Recently, I came across an in-depth analysis of AI applications in the military field, and some observations are worth pondering.

Simply put, the U.S. military has completely changed its traditional combat model in recent operations targeting a high-level official of a certain country. It’s no longer traditional large-scale bombing, but a "software-defined weapon" composed of the Palantir data platform, Anduril drone systems, and the Claude large model. The logic behind this is quite frightening.

Palantir acts as a "battlefield brain." Its ontological technology can convert chaotic information such as satellite images, communication intercepts, and social media data into real-time battlefield maps. Commanders no longer look at dense reports but at a living digital battlefield. Even more intense, they have deployed frontline engineers directly into the military, reducing system update times from months to just a few hours.

In terms of communications, SpaceX’s Starshield satellite constellation has become critical. When Iran cut off ground networks to create an information vacuum, these 480 encrypted satellites maintained communication through laser inter-satellite links. A portable terminal the size of a two-foot square can transmit petabytes of high-resolution imagery within seconds.

Interestingly, there is a split in AI ethics. Claude, due to safety guardrails, has been marginalized by the Department of Defense, while OpenAI and xAI are prioritized because they are "not bound by political correctness." But ironically, Claude still played a key behind-the-scenes role—processing massive intercepted Persian communications, identifying command chain breaches, and providing intelligence support for strikes. Analysts only need to ask, "If we do this, where is the target most likely to escape to?" and Claude can generate the optimal interception plan based on military theory and real-time intelligence.

Even more terrifying is the system developed by the Israel Defense Forces. Lavender can score millions of people, and Habosolah generates 100 strike targets daily. The scariest part is that humans only spend 20 seconds reviewing—just confirming whether the target is male. There’s also a system called "Dad Where Are You," which tracks when targets return home, because launching attacks at home is easier than at military bases.

On the operational level, drone swarms from Anduril and Shield AI can operate autonomously even when GPS and satellite communications are lost. Drones can seamlessly switch between different AI systems during flight—like updating apps on a phone. Soldiers wear mixed reality headsets developed jointly by Anduril and Meta, allowing them to see enemy skeletal postures, obscured targets, and even view real-time footage from drones.

The capital story behind this is also crucial. A16z-led venture capital has completed a $15 billion funding round, specifically investing in hard-tech companies like Anduril and Shield AI. Their logic is completely different—they’re not building a $100 million F-35, but mass-producing 10,000 autonomous drones costing $10,000 each. Software is prioritized; hardware is just the carrier.

Someone proposed the "Three Clocks" theory to reflect on this victory. The military clock has been pushed to the extreme—sensor-to-shooter time shortened from months to seconds. But the economic and political clocks are far from that speed. The exponential growth in drone consumption puts enormous pressure on supply chains. More critically, algorithms can precisely eliminate a leader, but they cannot change local people's minds or quell regional anger.

This marks the beginning of the software-defined geopolitics era. Warfare has become as easy as clicking a screen—low casualties, high efficiency—but at the same time, it dangerously lowers the political threshold for humans to initiate war. We have entered a battlefield where even human commanders are too overwhelmed to feel fear.