I don’t know how it’s already been another week, but it has. Everything feels weird - like it is all happening right in front of us and yet nothing much is happening at the same time. I guess that’s just the way history feels — like a Philip Larkin poem where ‘something is pushing us to the side of our own lives.’
Stocks tanked, tariffs are the most popular word around the globe, and Balaji says we’re basically giving the world to China. Free trade and the Bretton Woods paradigm that’s survived since World War 2 might finally be unravelling. Gavin Newsom starts a podcast and his very first guest is Charlie Kirk. The Department of Education staff is cut in half, but Tennessee will be fine - the new DoE secretary knows a mayor there real well.
Elsewhere, GMOs come to mosquitoes, David Sinclair scientifically breaks down fasting, and Bryan Johnson - the don’t die guy - finally reveals his endgame. A Blood Worm Moon, which has never before existed in the history of humanity, is cast down on North America, and science funding becomes a fall guy for the more woke humanities.
In the AI world, Dario says AI is writing 90% of all code within 3-6 months. Andrej says writing should all be for LLMs at this point, not humans. Sam says AI can now write creative metafiction well, and it’s hard to disagree.
And then there’s Zvi, who comes out with this epic explainer:
An AI produces a final output [X] via some method [M]. You can analyze [M] using technique [T], to learn what the AI is up to. You could train on that. Never do that.
You train on [X]. Only [X]. Never [M], never [T].Why? Because [T] is how you figure out when the model is misbehaving.
If you train on [T], you are training the AI to obfuscate its thinking, and defeat [T]. You will rapidly lose your ability to know what is going on, in exactly the ways you most need to know what is going on.
Those bits of optimization pressure from [T] are precious. Use them wisely.
We humans no longer have the highest median intelligence or creative output on the planet. What you have, use it wisely.
On to the reading!
Timely
America Is Missing The New Labor Economy - We’ll start with a more in-depth analysis of the new robot economy, how China is winning, and how incredibly far behind we actually are. Loaded with stats about servos and raw materials and genuinely frightening for anyone who thinks “Made in the USA” means it was made in the USA.
The Ozempocalypse Is Nigh - I had my own brief experiment with tirzapetide which you will find out about soon, dear reader. The generic versions that most people get at compounding pharmacies are now being regulated away. What happens next? And what’s the right way to think about these patents?
This Thing Will Fail - A bit of historical pessimism, a brief interlude with the word “thumotic”, and an excellent interpretation of how MAGA is very online and incompatible with the older “Strong Gods” it sometimes purports to serve.
The End of YC - The thing about building being so much easier these days is that anybody can build the dream they have in their head… only to find out their dream was kind of shit. Once we realize that, what comes next?
Timeless
Saving Our Souls - Our soul is meaning. Constructed, such as it is. We do still have agency.
The Hot Girl Economy - This feels like an odd one to put in the timeless section, but it’s a whirl. Everything about this screams Gen Z, right down to the total lack of capital letters (besides acronyms, obvi). A peak into what the world looks like if you’re an RHRG (really hot regular girl) that becomes a gaping question about how all of us confront youth and beauty and power.
How We Think About Safety And Alignment - OpenAI is still doing things, even if GPT-4.5 wasn’t talked about much. A look at their safety doctrine is worth a read for the same reason Sam Altman’s blog is: to read between the lines and keep an eye on things.
How I Use LLMs To Help Me Write Code - Don’t be fooled, this guide is NOT just about writing code with LLMs. It’s really an excellent set of guidelines for doing anything with an LLM. Doing it well is way harder than most people realize currently and starting here is a huge leg up.
Books
Tomorrow 3.0: Transaction Costs and The Sharing Economy by Mike Munger - A dive into the economic idea of transaction costs (triangulation, transfer, and trust), how businesses exploit these opportunities, and how these ideas are changing in a new economy where - as Tom Goodwin stated - “Uber, the world’s largest taxi company, owns no vehicles. Facebook, the world’s most popular media owner, creates no content. Alibaba, the most valuable retailer, has no inventory. And Airbnb, the world’s largest accommodation provider, owns no real estate.”
Tweets
Some good ones, so you don’t need to scroll!
🤣
The world is amazing. Cheers!