A whole bunch of European leaders were in the Oval Office monitoring the situation this week. Zelenskyy was there talking and wearing a suit. Trump called Putin. There seems to be some progress, but also still lots of Russian demands.
The other big news of the week has not hit many mainstream sources. OpenAI claims that GPT-5-Pro has now performed the first novel math proof by an LLM. If valid (and Dr. Bubeck is a researcher at OpenAI), this would mark a new stage in the AI journey - another Level 4 AGI.
VP Vance has a burger at Union Station. RFK and Hegseth do pushups and pullups, and I refuse to make fun of anyone’s pushup form. Newsom makes THE MAPS GREAT AGAIN.
Excel now has a copilot. Tesla releases the Model Y L. Vanguard eschews equities and tells the world to buy 70% bonds. The US and EU have another tariff deal. China tells Trump they won’t invade Taiwan. John Oliver complains about MAHA. X fixes a bug. Cracker Barrel decides to pull a Jaguar and make everyone hate them for their rebrand, even Newsom.
On to the reading!
Timely
I’m an award-winning mathematician. Trump just cut my funding - Terry Tao is one of the foremost geniuses on planet earth, and I’ve long enjoyed listening to some of his public explanations about his research. He’s soft spoken and entirely focused on doing maths. And NSF just cut his funding. The funding cuts at NSF and NIH are some of the worst outcomes of Trump 47 so far. Even if the administration’s desire to curb the cultural power of universities is sound, the current consequences are still dire.
Our Shared Reality Will Self-Destruct in the Next 12 Months - Prescience the way only Ted Gioia can deliver it. I’ve never agreed more or been so frightened by the implications. Ted expands on the behavior patterns we’ll all soon be in when we’re all asking: “Can I tell the difference between a fake AI video and a real video? Can I tell the difference between fake AI music and human music? Can I tell the difference between a fake AI book and a real book by a human author?” What happens next?
Austerity For Fertility - The majority has caught up to the problem of global birth rates now, and Bryan Caplan provides a countercultural solution. You won’t like it, but it’s logically sound and the argument helps us ask better questions about why fertility rates are where they are and what to do about it.
Global Crossing Reborn - Legendary hedge fund manager Kuppy Kupperman is calling it: AI is a bubble and it has to burst. $400 billion is a lot of capex for datacenter buildouts. But when I look at the quarterly earnings of the hyperscaler, there’s still an awful lot of free cash flow. If we cross some unknown threshold, this will keep scaling up. But many are realizing that an incremental step (not a Singularity) is also a possible outcome.
Timeless
Liberalism as the Shining City on the Hill - America has always been about our aspirations and the corrections we need to make when we (inevitably) fail to live up to them. A reminder of the centrality of this idea told through the state of Israel and the Palestine conflict. “Given that US support for every liberal democracy has become unpredictable, the rest of the West must find enough self-esteem to defend liberal values and liberal states while America gets its house in order. For too long, the status quo for “good liberals” in the West has been self-depreciating cynicism about the virtues of the free world. This is a cowardly position that shrinks from the challenge of never making the perfect the enemy of the good.”
Burner Phones 101 - I just love the course description: “Hosted by the Brooklyn Public Library, this Burner Phone 101 workshop introduced participants to phone-related risk modeling, privacy-protective smartphone practices, the full spectrum of burner phone options, and when to leave phones behind entirely.” What lockpicking tutorials were to the 1980s, this is for the 2020s.
Steve Yegge’s Google Platforms Rant - This is an oldie but a goodie. Yegge remains, in my mind, one of the best developer-bloggers ever. This one was a rant intended only internally for the old Google Plus team that was accidentally made public. I’ve been thinking a lot about accessibility in context to LLMs, and Stevey’s comments there still ring true. And I’m not sure Google has learned the platform lessons yet either.
GPT-5 Prompting Guide - Is this timeless? Definitely not. But prompting as a defined and learnable skill is, and this might be the best how-to guide on the subject yet.
Books
What Is Life? by Erwin Schrodinger - Ross Douthat interviewed Noor Siddiqui this week. She’s the founder of IVF screening company Orchid, and Douthat did an admirable job of holding back and letting her speak—exoterically, “safer babies through data”; esoterically, selection among possible persons (embryos) under a single objective function. (It’s amazing, go watch it.) It got me thinking a lot about the nature of genetics and heredity and where else do you turn for expertise on this than a physicist? Schrodinger gives a decomposition of genetics that installs information as a fundamental primitive of biology. Having such a framework reminds us that technical understanding is not the same as ethical adequacy, something Siddiqui conveniently elides. Technical reliability and risk models don’t adjudicate personhood, they only make the stakes legible. One of the tricks of our age is to mistake more bits for a moral argument.
Tweets
Some good ones, so you don’t need to scroll!
The world is amazing. Cheers!