If you shut up truth, and bury it underground, it will but grow.
—Émile Zola
🤔💭❓ One of the standard performance stats for cars is their 0-to-60 mph time (or 0-to-100kph here 🇨🇦).
I think there’s something similar with curiosity.
Some people are curious but have a slow 0-to-60. They wonder about things without doing much about it for a while. Others get curious and immediately start looking things up, asking questions, experimenting, learning a new skill, whatever. They get very good at going from ‘uh, I wonder if…’ to a useful 80/20 understanding surprisingly fast. 🏎️💨
It’s something I’ve tried to cultivate in myself. But a lot of your curiosity acceleration also depends on the world you live in. If I had been born in 1382 instead of 1982, satisfying my curiosity about almost anything would have been vastly harder.
And today’s world is incredible for this (if you take advantage).
I look at my oldest boy, and I’m amazed at how fast he moves when something grabs his interest.
When I was learning guitar as a teenager, learning a new song might mean hunting through random tab websites, often finding nothing, waiting until I saw my guitar teacher the next week, or remembering to ask a musician friend the next time I saw him.
There were upsides to that. When you have to work hard for something, you often remember it better.
But I look at how much my boy can explore in an afternoon. He’ll find a piano tutorial on YouTube and learn a new song. Then research fitness or physics or some author he likes with ChatGPT and Google. Then go vibe-code a 16-bit style video game or a simulated petri-dish universe.
His curiosity has a much faster 0-to-60 than mine did at his age.
Just yesterday I was talking to him about the exploration/exploitation trade-off: how much time you spend exploring new things versus going deeper on things you already know you like. There are only so many hours in the day, so you can’t maximize both.
But I realized that the trade-off has changed because exploration is so much faster and cheaper than it used to be.
An hour of exploration buys you more than it used to. You can now get answers, learn the basics of a new field, or start making something yourself in less time than a bus trip to the public library used to take. 🚍
Which makes me wonder if a lot of us are still operating with an old intuition about how much exploration is worth doing. 🤔
🛀💭🪞🚙 Why don’t all car side mirrors have little round convex mirrors?
Seems like that would help with blind spots (even on cars that have blind spot indicators — an additional visual cue seems useful, right?)
I never thought of it before, but now that I looked, you can buy these and add them to your mirrors. I’m considering doing that on my car.
I realize there’s a trade-off because you get a distorted view and your brain has to reconcile two different perspectives, but I’d be curious to try and see if it’s an improvement.
💚 🥃 The paid subscription includes the full feed, every podcast, Zoom Q&As with me, and access to the private Discord clubhouse, where a core group of us hang out and discuss the interesting stuff we find:
🏦 💰 Liberty Capital 💳 💴
🏴☠️ The AI Hardware Heist Economy: When a Truckload of Electronics Is Worth $38 Million
I don’t want to jinx anything or give anyone bad ideas, but I was thinking about how insanely valuable some truckloads of electronics must be these days.
A truck full of gold is very valuable and needs high security. But what is a truck full of RAM or NVIDIA GPUs worth? And there are lots of those moving around the world as part of the supply chain feeding hundreds of billions of dollars of datacenter capex.
I wonder what kind of security is involved with all this. Has organized crime started targeting the industry yet?
I wonder if you can buy memory that fell off a truck somewhere in South Korea on the black market or whatever 🤔
I looked it up, and it turns out: Yes, they’ve been paying attention (maybe they’ve been asking Kimi K3 for tips on how to do heists? 😅).
CargoNet specifically lists RAM, storage drives, and enterprise computing equipment as increasingly attractive targets. And just last week, WIRED reported on two California shipments of AI datacenter gear that were valuable enough to have private security vehicles escorting them. In both cases, attackers apparently rammed the security escorts out of the way, after which the semitruck drivers kept going and disappeared with the loads. Investigators suspect the drivers were in on it. 😬
FreightWaves says known reported thefts of datacenter/high-tech cargo have totaled $110.6 million since June 3 alone, including individual loads worth as much as $38 million.
There’s a twist: cargo theft isn’t necessarily becoming more common, but the losses are becoming much more concentrated in extremely valuable loads. CargoNet says reported theft incidents actually fell 26% year-over-year in Q2, while estimated losses more than doubled, from $135.7 million to $304.6 million.
AI is increasing the value density of certain parts of the economy, and of course, that attracts unwanted attention. I’m not sure everything around these trucks is adapting as quickly as the value of what’s inside them. 🚚 🚚 🚚
What does security look like for a $10m+ truckload of computer parts? Private escorts, GPS trackers hidden in the cargo, tighter checks on drivers and trucking companies, etc.
Where does all this stolen RAM, networking gear, and server hardware end up? Apparently some of these international crime rings have basically built their own parallel supply chains, complete with customers waiting for the stuff.
There must be some pretty weird black-market datacenters out there. 🤔
👨⚖️👨💼📰 Is AI’s New Bottleneck Political Permission? 🤔
In Will we ever have enough compute? in Edition #654, I wrote:
Maybe that’s one thing to watch out for. So far, every time we improve one constraint, another bottleneck shows up somewhere farther out (like whack-a-mole). Chips, memory, networking, power, grids, capital, whatever. As long as the bottleneck keeps moving rather than disappearing, that seems like evidence that we’re still a long way from “enough compute.”
