NUCLEAR WAR SIMULATOR: HOW IT WORKS

I bought Nuclear War Simulator on a Thursday night and did not sleep that night. That is not a dramatic opener, that is what actually happened. I downloaded it thinking I would poke at it for twenty minutes, figure out what the tool did, and then go watch something stupid. Instead I ended up at 3 AM staring at a rendered globe with about nine hundred simulated warheads in flight, the fallout plume from a surface burst in the Urals drifting east across Kazakhstan, and a casualty counter ticking upward every time I advanced the simulation clock by six hours.

There is something about this piece of software that does not let you close it. NUKEMAP is the famous one, the browser tool your nuclear-curious friends have been sending each other since 2012. Nuclear War Simulator is something else. It is the next order of magnitude. Where NUKEMAP shows you what one bomb does to one city, Nuclear War Simulator shows you what a full exchange does to a planet across several days, and it does it with enough physical detail that the results feel less like a toy and more like a very cold forecast.

This is an explainer on what the thing actually is, who made it, how it differs from the browser tools people already know, and why I think a simulator this dark has a legitimate educational reason to exist.

What it is

Nuclear War Simulator is a 3D desktop application that came out of early access in 2024 after years of development. You can get it on Steam. It runs in Unity and it looks, at first glance, like a strategy game. You have a globe. You have a time control at the bottom of the screen. You can place weapons, place targets, set detonation times, set burst types, and then press play and watch the exchange unfold in real time or at accelerated speed.

The core loop is not shooty gameplay. You are not commanding a side. You are building a scenario. Pick two or more countries. Assign them arsenals, either real ones based on current estimates or custom loadouts you define yourself. Set their targeting priorities. Configure their launch policies. Push the button. The simulator then runs a physics model of what happens next, from the first missile separation through the final fallout deposition hours later, and tells you how many people died, where, and how.

That is the summary. The detail is where this thing earns its reputation.

The casualty model

Most nuclear simulations use what is called a lethal area approximation. You pick a yield, you pick a burst height, you get a circle, and you assume everyone inside the circle dies and everyone outside lives. It is a fast calculation and it is roughly accurate at the high level. Nuclear War Simulator does not do this.

Instead it uses population density grids at one-kilometer resolution. For each detonation it computes blast overpressure, thermal flux, and prompt radiation as continuous fields, then integrates those fields against the population grid to estimate casualties at each location. Somebody standing two kilometers from a 300 kiloton airburst has a different probability of survival than somebody standing four kilometers out, and the model tracks that gradient rather than drawing a hard line.

It also separates casualties into deaths and injuries, which matters because mass casualty injury is what breaks a medical system. You can kill a city with one bomb. You can destroy a country's ability to treat burn victims with two or three. The simulator shows you both numbers and the gap between them is where the long term consequences live.

The fallout model

This is the part that kept me awake. NUKEMAP has a fallout overlay. It is useful but it is static. You get a plume shape based on prevailing winds at your chosen ground zero, and that is that. Nuclear War Simulator runs an actual atmospheric dispersion simulation. It pulls real wind data, including altitude-resolved wind speed and direction, and it tracks fallout particles of different sizes as they settle out over time. The plume does not just point one direction. It shears, it curls, it drops heavy particles early and carries fine particles for hundreds of kilometers.

You can scrub the time slider forward and watch the plume evolve. A surface burst at midnight in one country becomes a radiation event in a neighbor's capital eighteen hours later. The dose contours update live. You can see the fifty percent lethal dose boundary move across borders over the course of a simulated day. I have never seen that dynamic visualized anywhere else outside of classified-looking defense modeling software.

How it differs from NUKEMAP

NUKEMAP is a browser tool. It is free, it loads instantly, and it answers one question: what does this weapon do if I drop it here. It is the right tool for that question and for first contact with the scale of nuclear effects. I love NUKEMAP. I have written about NUKEMAP. It is the best entry point to understanding what these things actually are.

Nuclear War Simulator answers a different set of questions. What if two countries exchange fully? What if the exchange is counterforce versus countervalue? What does the timeline look like hour by hour? Which populations die from fallout versus blast versus thermal? How do command and control decisions propagate across a six hour window? You cannot ask those questions in NUKEMAP because NUKEMAP does not simulate time, or multiple targets, or arsenals, or dispersion dynamics. It is a single-shot visualization.

The other big difference is the 3D globe. NUKEMAP is a 2D Google Maps overlay. Nuclear War Simulator is a rotatable planet with real topography and a day-night cycle and cities you can zoom all the way down to the street level in. The weapons travel. You see the trajectories arc. You see the MIRVs separate. You see reentry. When a warhead detonates the camera can follow it. That sounds lurid and in screenshots it kind of is, but in the act of using the software it reads as solemn rather than exciting. The 3D view is not there to make it cool. It is there because geography matters when you are modeling fallout and the curvature of the earth matters when you are modeling ballistic trajectories.

The developer

Nuclear War Simulator was built by Ivan Stepanov. One person. He is a German developer who started the project as a solo effort and has continued to run it as an essentially one-man studio, though he has collaborated with subject matter experts and taken on limited help over the years.

Stepanov's background is not in nuclear policy. He came to the project the way a lot of good simulators get built, which is that he got interested in a specific problem and spent an unreasonable number of years refusing to stop working on it. He has talked in interviews about being motivated by the Ukraine war and by watching public discourse around nuclear weapons become detached from physical reality. His stated goal with the tool is to let regular people and policymakers see what these weapons actually do, at scale, in a way that is hard to dismiss.

He works with Alex Wellerstein, the creator of NUKEMAP, and with other researchers who help validate the models. The simulator's effects calculations reference the same Glasstone and Dolan data (The Effects of Nuclear Weapons, the Department of Defense reference text) that most serious nuclear effects tools rely on. This is not a guy in a basement making numbers up. It is a guy in a basement citing the standard literature and implementing it correctly.

That matters for credibility. Nuclear simulation is a field where you can get in trouble fast, both politically and accurately. A solo dev choosing to do this seriously, with real references and real domain experts in the loop, is the only reason the tool has any standing at all.

The creepy educational value

I want to say something about the ethics of this software because it is the question I get every time I talk about it. Is it okay to play a game about nuclear war. Is this glorifying the unthinkable. Is this software that should exist.

My honest answer is yes, and I think the question itself reveals the problem the software is trying to solve. We do not talk about nuclear weapons. We have built an entire civilization on top of a pile of city-destroying devices and we handle the pile the way a household handles a drawer of tangled cables, by mostly pretending it is not there. The Cold War ended, the tension eased, and an entire generation has grown up without the school drills and the public awareness that made these weapons feel real to their parents. And then they did not go away. The arsenals are still there. The policies are still there. The launch authority is still held by a small number of aging politicians whose decision-making processes are largely opaque.

Nuclear War Simulator does not make nuclear weapons fun. I have used it for probably fifteen hours now across various scenarios and at no point have I felt entertained. I have felt queasy, I have felt sober, I have felt confronted with numbers I did not want to see. That is the point. That is the same reason NUKEMAP went viral and the same reason NUKEMAP is still a useful teaching tool. Making the abstract concrete is the single hardest and most important job in public education about existential risk.

Interactive simulation does this better than any other medium. A documentary about nuclear war is a story somebody else tells you. A simulator is a situation you walked into yourself. You chose the targets. You set the launch time. You watched the plume roll over a city you recognize. You get to feel your own hand on the thing in a way that watching Threads or The Day After cannot replicate. That shift from passive to active changes how the information lodges in your brain.

There is a version of this argument that says simulating mass casualty events is inherently grotesque and should not be done. I respect that position and I disagree with it. The alternative to simulating nuclear war is not a world where nuclear war cannot happen. It is a world where nuclear war cannot happen in anyone's imagination, where the public cannot picture the thing, where the decisions get made by a small group of people who can picture it and communicate to a public who cannot. That is worse. An informed citizenry is harder to scare and harder to manipulate, and the only way to inform a citizenry about an unprecedented event is to give them tools to visualize it.

What I did with it

For the record, here is what I actually did with fifteen hours of this software. I modeled a hypothetical India Pakistan exchange based on current public arsenal estimates. I modeled a limited US Russia counterforce strike. I modeled a single warhead accidental launch on a major European city, just to see what the knock-on fallout pattern would look like. I modeled a tactical battlefield use scenario in eastern Europe. In every case the numbers came out larger and more spatially spread than I had expected going in. That is the consistent experience people report with the tool, and it is the same experience I had with NUKEMAP a decade ago at a smaller scale. Your priors about these weapons are almost certainly wrong in the direction of underestimating them.

I also did the thing you are not supposed to talk about, which is I ran scenarios involving my own city. I grew up in a place that is not a strategic target by any reasonable definition and I still wanted to see what a near miss on a nearby military installation would do to it. The answer was not what I expected. The blast effects were survivable at that range. The fallout was not. Seeing that distinction on the map, with my hometown's actual road network rendered underneath the dose contours, did something to my understanding of nuclear risk that ten years of reading books about it had not done.

That is what this software is for. It is not a game in the entertainment sense. It is an instrument. A well-built, scientifically grounded, unsettling instrument for making a piece of physics legible to people who do not have the background to read the technical literature. If you have the stomach for it, I think everyone who votes should spend an hour with it. If you do not have the stomach for it, that is also a completely valid response, and I would argue it is the response the software is trying to produce.