Thinking On Paper

Pete Hegseth signed a memorandum to every senior official at the Pentagon directing the Department of War to become an AI-first warfighting force. Six weeks later America was at war with Iran and AI was identifying targets. The memo outlines seven pace-setting projects — Swarm Forge, Agent Network, Ender's Foundry, and four more — and contains one sentence that defines the entire document: "We must accept that the risks of not moving fast enough outweigh the risks of imperfect alignment." Mark & Jeremy get into what every section actually means, why Anthropic was designated a national security supply chain risk for refusing to remove two guardrails, and the question the memo never answers: should a machine ever be allowed to pull the trigger?
--

🎧 Listen to every podcast⁠
📺 Follow us on ⁠Instagram⁠
🏠 Follow us on ⁠X⁠
🏠 Follow Jeremy on ⁠LinkedIn⁠

To suggest guests or sponsor the show, please email: [email protected]
--
Chapters
(00:00) Artificial Intelligence Strategy for the Department of War
(00:58) Executive Order 14179: America's AI Military Dominance
(01:59) China And AI Arms Race
(04:36) Anthropic & Eliminating Bureaucratic Barriers
(07:20) The 7 Pace Setting Projects (PSPs) In The Memo
(08:28) 100% LLM Kill Chain Capability
(10:22) Palmer Luckey
(11:53) Intelligence & The AI Open Arsenal
(13:57) The War Time Approach To Blockers
(16:46) AI Talent Acquisition At The DOW
(18:54) We must accept that the risks of not moving fast enough outweigh the risks of imperfect alignment

Iran posted an AI-generated LEGO propaganda video mocking Trump and Netanyahu. The White House fired back with Grand Theft Auto, Call of Duty and a Wii Sports video of Iranian military sites being destroyed.

A senior White House official told Politico they were "just grinding away on banger memes, dude." 

A hundred million views later…

- WATCH ON YOUTUBE -

This is the AI slop propaganda war, playing out in public - mostly on Twitter -  as the bombs drop, the drones fly and the smoke and mirrors of a confused story evolve.

Operation Epic Fury has killed hundreds, triggered one of the largest oil and energy shocks in history, and will reshape the Middle East - and global politics - for decades to come. This is the meme war that accompanies it.

We react to all of it. The Iran LEGO propaganda video. The White House GTA Iran meme. The deleted Call of Duty airstrike video. The Wii bowling Iran war clip. 

Why did Iran use LEGO? Is this propaganda, or kids playing video game make believe? What does it mean when governments reach for children's toys and video game aesthetics to sell a real war to a generation raised on screens?

The answer is uncomfortable. It is a desensitisation of death. It is a military hiring video dressed as a meme.

This is part of our on-going AI and the War Machine Season.

Please enjoy the show.
--

🎧 Listen to every podcast⁠
📺 Follow us on ⁠Instagram⁠
🏠 Follow us on ⁠X⁠
🏠 Follow Jeremy on ⁠LinkedIn⁠

To suggest guests or sponsor the show, please email: [email protected]

Timestamps

(00:00) What Is Propaganda?
(00:36) Iran Lego Propaganda Video
(02:45) Reaction
(06:55) Whitehouse GTA Iran War Video
(09:07) Epic Fury - US Wii Sports Video
(13:22) Call Of Duty Iran War Video

Space To Grow, By Matthew Wienzierl and Brendan Rosseau. In 2009 an active American satellite collided with a defunct Russian one at 22,300 miles per hour. The resulting debris field created over 150,000 pieces of space junk that won't decay for a century. Nobody paid for it. Nobody cleaned it up... because nobody had to.

That is the tragedy of the commons in orbit, and it is getting worse. 

Conjunctions, the close-passing events that require satellites to burn fuel to avoid collision, grew from 1.7 million in 2020 to 4 million in 2022. Elon Musk has applied for a licence for a million objects. 

This episode covers: 

Kessler syndrome: the tipping point where collisions become unstoppable, and how close we are
Why the insurance market is not pricing orbital collision risk
Astroscale: the company using magnetic docking to clean up space junk, and the Siberian rocket explosion that nearly ended them
Why the economic solution to the tragedy of the commons breaks down completely in orbit
NASA Administrator Jim Bridenstine on SpaceX: "there is only one thing worse than a government monopoly"
Whether a mission-driven monopoly plays by different rules
-
⁠Listen to every podcast⁠

Follow us on ⁠Instagram⁠

Follow us on ⁠X⁠

Follow Mark on ⁠LinkedIn⁠

Follow Jeremy on ⁠LinkedIn⁠

Read our ⁠Substack⁠

Email: [email protected]

--

Timestamps

(00:00) How 150,000 pieces of space junk ended up in orbit and why nobody cleaned them up
(06:21) Kessler syndrome explained: the tipping point where collisions become unstoppable
(10:57) Why the insurance market is not pricing orbital collision risk
(13:50) Government intervention, the Moon Treaty and the five-year deorbit rule
(20:26) Active debris removal: magnets, robots and who is building the solutions
(22:37) Astroscale: how one company is trying to clean up space junk commercially
(24:53) Who pays to clean up orbit when the market has no incentive to
(26:26) Is SpaceX a monopoly and does that matter for the space industry
(29:08) NASA Administrator: there is only one thing worse than a government monopoly
(33:04) Space governance, coordination and whether the tragedy of the commons can be solved in orbit

Spin qubits could scale quantum computing using the same semiconductor fabrication lines that print 50 billion transistors on an Nvidia chip. No new manufacturing paradigm required.

Brandon Severin, Oxford PhD and founder of Conductor Quantum, joins Mark and Jeremy to explain why that matters. 

You need hundreds of reliable qubits for meaningful quantum computation. The industry has dozens. Spin qubits, built from modified transistors, controlled by classical voltages, no lasers, no vacuum, may be the most practical path to millions.

This episode covers:

Why qubit fidelity and coherence time determine what a quantum computer can actually do
How AI automates the calibration problem that makes human-controlled quantum scaling impossible - "you can't have a billion Brandons"
Why trapped ions vs spin qubits is the wrong debate
What Google's quantum algorithm result actually proved, and why it matters
Why the physicists who understand semiconductor manufacturing may unlock use cases pure quantum researchers never reach
The two camps dividing the quantum industry: build one qubit at a time, or build for a million

Also: quantum startup culture vs the AI boom, Brandon's Y Combinator experience, and why scaling quantum looks more like building a rocket ship than climbing a ladder.

--
Brandon Severin: https://www.conductorquantum.com/

--
⁠⁠Listen to every podcast⁠⁠

Follow us on⁠ ⁠Instagram⁠⁠

Follow us on⁠ ⁠X⁠⁠

Follow Mark on⁠ ⁠LinkedIn⁠⁠

Follow Jeremy on⁠ ⁠LinkedIn⁠⁠

Read our⁠ ⁠Substack⁠⁠

Email: ⁠[email protected]⁠

--

Timestamps

(00:00) Introduction: spin qubits and the quantum scaling problem
(03:47) Trapped ions vs spin qubits: fidelity, coherence, and tradeoffs
(06:14) What qubit fidelity means and why it determines scaling limits
(08:25) What is a spin qubit? Building from the transistor up
(11:06) Semiconductor fabrication as quantum computing's manufacturing advantage
(15:00) The quantum circus: superposition, measurement, Schrödinger's cat
(17:17) Shuttling qubits — moving electrons across a chip
(20:33) How AI automates quantum calibration (the control problem)
(25:00) Quantum scaling vs AI scaling: the GPU parallel
(29:08) Quantum startup culture and the AI generation gap
(32:59) Building for a million qubits — rocket ships vs ladders
(36:52) Why quantum is taking so long: talent, concentration, and meaning
(39:43) What seems impossible now that will be routine in 20 years

Quantum computing has been five years away for thirty-five years. Infleqtion CTO Pranav Gokhale makes the case that 2028 is different, because for the first time, logical qubits actually exist.

He explains what a logical qubit is using a Wi-Fi analogy: noisy physical qubits are like corrupted data packets; logical qubits are the clean, usable signal that comes out the other end. Humanity had zero logical qubits before 2023. Infleqtion now has twelve, with a public roadmap to thirty in 2026 and one hundred in 2028.

At one hundred logical qubits, material science and drug discovery become the first real quantum use cases - and the timeline stops being a prediction and starts being an engineering problem.

--
Other ways to connect with us:
⁠Listen to every podcast⁠

Follow us on ⁠Instagram⁠

Follow us on ⁠X⁠

Follow Mark on ⁠LinkedIn⁠

Follow Jeremy on ⁠LinkedIn⁠

Read our ⁠Substack⁠

Email: [email protected]