Scuttlebutt

Scuttlebutt

Autonomous weapons have existed for thousands of years, says Anduril's founder

11 arguments: why Anduril will not build humanoids or launch rockets, why 80% of big defense programs went to 5 firms, and why AI does not change the hardest part of a founder's job

Scuttlebutt's avatar
Scuttlebutt
Sep 29, 2026
∙ Paid

Moonshots with Peter Diamandis · Sep 28, 2026 · recorded Sep 25, 2026 · 36m · 25m saved

With Palmer Luckey, Founder of Anduril Industries and Oculus VR

Autonomous weapons have existed for thousands of years, says Anduril's founder

Until a few years ago, 80% of major US defense acquisition programs went to five companies, Palmer Luckey said, which is why investors doubted the government could ever buy from Anduril at scale.

Most defense contractors are paid to develop a system and then build it. Luckey said Anduril spends its own money to build products first and then sells them, and that program managers now give it contracts because it delivers on schedule.

"We use our own money to build products, then we sell them."

Palmer Luckey, founder of Anduril Industries and of Oculus VR, on Moonshots with Peter Diamandis, spoke at the Moonshots Live summit, where Anduril had just won an XPRIZE for autonomous wildfire response. Anduril builds autonomous systems for the US military, from robot submarines to aerial vehicles.

The full interview is covered here so you can skip it. 36 minutes of audio, 11 minutes of reading.

Here are the 11 arguments that matter.


This Scuttlebutt Summary is available in full to premium subscribers. If you're not one already, please consider subscribing to keep reading.


Key Takeaways

  • Autonomous firefighting has been possible for over a decade; politics, not technology, held it back

  • He is not worried about AI doing the killing, but about evil people using moderately competent AI

  • Weapons that make their own kill decisions have existed for decades, from the Aegis system to Vietnam-era missiles

  • Chat-style foundation models are the wrong tool for weapons because they are not deterministic or auditable

  • Defense robots will stay specialized, because a small edge over an enemy is worth paying for

  • Anduril will not build humanoids, small attack drones or rockets, because others already do them well

  • Anduril's first pitch deck promised to save taxpayers hundreds of billions a year

  • AI does not change the hardest part of a founder's job, he said, and will not for about 10 years

1. A Wildfire XPRIZE

A host opened by noting that Anduril had won the wildfire XPRIZE, worth $2.2 million, a million of it from Lockheed Martin. Luckey said Anduril competes with Lockheed's fighter jet division and partners with it on payloads.

Possible for over a decade

One of Anduril's first two products was an autonomous firefighting tank, started in 2017 and finished in 2019.

"It's insane that we let fires become conflagrations before you put them out." — Palmer Luckey

He said people claim the technology has only now made this possible.

"But it's, I mean, this has been possible for over a decade." — Palmer Luckey

What changed, he said, is that the problem has become politically impossible to ignore.

Detect, classify, reject, match

The test required detecting and classifying fires and rejecting false ones, such as a cigarette, a campfire or a road flare, then sending the most cost-effective asset that could arrive in time. Military work follows many of the same steps, he said, but timing matters far more with fire.

"You know, a bomb is a bomb whether or not it takes a day or an hour or a minute to get there." — Palmer Luckey

A prize for talking to animals

A host said Luckey is funding a new $10 million XPRIZE, to be announced in a month. Luckey kept the details secret but said it concerns understanding how animals think and communicating with wild animals 10 times better than today's best, in their own language.

"But suffice to say that I'm not interested in joining the race to push AI forward with my money, I think that the artificial intelligence side of models is pretty well covered by plenty of people." — Palmer Luckey

He said large language models distill humanity's way of thinking, other species could teach a lot, and the work may help build more power-efficient AI.

2. Evil People, Not AI

User's avatar

Continue reading this post for free, courtesy of Scuttlebutt.

Or purchase a paid subscription.
© 2026 The Golden Age LLC · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture