Hey there,
Welcome to where I write about my journey from a stable Big Tech Software Engineering job to the wild and volatile world of Venture Capital.
If you haven’t already, subscribe to be the first to get my weekly posts:
So I just read this absolutely terrifying report (https://fourthoffset.ai/) about AGI development and national security that has me questioning whether I should be stockpiling canned goods instead of reviewing pitch decks. Like, seriously...
🤖 The AGI "Critical Scale" Problem
Let me break this down: we're rapidly approaching what experts are calling a "critical scale" in AGI development. Think of it like the "critical mass" concept from nuclear physics, but for artificial intelligence. Once a foundation model hits this scale, it could essentially automate its own R&D.
What does that mean in practical terms? We're talking about the equivalent of HUNDREDS OF MILLIONS of scientists and engineers working 24/7. The report literally describes it as "10,000 Manhattan Projects" running simultaneously.
Let that sink in for a moment... I'll wait.
Yeah, I'm freaking out too.
🪖 Military Offsets
Military strategists talk about "offset strategies" - basically technological advantages that give one country an edge over others. We've had three major ones in modern history.
The First Offset: Nuclear Deterrence (1950s) - AKA "We Can't Match Your Army So We'll Just Nuke Everything"
Picture this: it's the 1950s, the Cold War is getting frosty, and the U.S. is looking at the Soviet Union's massive conventional forces like "well... f*ck."
Instead of trying to match soldier-for-soldier (which would have been economically insane), America went with what I'm calling the "big stick energy" approach - doubling down on nuclear weapons. The logic was basically: "Sure, you have more tanks, but we can turn your country into a parking lot, so maybe let's all just chill?"
The Second Offset: Precision Strike & Stealth (1970s-80s) - AKA "We'll Just Be Sneakier Than You"
By the 70s, the Soviets had achieved nuclear parity (awkward), so America needed a new trick. Enter precision-guided munitions, stealth technology, and advanced surveillance systems.
This was the "work smarter not harder" phase. Why send 100 bombers when one stealth aircraft could do the job? Why waste 100 bombs when one smart bomb could hit the target? It was all about efficiency and technological edge.
The Third Offset: AI & Human-Machine Teaming (2010s) - AKA "The Robots Are Our Friends Now"
Fast forward to the 2010s, and competitors like China and Russia have basically copied our homework from the Second Offset. They've got their own precision weapons, stealth tech, and battle networks. The military advantage has eroded, and America is having another "oh crap" moment.
The solution? Artificial intelligence, autonomous systems, human-machine teaming, and swarm technologies. It’s like the military version of “if you can’t beat ‘em with conventional tech, bring in the robots.”
🚀 The "Fourth Offset" (AKA Why I Can't Sleep Tonight)
Strategists postulate that AGI development could become the fourth offset.
What could an AGI at critical scale enable?
Unlimited cyber operations (goodbye internet privacy)
Unstoppable information warfare capabilities
Left-of-launch capabilities to track nuclear subs
And at the extreme end... missile defense systems that could potentially NEGATE NUCLEAR WEAPONS
That last one is particularly terrifying because it could completely upend the concept of mutually assured destruction that's kept us from nuclear war for decades. Whoever gets there first would have "unprecedented national security policy options" - diplomatic speak for "could do whatever the hell they want on the global stage."
“Big Plastic” Hates Them
From water bottles to shrink wrap, Timeplast has a $1.3T market opportunity with its patented plastic that dissolves in water. But the clock is ticking to invest. You have until midnight, July 31 to become a Timeplast shareholder as they expand globally.
This is a paid advertisement for Timeplast’s Regulation CF Offering. Please read the offering circular at invest.timeplast.com.
⚡ The Weird Plot Twist: It's All About Electricity
Here's where my tech brain gets particularly interested. You'd think the bottleneck would be algorithms or talent, right?
Nope. It's power. Literal electrical power.
The report identifies that constructing a 5 GW data center campus is now the principal bottleneck to scaling foundation models. For context, that's roughly the output of 5 nuclear reactors. We're talking MASSIVE energy requirements.
The US actually leads China in algorithmic innovations and advanced chips, but we're hitting infrastructure constraints. The recommendation? Invest in on-site power sources like small modular nuclear reactors, next-gen geothermal plants, or gas turbines with carbon capture.
Not exactly the kind of infrastructure discussion I expected to be having when I left my cozy SWE job, but here we are...
💰 The Insane Price Tag
Creating a foundation model at "critical scale" could cost HUNDREDS OF BILLIONS of dollars. That's with a B. Multiple Bs.
But here's the kicker - US algorithmic innovations could potentially reduce these costs by 4X, which means our algorithms alone are worth hundreds of billions to countries like China. No wonder these models and the algorithms to train them would be prime targets for theft.
It's like the nuclear secrets of the 21st century, except potentially more impactful. And I thought protecting my company's codebase was stressful...
🏭 Our Manufacturing Achilles' Heel
Even if we win the AGI race, we might lose the implementation race. The report states that China can acquire high-end weapons systems and equipment 5-6X faster than us. Their shipbuilding capacity is a mind-boggling 230X larger.
TWO HUNDRED AND THIRTY TIMES. That's not a typo. That's a "holy sh*t" moment.
So we could develop the most advanced AGI in the world, but if we can't turn those innovations into actual systems and capabilities quickly enough, we're still screwed. It's like having the world's best blueprint for a lightsaber but only a 3D printer from 2010 to build it with.
🤔 What Are We Actually Doing About This?
The report recommends establishing a modern equivalent of the National Defense Research Committee to adapt AGI-driven R&D into field-ready applications. It also suggests using public funding, tax credits, and insurance guarantees to de-risk investments in adaptable manufacturing facilities and onshore weapons supply chains.
But here's my concern: organizational inertia. The report warns that even if we get a first-mover advantage in AGI development, our bureaucratic slowness could allow China to catch up and potentially surpass us in implementation. We’ve never had governmental grid lock in the US …
As someone who's worked in both Big Tech and now VC, I've seen firsthand how slow large organizations can be. The thought that our national security could hinge on our ability to move quickly and decisively is... concerning, to say the least.
😱 My Personal Existential Crisis
I'm sitting here wondering if I should be focusing my investments differently. Should I be looking more at infrastructure plays? Energy startups? Companies working on manufacturing automation?
The venture world tends to get excited about consumer apps and SaaS platforms because they scale quickly, have nice margins, and predictable exits. This report has me thinking about the deeper tech that underpins our national security.
Maybe I've been thinking too small. Maybe the real moonshot investments aren't just about making life more convenient or businesses more efficient, but about ensuring we maintain technological superiority in an increasingly competitive global landscape.
Or maybe I'm just having an existential crisis after reading a particularly alarming report. It wouldn't be the first time.
Either way, I think I need another coffee. Or maybe something stronger.
Until next time!
Signing off and signing zero checks,
SWEdonym
P.S. What do you think? Am I overreacting to this whole AGI national security situation? Are you also lying awake at night wondering about the geopolitical implications of foundation model scaling laws? Drop a comment below - misery loves company, especially the existential kind.
Rate this post
P.P.S: I want to make this better for you, add a comment below with your thoughts on what you love, hate, or want more of.