One destroyed data center can derail a nation's AI.
We built systems so efficient they became brittle. When Yandex's supercomputers went dark, we learned the real vulnerability wasn't the redundancy we engineered — it was the concentration of power we allowed. The collective brain we trusted was actually a single point of failure wearing a crowd's face.
Key Insights
1
Crowds grow smarter by diversifying their inputs, but we've consolidated our inputs into fewer hands, which means a single targeted blow now silences entire categories of decision-making that used to spread across thousands of smaller actors.
2
Yandex trained its AI on centralized hardware because centralization optimizes for efficiency and cost, not resilience, which means the system worked perfectly until the moment it catastrophically didn't — and by then, downstream decisions across industries were already locked into its predictions.
3
The next vulnerability isn't technological, it's structural: anyone who understands where the single points are positioned will find them, and the crowd that depended on that concentration will discover it had no backup brain at all.