below is the smallest world in which that line exists — and it is exactly as sharp as it feels.

site percolation · 80×80 square lattice · every number computed live

does not span
view
Density of the willing · p
55.0%
0% pc ≈ 59.3% 100%
receptive sites
pockets (clusters)
largest cluster
largest ÷ receptive
jump to
Each site is independently receptive with probability p. Drag the density slowly through 59.3% and watch the largest cluster: a pocket, a pocket, a pocket — the world.
What this grid has seen · accumulated from your samples
spanning odds largest-cluster share

Every drag and resample logs one honest sample at its density. Sweep the slider a few times and the knee at pc draws itself. (Storm-damaged grids aren't logged — they're no longer fair samples.)

θ(p) = P(a site belongs to the infinite cluster)  ·  θ = 0 for p < pc,  θ > 0 for p > pc The sharpest kind of change: not steep, but off-then-on. For the infinite square lattice pc ≈ 0.592746 — known only numerically. On this finite grid the switch is slightly blurred, which is why the same density can span one resample and not the next.

the lattice

Pockets and the world.

Below the threshold, mathematics is brutal about what effort buys. Cluster sizes are exponentially bounded: there is a characteristic pocket size, and the odds of a pocket much larger than it decay exponentially. You can work harder inside this regime — better pamphlets, warmer meetings — and what you get is a richer pocket. A livelier scene, a denser correspondence, a deeper loyalty among the already-reached. What you cannot get, at any price payable in local effort, is reach. The idea still dies at the pocket's edge, because the edge is where the willing run out.

Above the threshold, an object of a different kind exists: a single cluster comprising a finite fraction of everything, from which almost anywhere can be reached. It does not grow smoothly out of the biggest pocket; at p_c the pockets suddenly find each other. This is why identical tactics fail for years and then succeed: the tactics were always operating within the medium, and reach was never a property of the tactic. It is a property of the medium's density — which was rising, invisibly, the whole time.

what to try

Sixty seconds at the threshold.

Sweep it. Drag the density slowly from 50% to 65% and watch the largest ÷ receptive readout. Through the low fifties it crawls — a few percent, the biggest pocket in a world of pockets. Around 59% it lurches, and by 62% the largest cluster usually holds most of every receptive site on the grid. A seven-point change in ingredients; a categorical change in reach. Each pass also plots itself on the curve below the grid, and the knee appears exactly where you felt it.

Spark it. Fire the left edge at 50%: the front eats a few columns of pocket and starves. Fire it at 65%: it crosses. Same spark, same rules of spread — the only difference is the medium. Then storm it: set the density to 60%, confirm the badge says it spans, and remove a random 2% of the willing. Watch the spanning cluster shatter back into pockets, often on the first strike. Victory just past the threshold is a filament, not a fabric — connected through narrow bridges any one of which is load-bearing. And resample a few times right at p_c: the same density spans one draw and not the next. At the critical point, history is genuinely written by luck.

the medium

The threshold is a property of the medium.

p_c ≈ 0.593 is not a truth about any message. It is a truth about the square lattice — about who can hear whom. Rewire the neighborhoods and the threshold moves: on a triangular lattice, where each site has six neighbors instead of four, it drops to exactly one-half. Add even a sprinkling of long-range ties — connections that leap across the grid instead of crawling along it — and it falls drastically. On networks with genuine hubs, the kind that broadcast to thousands, the threshold plummets toward zero.

This is why one talk-show mention outweighs a thousand pamphlets. The pamphlet raises p by a rounding error; the broadcast rewires the lattice, stitching distant pockets together with bonds that no amount of door-knocking could lay. Every spreading effort faces the same fork: you can raise the density of the willing, one costly conversation at a time — or you can change the geometry of who hears whom, and move the threshold to meet you.

reading defeats

Reading your defeats.

Pre-threshold failure looks total because reach is the only thing visible from outside, and below p_c reach is identically zero. But look at the grid at 55%: it is full. Forty-odd percent of everything is receptive, organized into real pockets with real internal life. Raising the density from 40% to 55% changes nothing an outsider can measure — and it is most of the work. The pockets are the progress. The year everything "suddenly worked," almost nothing happened; the decade in which nothing worked is when everything happened.

The strategic question, standing in the ashes of another campaign that died at the pocket's edge: raise p, or rewire the lattice? Persuade more people to pass it on, or change who can hear whom? And the moment the answer finally arrives — the first year the thing spans — remember the storm button. A movement just past its threshold is at its most euphoric and its most filamentary at the same instant. Feeling unstoppable is not the same as being fabric.

the mapping

Mathematics ↔ life.

MathematicsLife
receptive siteA person willing to pass it on — not a believer, just a conductor.
density pThe prevalence of the willing: the slow, invisible ingredient that years of pocket-work actually change.
clusterA pocket where the idea circulates and dies — the scene, the cell, the mailing list with no exit.
spanning clusterThe movement that can reach anyone from anywhere: a different kind of object, not a bigger pocket.
pcThe invisible line between "nothing works" and "everything works" — crossed without ceremony, noticed only afterward.
the filamentary giantA young movement's fragility at the very moment it first feels unstoppable: spanning, but by threads.

where the metaphor tears

Three honest failures.

Real social networks are not lattices.

The square grid is the conservative case: every site has exactly four neighbors, and reach must be earned locally. Real networks have heavy-tailed degree distributions — a few people connected to thousands — and on such networks the percolation threshold can be pushed toward zero. In hub-rich media, "everything spreads," and the drama of the sharp threshold largely dissolves. If your world runs through broadcasters, the lattice understates how easily things travel — and how easily bad things do.

People are not fixed sites.

Here each site's receptivity is set by an independent coin flip and never changes. Real people watch their neighbors: willingness to pass something on rises as adoption becomes visible, and collapses when it stalls. That is the territory of threshold models and cascades — where p moves while you measure it, and where the act of spreading changes the medium it spreads through. Percolation freezes precisely the feedback that makes social contagion interesting.

Paths are not persuasion.

Percolation is about connectivity, nothing more. The spanning cluster means a route exists from anyone to anyone through willing intermediaries — it does not mean the message survives the journey intact, or that anyone at the far end is convinced. Reaching everyone is not converting everyone; it only means the road exists. Many movements have crossed their percolation threshold and discovered that reach was the easy part.