What I am Reading
Are You Hunting Antelope or Field Mice?
The image comes from a story about lions. A lion can spend all day catching field mice and still starve, because the energy it burns chasing them is more than the energy it gets back. It has to hunt antelope, which are harder, riskier, and rarer, but actually worth the effort. I have watched smart people fill entire careers with field mice. The email, the meeting, the small deal that feels productive because it is finishable. This piece is short and it will make you uncomfortable in a useful way.
How to Have Enough: Wendell Berry on Creativity and Love
Maria Popova on the farmer poet Wendell Berry and the Sabbath poems he wrote walking his Kentucky land on Sunday mornings for decades. The subject is enough, which is a strange word to put next to creativity and love, and that is exactly why it works. Most of the ambition I see, including my own, is built on the assumption that more is the natural direction of a life. Berry is one of the few writers who makes the opposite case without sounding like he is giving up on anything. I read it twice in a week, which almost never happens.
Rough Night of Sleep? Try the 20 Minute Move Doctors Use to Restore Their Energy.
A bad night of sleep hits the hippocampus first, which is why your memory and spatial sense go before anything else does. The fix the sleep doctors here recommend is not more coffee. It is twenty minutes of moderate exercise, ideally in the morning, which a small study found restored cognitive performance in sleep deprived subjects better than rest did. A short nap helps if you can get one. Coffee cut off around noon so you do not wreck the next night too. Practical, evidence backed, and the kind of thing I wish I had known in my twenties.
When's the best time to drink coffee?
The internet has decided you should wait ninety minutes after waking before your first cup so caffeine does not collide with your morning cortisol. The researchers here are less convinced. Caffeine peaks in the blood thirty to sixty minutes after you drink it, so the honest answer is to drink it an hour before whatever you need to be sharp for. Dose matters as much as timing, and for habitual drinkers the morning cup is mostly reversing overnight withdrawal rather than adding anything. The one rule everyone agrees on is the boring one: not too close to bed.
Is AI Reasoning Right for the Wrong Reasons?
When a model produces a correct answer to a hard problem, did it reason its way there or did it arrive by a route that only looks like reasoning from the outside? The intuition that these systems think is stronger than it has ever been. The science of what is actually happening inside them is far from settled. I find this genuinely unresolved, and I am suspicious of anyone on either side who claims it is not.
Why Do Humanoid Robots Still Struggle With the Small Stuff?
A decade of enormous progress in humanoid robots, and the things that still trip them up are small. Uneven ground, a slippery handle, an object that is not where the training data said it would be. The argument here is that getting to widespread use may require going back to fundamentals that the demo era skipped over. Physical AI is one of the areas Neman Ventures spends the most time on, and the gap between what a robot does in a video and what it does on the fifth day in a real warehouse is where most of the investment risk lives.
What I Am Listening to:
Are We Thinking Correctly About AI Intelligence? | The Joy of Why
Steven Strogatz and Janna Levin sit down with Melanie Mitchell, the Santa Fe Institute computer scientist who has spent years arguing that AI does not think or reason the way humans do and that our benchmarks are measuring the wrong things. She is not a skeptic in the dismissive sense. She is precise about what the systems can do and pointed about where the language we use to describe them is doing work the evidence has not earned. After the Navier-Stokes announcement, this is the conversation to have in your head.
Stealing Reasoning Traces from Proprietary LLM APIs | Machine Learning Street Talk
Tim Scarfe talks with Ilia Shumailov, formerly of Google DeepMind, and Alexander Panfilov of the Max Planck Institute about a security hole that sounds almost too simple. Model providers hand back encrypted reasoning state so conversations can be resumed. Those blobs can be replayed across users and sibling models, and a smaller model can ask the provider to decrypt them and then repeat the hidden reasoning in plain text. The consequences run from leaked private data to a reusable jailbreak to poisoned agent traces. If you have any exposure to AI security or defense applications, this one matters.
Gundry MD Ultimate H2: Hydrogen Water for Gut Health, Energy, and Longevity
Steven Gundry, the former cardiothoracic surgeon behind the Plant Paradox, makes the case that molecular hydrogen dissolved in water reactivates one of the body's own antioxidant defense systems and supports weight management along the way. I listened because I keep hearing about hydrogen water from people I respect and I wanted to understand the claimed mechanism. Gundry is selling the tablets he is describing, so calibrate accordingly. The mechanism argument was more interesting than I expected and the evidence base is thinner than the marketing suggests, which is roughly where I land on most of this category.
What I Am Watching:
How Obsession Built Christopher Nolan
A long look at Nolan's life, working methods, and philosophy, built around Tom Shone's book The Nolan Variations and organized around a single claim: obsession is the engine of the whole career. What I found useful was how unromantic it is. The obsession is not a mood. It is a set of practices, most of them tedious, that he has repeated for thirty years without much regard for whether anyone else was watching. Founders will recognize a lot of it.
Physics has misunderstood time
Most physicists will tell you the flow of time is an illusion, a trick your brain plays on the block universe. Avshalom Elitzur, the quantum physicist behind the Elitzur Vaidman bomb tester, thinks that orthodoxy is wrong and that time's passage may be the most fundamental ingredient in the universe rather than an artifact of how we experience it. He makes the case through research into a sub quantum realm where becoming is basic. If he is right, physics needs redesigning. Even if he is wrong, it is one of the more bracing twenty minutes I have spent this year.
How OpenAI Broke Mathematics In Just 88 Hours | Navier Stokes Problem Explained
OpenAI did what the greatest human minds couldn't for over a century: it cracked the Navier-Stokes problem, one of math's seven million-dollar Millennium Prize problems. But instead of a clean victory, it set off a bitter fight over who really deserves the credit.
