Kurzweil Scorecard: The Weapons Got Precise. The Casualties Got Worse.

In 2005, Ray Kurzweil looked at the smart bombs of the first Gulf War and saw a hopeful arc. Weapons were getting more intelligent, and intelligent weapons, he reasoned, hit only what they aim at. The future of war would be precise, remote, and — this is the part worth sitting with — less deadly. He predicted warfare would migrate to two exotic frontiers: cyberweapons and nanobots.

Twenty-one years later, the cyberweapons are real. The nanobots are not. And the central promise — that precision would mean fewer bodies — has been inverted by the cheapest flying robots ever built. In 2025, drones accounted for roughly 80 percent of battlefield casualties in Ukraine, and the year was the third-deadliest for organized violence since the Cold War ended. Kurzweil got the technology of precision right and the consequence of precision exactly backward.

The predictions

Three claims, drawn from The Singularity Is Near (2005), anchor this batch.

The first: warfare would move toward nanobot-based weapons and cyberweapons (ch. “The Impact”), with a stated horizon of the 2030s.

The second, more specific and dated to the late 2030s and 2040s: as nonbiological intelligence predominates, cyberwarfare over information, communication, command, and control will become a primary determinant of military success (ch. “on Warfare: The Remote, Robotic, Robust, Size-Reduced, Virtual-Reality Paradigm”).

The third is not a forecast but a trend claim Kurzweil made about his own moment: that as weapons have become more intelligent, warfare has trended toward more precise missions with fewer casualties, including collateral casualties (same chapter). He doubled down in The Singularity Is Nearer (2024), devoting pages to Steven Pinker’s long-peace thesis. There, citing Pinker, he notes that pre-state societies “averaged a rate of death in warfare of 524 per 100,000 people per year,” while twentieth-century Germany, Japan, and Russia — countries hit hardest by two world wars — saw annual war-death rates of just 144, 27, and 135 per 100,000. The arc of history, in his telling, bends toward less violence, and technology is the bow.

Where we actually are

The cheap-precision turn Kurzweil didn’t name. Start with what is actually killing people on the modern battlefield, because it is neither a nanobot nor a cyberweapon. It is a first-person-view quadcopter that costs a few hundred dollars and carries a shaped charge. In 2025, Ukrainian FPV and bomber drones carried out roughly 820,000 confirmed strikes; in December alone, more than 33,000 Russian personnel were reported hit by them. UN human-rights monitors found that civilian casualties from short-range drones rose 120 percent over the year, naming FPV drones the most dangerous weapon for civilians in the country.

These machines are precise in exactly the way Kurzweil described — a single operator, or increasingly the drone’s own onboard model, steers a warhead onto one specific target. Ukrainian developers now field drones whose onboard AI locks a designated target and flies the terminal attack leg autonomously when the radio link is jammed, using optical-flow tracking to hold the lock through a high-speed dive. In 2024, by the account of a senior Ukrainian defense-industry figure, ten fully autonomous quadcopters engaged a “Terminator mode” and killed without a human in the firing loop — the first autonomous killings of humans by machines acting on their own targeting decisions.

That is the precision Kurzweil foresaw. But its arrival did the opposite of what he expected. When a precise weapon costs as much as a laptop, precision does not ration violence — it democratizes it. The Uppsala Conflict Data Program counted 245,000 battle deaths in 2025, up from 188,000 in 2024, across a record 65 state-based conflicts. The smart weapon didn’t make war surgical and rare. It made killing abundant.

The patent record shows the same civilian camouflage. Search the 9.3-million-document US patent corpus for the explicit category — “autonomous weapon system” — and you find almost nothing: a handful of filings a year, never more than three. But the enabling layer is enormous and grows fast. Drone-swarm patents climbed from a trickle in the late 2010s to twelve grants in 2023 and double digits every year since. Automatic-target-recognition filings run twenty to thirty a year. Cyberattack-detection patents jumped from a couple in 2018 to roughly two dozen annually by 2022 and have held there.

Read the actual inventions and the dual-use blur is total. US 12,608,024, granted in April 2026, describes an autonomous drone swarm system with AI-driven coordination: command drones carrying large-language-model processors, subordinate drones organized in a “Queen-Worker” hierarchy, the whole formation talking over an “encrypted self-healing mesh communication network” using laser, radio, and visual links. The claims read like a specification for the self-organizing secure military network Kurzweil said autonomous warfare would require — but the patent is coded for logistics and inspection, not combat. Its neighbors in the swarm category are patents for seismic oil surveying, seafloor harvesting, insect management, and search-and-rescue. The military future arrived wearing a commercial badge, the same pattern this scorecard found when it looked at battlefield robots.

Cyberweapons: real, but not yet the main event. Kurzweil’s cyber prediction has two parts, and they score differently. That warfare would acquire a permanent cyber dimension — verified, and ahead of his 2030s clock. Three years of the Russia-Ukraine war have produced a near-constant cyber campaign: Russian operations against critical infrastructure, a roughly 70 percent surge in attacks on Ukrainian civilian and defense systems in 2024, and Ukraine’s volunteer “IT Army” claiming the largest distributed denial-of-service attack on record against Russian banks.

But his stronger claim — that cyber would become a primary determinant of military success — is where the early evidence cuts against him, even though his deadline is the 2040s and the verdict must stay open. The pattern so far is that cyber operations are supplementary, not decisive. An analysis in Lawfare of cyber’s combat effectiveness after Ukraine concluded that the lesson is about defensive resilience and operational tempo, not offensive decisiveness, noting Ukraine drove a 95 percent decrease in high-severity incidents from 2022 to 2024. The most concrete cyber-to-kill chain anyone has documented — an alleged 2025 breach that forwarded drone-location data to a Russian assault brigade — is telling precisely because the cyber operation existed to feed a drone strike. Even at its most lethal, the cyberweapon served the quadcopter. The academic literature has been making this point for years: the most-cited recent work in the field, Lennart Maschmeyer’s “subversive trilemma,” argues that cyber operations systematically fall short of the strategic expectations placed on them.

The nanobots never showed. The third leg of Kurzweil’s weapons forecast — nano-based arms — remains entirely theoretical. In The Singularity Is Nearer he sketches the threat vividly: “Nano-based weapons could include tiny drones that deliver poisons to targets without being detected, nanobots that enter the body in water or as an aerosol and tear it apart from within, or systems that selectively target certain groups of people.” None of this exists. Notice, though, the first item on his own list — “tiny drones.” He was circling the right answer and overshot it by several orders of magnitude in miniaturization. The decisive small autonomous weapon turned out to be ten inches across, not ten microns.

The scorecard

What Kurzweil missed (and what he nailed)

The pattern in this batch is unusually clean, and it is a pattern about cost, not capability. Kurzweil’s instinct for what technology would do to war was sound. Weapons did get intelligent. Targeting did get autonomous. Combat did move toward remote, networked, self-organizing systems — there is a granted patent describing an LLM-piloted drone swarm on an encrypted mesh, which is close to a literal rendering of his 2005 sentence. His error was in price, and price changed everything downstream.

He imagined precision as the property of expensive, exquisite systems — the cruise missile, eventually the nanobot — wielded by a few advanced states. Expensive precision rations itself; you fire a million-dollar weapon only when the target is worth a million dollars, and casualty counts fall. But precision did not stay expensive. It collapsed onto a hobbyist airframe and a commodity vision chip, and cheap precision behaves like the opposite of its costly ancestor. It proliferates. It saturates a battlefield. It puts a guided warhead within reach of an infantry squad, a militia, a teenager with a controller. The long-peace thesis Kurzweil endorsed assumed the trend lines of the twentieth century would extend. Instead, the democratization of precision broke them: 2025 set a post-1946 record for the number of state-based conflicts even as each strike became more accurate.

That is the forecasting lesson worth carrying forward. The hardest variable to predict is not whether a capability will exist but what it will cost once it does — and cost, more than capability, decides who gets to use a technology and to what end. Kurzweil read the capability curve almost perfectly. He missed the cost curve, and the cost curve is where the bodies are.

Method note

This scorecard draws on a corpus of 9.3 million US patents, roughly 357 million scientific papers, and battlefield and conflict-trend reporting gathered for this piece. Patent counts come from full-text searches of the grant and application record; the swarm-coordination patent quoted is read from its own claims and abstract. Casualty and conflict figures come from UN human-rights monitoring in Ukraine, Ukrainian and Western defense reporting, and the Uppsala Conflict Data Program as carried in June 2026 coverage. Kurzweil’s predictions are quoted from The Singularity Is Near (2005) and his updates from The Singularity Is Nearer (2024). Verdicts are the author’s own.