Deterrence in the 21st Century
There’s a poignant section in Feynman’s memoirs, Surely You’re Joking, Mr. Feynman, where he describes sitting at a cafe in Manhattan after helping build the nuclear bomb, watching families enjoy their afternoon, while he lives with the dread of knowing he helped create something capable of turning the whole city into dust in seconds.
I was a committed pacifist in my early 20s. I viewed defense companies as war-profiteering machinery that maximized global conflict.
That changed with Russia’s invasion of Ukraine in 2022. It made me realize diplomacy alone might not be enough for liberal democracies to stand up to authoritarian regimes. Watching Ukraine arm itself and fight back against a giant who expected a quick crushing victory was eye-opening. It was also the first time we saw drone warfare take center stage in a real conflict.
What sealed it for me was the 2025 India-Pakistan conflict, the first time both nations deployed autonomous systems against each other.
The conclusion I landed on: liberal democracies have a legitimate interest in arming their militaries with superior defense technology. Democratic governments offer the best chance to ensure these tools don’t fall into the wrong hands. I don’t think there’s another way.
As we get closer to field trials of our own systems (SIGINT and counter-drones) for military and intelligence customers this year, I want to write down a few broad insights. These are the things I think give democracies the best shot at staying ahead as we enter a world where drones and autonomous systems increasingly run the show.
This isn’t meant to be moralistic. It’s a living document I’ll keep updating as my beliefs sharpen. It also has a few questions I can’t stop asking myself.
Musings on Autonomous Warfare (or how democracies keep the lights on)
-
Cheap, mass-produced drones and missiles offer asymmetric advantages over traditional aircraft carriers, tanks, and infantry. The cost to produce the former is an order of magnitude lower, while the capability gap keeps shrinking.
-
Scale becomes a superpower. A large inventory of simple autonomous systems can beat a smaller set of complex, expensive ones. Quantity has a quality all its own.
-
Reducing system complexity is the key unlock, and it comes from two moves: cut the number of components, and cut the number of exotic components that require specialized factories to make.1
-
Most of what modern defense systems need already exists in commercial tech. For example: face recognition algorithms in Snapchat, miniaturized electronics, cheap sensors in iPhones. A team of good software engineers can build precise, cheap systems that outperform legacy hardware that cost billions.
-
We will soon have surveillance at a scale and efficiency never seen before. I’m not sure our existing laws and institutions are ready for the edge cases that come with it. We need to educate lawmakers carefully. I hope we can play a small role in helping legislatures and military institutions navigate this well.
-
We must protect civil liberties. These systems cannot be turned inward for domestic mass surveillance. I would consider my startup a failure, even a financially successful one, if our systems made democracies more authoritarian. The formulation I keep coming back to: use autonomous systems for national defense in every way except those that make us more like our adversaries.
-
Wherever possible, pursue open dialogue and traditional diplomacy before offensive action. I’d even advocate for mutual disclosure of advanced systems before any military engagement. Adversaries will assume a system built in secret is a superweapon, which defeats the whole purpose of deterrence.
- Privatizing defense tech is broadly good. Market incentives are a major structural advantage liberal democracies hold for three reasons:
- Legacy defense tech was built by large, bloated organizations with no incentive to move fast or keep costs down. This is largely a product of cost-plus contracting. A startup has a structural edge over prime contractors, if nothing else, by saving taxpayers money.
- Military tech consistently spills into civilian life. The Internet, GPS, rockets. A lot of what we’re building will find commercial applications within a few years.
- Liberal democracies have market-oriented economies. Authoritarian regimes favor central planning. That structural difference naturally produces more defense tech startups on our side.
-
One question I keep returning to: is a private limited company the right structure for a defense tech startup in India? I’m not sure pure profit maximization avoids the negative externalities that come with building defense tech. Something like a Public Benefit Corporation that formally indexes on mission alongside revenue might be a better fit.
- The largest question I’ve been sitting with for years: do we live in a world that structurally favors offense over defense? It is easier to build a weapon of mass destruction than to build a defense against one. If that’s permanently true, we’re doomed. But I think we can build systems that maintain the offense-defense balance. It’s worth being delusionally optimistic here.
Footnotes
-
One underappreciated reason the Allies won World War II: American automaker factories were converted into weapons production lines. The machines building Ford and GM cars started building wartime systems. Industrial scale, already built for commercial use, became a decisive military advantage. ↩