Start with the map network scientists draw, not the one the tourist board prints. In graph terms, a node’s betweenness centrality measures how much flow is forced to pass through it — the higher the number, the more the whole system depends on that one point staying up. Zurich is what happens when a single high-centrality node is fortified, taxed, and monetised for two thousand years.

The Wikipedia record of the city is, read coldly, a topology document. The Romans put a castellum on the Lindenhof — a morainic hill dominating the SE–NW waterway of Lake Zurich and the river Limmat — and named the customs post Turicum. The surviving tombstone detail (c. AD 185–200, discovered in 1747) reads P·P·STA·TVRICEN: head of the Zurich customs post, the office that collected the 2.5% Galliarum value tax on everything moving down the Limmat. That is not a founding myth. That is an edge in a flow graph with a toll booth bolted onto it.

The mechanism never really changed; only the payload did. A tax point becomes a vicus, the vicus a walled town of 38 hectares, the town a Confederacy canton, the canton — per the same record — “a hub for railways, roads, and air traffic,” with both the busiest station and the busiest airport in the country. Same hill, same corridor, higher throughput. Zurich Hauptbahnhof, 408 m above the sea, is the modern Lindenhof: kill it for a day and you do not lose a building, you lose a national schedule.

←TODAY: One municipality of 436,551 people packed onto 87.93 km² sits on the single busiest rail and air node in Switzerland — a 2,000-year-old chokepoint still doing its original job. →3012: The Zurich-3012 city survives because it published its own dependency graph and stopped pretending the Hauptbahnhof was redundant. Fulcrum: A node is only a single point of failure until someone draws it; the drawing is the defence.

Here is the part that matters for a design office. The city already hands you the graph. Open Data Zürich (data.stadt-zuerich.ch) publishes 275 datasets under Bevölkerung, 279 under Umwelt, 127 under Mobilität — free for commercial reuse. That is the raw adjacency list of a working city: where the flows are, where the density is (4,965 people per km²), where the queues build. Most practices never touch it. They design into the topology without ever mapping it.

Atelier: Every office running a coordinated BIM model already owns a single point of failure and calls it a feature. As PAZ’s own Archicad — En Obra panel puts it, the Virtual Building’s landmark is “the day-to-day office… where the single model quietly holds a project together” — which is exactly the property that sinks the project when that model, or the one person who understands its structure, goes down. Your Monday move: open data.stadt-zuerich.ch, pull one Mobilität dataset for your next site’s catchment, and drop it into your context model — so the city’s real flow graph is in the file before the first sketch, not argued about at Bauleitung.

Hack: Rank the chokepoints in your own supply before an outage ranks them for you. The lesson is graph centrality: the node everything routes through is rarely the one on your org chart. Sketch your dependency graph — sources, people, tools, the one unlabelled server — as edges, and let the math surface the hidden hub.

import networkx as nx
G = nx.Graph([("Lake","Limmat"),("Limmat","Lindenhof"),("Lindenhof","HB"),("HB","Airport"),("HB","Roads")])
bc = nx.betweenness_centrality(G)
print(sorted(bc, key=bc.get, reverse=True)[:3])  # your top 3 single points

Run it on the toy graph and “HB” tops the list — the Hauptbahnhof, unsurprisingly. Run it on your office and the third name on that list is usually the one nobody defends: the freelancer who owns the Grasshopper definition, the license server, the single distiller job.

I write this from far enough downstream to tell you how it ends. We did not run out of compute; we ran out of intact bandwidth and intact people who remembered how the old system worked. Zurich’s advantage was never the hill — it was that eventually someone drew the graph and taxed the truth of it. Draw yours before the outage draws it for you: not the architecture diagram, the dependency graph. The third single point you didn’t know you had is the one worth finding.

Source: en.wikipedia.org