Around 1185 BCE, a scribe in the palace of Ugarit — a rich coastal city on what is now the Syrian shore — pressed a cuneiform tablet with an urgent message. The king of Alashiya, on Cyprus, had written asking where Ugarit's ships were. Ugarit's king wrote back that his fleet was away in Lycia, that enemy ships had already appeared off his coast, that seven of them had done damage, and that he had no forces at home to respond.
Another tablet, unfired and never sent, was found in the ruins of the house of Rapanu. It contains a draft reply to a plea for grain from the Hittite court. The grain is promised. There is no evidence it ever arrived.
Ugarit burned shortly after. The tablets survived because the fire that destroyed the city baked the clay hard enough to preserve what the scribes had been writing in the last weeks of their world. They describe, in real time, a palace administration trying to manage a cascade it did not understand and could not stop. Ships that should have been home were somewhere else. Grain that should have been moving wasn't. Enemies that should have been distant were on the coast. Every request for help was being answered with another request for help.
This is what institutional collapse looks like from the inside. Not drama. Paperwork.
I'm opening with Ugarit because the Late Bronze Age collapse — the staggered, decades-long unraveling of the eastern Mediterranean palace societies around 1200 BCE — is the best-documented example we have of what happens when a densely connected, prosperous, complex system runs out of buffer. And because the analytical structure of that collapse, as it's now understood by historians and network scientists, maps onto the world of 2026 in ways that should make us uncomfortable.
Not because we are doomed. We aren't. But because the mechanism is recognizable, the warning signs are legible, and the window for doing something about them is almost always shorter than the people living through it believe.
The System That Failed
For most of the 20th century, the standard story of the Late Bronze Age collapse was simple. The Sea Peoples came. Cities burned. A dark age followed. It was a story of invaders and victims, with a clear cause and a clear effect.
That story is no longer what the evidence supports. What the evidence supports is stranger and, for our purposes, more useful.
The eastern Mediterranean around 1300 BCE was one of the most interconnected regions the ancient world had ever seen. Egypt, the Hittite empire, Mycenaean Greece, the Levantine city-states, Cyprus, and Assyria were bound together by diplomatic correspondence, royal marriages, tribute networks, and a continuous flow of goods through a maritime trading system of remarkable sophistication.
Uluburun shipwreck, c. 1320 BCE
for cargo on a single ship
to systemic collapse
The Uluburun shipwreck — a single merchant vessel that went down off the Turkish coast around 1320 BCE — was carrying Cypriot copper, Anatolian tin, Canaanite jars, Egyptian faience, Baltic amber, African ebony, ivory from both hippopotamus and elephant, and glass ingots of a kind found nowhere else in the archaeological record. A single boat. One trading voyage out of thousands.
This wasn't just trade. It was the physical substrate of palace power. The legitimacy of Bronze Age rulers rested on their ability to acquire, redistribute, and display prestige goods that came from somewhere else. The bronze that gave the era its name required tin from Afghanistan or Cornwall and copper from Cyprus, brought together in workshops that depended on continuous long-distance exchange. A Hittite king who could not supply bronze to his armies was not a Hittite king for long. A Mycenaean palace that could not move olive oil and textiles through its networks was a palace whose scribes stopped being paid.
Complexity and prosperity were the same thing. So were complexity and fragility.
Then, beginning sometime in the late 13th century, the system began to fail.
Climate proxies from across the region — pollen cores, cave deposits, tree rings — record a shift toward drier conditions. The most precise evidence comes from central Anatolia, where dendrochronologists have identified an unusually severe multi-year dry episode around 1198 to 1196 BCE, the kind of concentrated drought that overwhelms the risk-buffering practices of even well-organized agricultural societies. The palace economies had grain storage and redistribution systems designed to handle a bad year. They were not designed to handle three bad years in a row on top of a trade network that was already under stress.
The Ugarit tablets show what that stress looked like at the ground level. Urgent appeals for grain from multiple directions. Promised shipments that were delayed or never arrived. Small-scale raiding that the palace could normally have absorbed but suddenly couldn't, because the fleet was elsewhere, the treasury was thin, and the reinforcements that should have come from allies were occupied with their own problems.
Within roughly fifty years, most of the major centers of the eastern Mediterranean had been destroyed, abandoned, or reduced to a scale that their predecessors would not have recognized. Ugarit never came back. The Hittite capital at Hattusa was burned and left empty. Mycenaean palace civilization ended. The long-distance trade in tin and copper collapsed. Writing systems were lost in some regions and would not return for centuries. The Egyptian state survived but emerged from the crisis visibly diminished, never again the regional power it had been under Ramesses II.
Something had broken, and it wasn't one thing. It was the accumulated interaction of several things at once.
The Sea Peoples Problem
This is where the Sea Peoples come in, and where the analogy to the present starts to sharpen.
The Sea Peoples are known primarily from Egyptian sources — inscriptions on the walls of the mortuary temple of Ramesses III at Medinet Habu, a few other royal texts, a scatter of references in diplomatic correspondence. The Egyptian reliefs show naval battles against confederated groups with names like the Peleset, the Shekelesh, the Denyen, the Tjeker, the Weshesh. Ramesses III claims to have defeated them in a great battle around 1175 BCE, saving Egypt from a fate the inscriptions describe as having already befallen the Hittites and the Levantine coast.
For a long time, the Sea Peoples were read as the cause of the collapse — a sudden exogenous invasion that shattered the eastern Mediterranean. The image was of a great migratory wave crashing against coastal states that could not resist it. The reliefs even show ox-carts carrying women and children, suggesting whole populations on the move rather than simple raiding parties.
The problem with this reading is that it doesn't survive contact with the rest of the evidence. The collapse was not a single event. Different cities fell in different decades. Some were destroyed violently, some were abandoned peacefully, some were reoccupied at reduced scale. The timing doesn't fit a single invading force. The Hittite capital was destroyed before the Egyptian battles. Some Mycenaean centers fell and were reoccupied. The "wave" is not really a wave when you look closely at the chronology.
Modern ancient DNA work has added an unexpected piece of evidence. Analysis of early Iron Age burials at Ashkelon, a Philistine city on the coast of what is now Israel, has identified a detectable European-related ancestry component in individuals from the period shortly after the collapse. Within a few generations, that signature had been diluted by intermarriage with the local population. The DNA tells a more interesting story than either the traditional invasion narrative or the skeptical "no real migration" counter-narrative. Real people moved. They came from somewhere, and they settled. But they didn't replace the populations they encountered, and they didn't sustain their distinct identity for long.
So what were the Sea Peoples, if not a master cause?
The most persuasive contemporary reading treats them as something more interesting than an invasion. They were a mobility phenomenon — displaced populations, opportunistic raiders, mercenaries without employers, merchant communities whose trade had collapsed, coastal groups whose own homes had failed — moving through a maritime system whose coastal nodes had already lost the capacity to defend themselves.
The Sea Peoples didn't cause the collapse. They were the shock-transmission mechanism that propagated it across nodes that were already weakened. The drought had thinned the buffer. The trade disruption had hollowed the palace treasuries. The small-scale raiding that a healthy Ugarit could have swatted away was fatal to a sick one.
The Sea Peoples are not a warning about invaders. They are a warning about what happens when a stressor that would have been absorbable in normal times encounters a system that has lost its ability to absorb.
In network-science terms, the Late Bronze Age eastern Mediterranean was robust against single-node failures and fragile against paired ones. A recent computational study of the region's trade and diplomatic connections found that the system could sustain the loss of any one major city without cascading. But certain combinations of two simultaneous failures — the wrong two nodes, at the wrong time — would propagate damage through the whole network. Collapse wasn't lurking behind any single shock. It was lurking behind the specific coincidence of two.
That finding is the analytical centerpiece of this essay. Hold onto it.
How Complex Systems Actually End
Three ideas from resilience theory explain why the Bronze Age pattern is not a historical curiosity but a recurring feature of how complex societies fail.
The first is nonlinear failure. Systems don't degrade smoothly. They absorb shocks, and absorb shocks, and absorb shocks, and then they cross a threshold and reorganize into something simpler very quickly. From inside the system, the moment of crossing is almost invisible — until you're on the other side of it. The Ugarit scribes were still filing grain receipts the week the city burned.
The second is tight coupling. When critical flows can't be substituted — when there is no alternative route for the copper, no alternative source for the tin, no alternative granary for the palace — the system concentrates damage at its chokepoints. A healthy system has slack. A brittle system has optimized the slack away in exchange for efficiency. The Bronze Age had optimized its slack away. Its efficiency was its fragility.
The third is the diminishing return to complexity. The archaeologist Joseph Tainter's formulation is that complex societies keep investing in more complexity to solve problems, and each additional layer of complexity yields a smaller benefit than the one before it. Eventually, the cost of maintaining the system exceeds what the system can produce, and simplification becomes the least-bad option. Collapse, in Tainter's reading, is a rational response to a bad problem — not a failure of will or intelligence, but the choice of people who have run out of affordable alternatives.
Put these three ideas together and you have the shape of Bronze Age collapse. A complex, tightly-coupled, diminishing-returns system encountered a coupled stressor it could not buffer, crossed a threshold nobody saw in time, and reorganized at lower scale. The palaces fell. The trade networks dissolved. Writing was lost. The scribes who had been filing grain receipts stopped being scribes, because nobody was paying them, because the system that paid them was gone.
This is not ancient history. This is a description of how complex systems end.
Two Nodes, Loading
Which brings us, finally, to the present.
I wrote in my last dispatch about the war in Iran and the Duration Paradox — the observation that modern firepower can destroy in days what takes years to replace, and that the difference between a short war and a long one is the difference between a recoverable shock and a structural one. That analysis stands, and I won't reproduce it here. What matters for this essay is a narrower point: the war has already put one major global chokepoint under sustained stress.
The Strait of Hormuz carries roughly 20 million barrels of oil per day — about 20% of global petroleum liquids consumption — and about a fifth of global LNG trade, most of it from Qatar. There is no alternative route that can absorb anything close to that volume. Pipelines bypass a fraction. Strategic reserves buffer a few months. The IMF's assessment of the 2026 disruption describes it as the worst-ever energy-supply shock, and its economic modeling assumes growth downgrades and inflation upgrades even in the scenarios where the war ends soon. If it doesn't end soon, the shock becomes a regime.
Hormuz is one node. It's the obvious one. It's the one everyone is watching.
The second node is less visible and more interesting.
Somewhere in the world right now, a large insurance company is running a pilot program that uses generative AI to process routine claims. The pilot works. The claims get processed faster, with fewer errors, at a fraction of the cost. The company's claims department employs several thousand people, most of them doing work that the pilot has just demonstrated is automatable. Nobody has been laid off yet. Hiring has slowed. Some open positions are being left unfilled. A consulting firm has been brought in to map the department's workflows. The employees can feel what's coming, and are beginning to update their resumes, and are telling their friends at other companies that things feel strange.
This scene is playing out in clerical departments, customer service centers, paralegal pools, bookkeeping firms, and back-office operations across the developed world right now, in 2026. Not as science fiction. As a current phenomenon being documented in real time by the IMF, the OECD, the ILO, and McKinsey, each of which has published estimates that converge on a similar range: somewhere between a quarter and 40% of current jobs have significant exposure to generative AI, with the highest concentrations in clerical and administrative work — which, it should be noted, is one of the largest sources of female employment in advanced economies, and the single largest source of middle-income employment for workers without college degrees.
generative AI exposure (IMF)
by 2030 (McKinsey)
at highest automation risk
The labor-market effects of this are not yet showing up as mass unemployment. They are showing up as hiring freezes, as wage compression, as the quiet disappearance of entry-level positions, as the kind of anxious stability that makes people feel simultaneously secure and doomed. Whether this becomes a full displacement shock by 2030 or a more gradual transition stretched over fifteen years depends on factors that are hard to forecast: the speed of enterprise adoption, the capital costs of deployment, the political response, whether the technology continues to improve at its recent pace or hits a plateau.
But here is the thing that matters for the Bronze Age parallel. Even if we don't know the exact timing, we know the mechanism. And we know that the second node is loading.
An energy shock that would have been absorbable in a healthy labor market becomes much harder to absorb when the labor market is simultaneously reorganizing. A labor displacement that would have been manageable in a low-inflation environment becomes much harder to manage when energy prices are eating household budgets. The fiscal capacity that would normally buffer either shock gets consumed by trying to buffer both.
This is the coupled-failure scenario. Not that either stressor on its own is civilization-ending. Neither is. A sustained energy shock is recoverable. A rapid AI-driven labor transition is manageable with the right policies. The Bronze Age warning isn't about either of these in isolation. It's about what happens when they hit a system at the same time. The network-science finding — that paired node failures cascade in ways single failures don't — is the specific analytical concern.
We don't know whether this is what's going to happen. We know it's the shape of the risk.
What's Different, What Isn't
It's important to be honest about the ways the analogy doesn't hold.
Modern states have capacities the Bronze Age palaces lacked. We can build strategic reserves. We can diversify supply chains, even if we usually don't until after a crisis forces us to. We can surge production, slowly. We can redistribute through fiscal policy. We can deploy monetary tools that didn't exist in 1200 BCE. The Ugarit palace had a grain storehouse and a fleet; we have central banks, social insurance systems, and international coordinating institutions. The comparison is not that our situation is as fragile as theirs. It isn't.
The comparison is that the mechanism of failure is the same, and the warning signs are the same, and the historical pattern is that institutions build redundancy after shocks rather than before them. The Bronze Age polities didn't fail because their leaders ignored every risk. They failed because their leaders optimized for normal conditions until normal conditions ended. Modern institutions have the same tendency. The people running the global system in 2026 are not less intelligent than the scribes at Ugarit. They are doing the same thing the scribes were doing — filing the grain receipts, answering the correspondence, handling this week's problem with this week's tools — because that is what their jobs require, and because the scale of reorganization that would be needed to genuinely buffer against coupled failure is politically impossible until after a coupled failure has happened.
The honest version of the forecast is that we probably won't build the buffers in time. Not because the analysis is wrong, but because it never is.
Complex systems don't fail because leaders ignore every risk. They fail because leaders optimize for normal conditions until normal conditions end.
◆ End of dispatch
The Late Bronze Age collapse was not the end of the world. It was the end of a world. The eastern Mediterranean reorganized, slowly, into something different — smaller polities, simpler economies, different trade patterns, different technologies. Iron replaced bronze, in part because the copper-and-tin trade had collapsed and iron was locally available. New peoples settled where old ones had been. The Phoenicians eventually rebuilt long-distance trade, the Greeks eventually developed the polis, the Israelites eventually wrote down their scriptures. The thing we call Western civilization was born out of the wreckage of what the collapse destroyed.
The people living through it didn't know that. They knew their grain shipments were late and their walls had been breached and their kings couldn't help them.
This is the deeper warning. Not that collapse is apocalyptic — it usually isn't — but that reorganization is brutal for the people inside it, and that the reorganization, when it comes, is almost always more sudden and more sweeping than the incremental signals leading up to it would suggest. The scribes at Ugarit were not stupid. They were doing their jobs. Their jobs just stopped existing in a context they could not have predicted from the data they had.
The question we should be asking in 2026 is not whether the modern global system is going to collapse. It almost certainly isn't, at least not in the way the word implies. The question is whether we can see the second node loading in time to build the buffers that would prevent the first node's shock from cascading into a structural reorganization we don't control.
The Bronze Age answer to that question was no. They couldn't see it. The network-science reconstruction of their system is something we can only do now, with computers and three thousand years of hindsight. They were inside it, and they had to rely on the grain receipts and the diplomatic correspondence and the reports from the coast. When the collapse came, the last thing many of them did was sit down and write a tablet asking for help from someone who was already beyond helping.
The difference is that we have the tools to see it. The IEA publishes the electricity demand curves. The IMF publishes the growth scenarios. The OECD publishes the labor-exposure estimates. The network scientists publish the coupled-failure models. The information is available. Whether we use it is a choice.
That's the warning from Ugarit. It's three thousand years old. It's also the most relevant thing you'll read this week.