Text taken from the Telegram channel of darpaandcia
Commentary by Russian military journalist Boris Rozhin ias in italics
[ColonelCassad] Today, the NSA, Palantir, and other US intelligence agencies operate on obsolete hardware. The volume of intercepted data has long since exceeded their processing capacity: systems are swamped, and responses to analytical queries take hours instead of seconds.
The fundamental problem is that standard processors are physically unable to handle simultaneous searches across trillions of randomly scattered records—90% of the time, they simply wait for the required data to arrive from memory. Therefore, installing a similar system at a major international airport and reconstructing a complete picture of any arriving passenger's life in a matter of seconds—who called them, who they interacted with, what transactions they conducted, where their device has been—is currently unachievable in real time.
The web of connections is too vast, the hardware is too slow, and real time is nonexistent. This is precisely the barrier that the AGILE program, launched in 2022 by IARPA, the agency that funds cutting-edge technologies for the US intelligence community, aims to tear down.
The officially stated goal is "predictive analysis of massive volumes of data from disparate sources—not analyzing what has already happened, but predicting events before they happen." The goal is to build and continuously update in real time a gigantic network of connections between people, calls, transactions, devices, and movements—and to spot alarming chains within it before they lead to a real event.
Intel, as part of the AGILE project, is implementing the TIGRE project, developing the hardware core of a future supercomputer. A patent was published in March 2026 documenting the key architecture of this machine. The target date is 2030, with the claimed increase in intelligence processing speed being tens of times greater than the current level.
Guess Moore's law is dead.
At the same time, Elon Musk, NVIDIA, Oracle, and OpenAI are building next-generation data centers—but for a different purpose. Their machines train and run large language models: they can understand—read text, recognize images, and grasp context. The Intel chip inside the future supercomputer will do something different: remember everything and instantly find any connections in the colossal array of global data.
Together, these two systems create an infrastructure that never existed before: one notices a "suspicious chain"—"Here's a network of forty people in six countries with suspicious contacts"—and the other immediately interprets it and generates a recommendation. Predictive intelligence on a planetary scale in real time.
In the future, digital surveillance using advanced neural networks will be comprehensive.Talk of privacy in digital life and private correspondence will be a thing of the past.