| ◉ WILDCARD | Traction! — Deep Dive | ~20 min |
| ◈ SCIENCE | Malaria Parasite Caught in the Act of Breaking and Entering | ~8 min |
| ◈ SCIENCE | Experience Makes Disruption Last Longer | ~7 min |
| ⬡ AI & PRODUCT | Open Weights Crosses the Frontier | ~12 min |
| ◆ DESIGN | Cloud House | ~3 min |
| ◈ SCIENCE | Ice Age Europeans Were More Connected Than We Thought | ~5 min |
| ◉ WILDCARD | The Negative Value of a Billionaire Tax | TAKEAWAY |
Jennifer Vail has spent a career as a tribologist reducing friction in jet engines and car emissions — and occasionally increasing it to remove dental plaque. Her essay argues that the force that resists motion is the very thing that has propelled every human technological leap: chariot axle grease, Viking hull geometry, the wheel’s static friction enabling forward motion. Without resistance, no grip; without grip, no traction.
The frictionless metaphor has colonised product and software thinking so thoroughly that reading an actual tribologist’s account of friction’s indispensability feels almost corrective. Worth every minute of the full read.
Read on Aeon →Plasmodium falciparum invades red blood cells through a molecular portal it assembles, squeezes through, and dissolves behind itself — a door that vanishes to avoid immune detection. Cryo-EM has now imaged this portal in the act, showing for the first time the exact structure of a mechanism previously inferred only from timing and genetics. Understanding the door’s architecture is a direct route to designing drugs that jam it before it disappears.
Read on Quanta →The dominant story in science sociology — that disruption peaks young — turns out to be incomplete. A PNAS study of 200,000+ scientists finds that experienced researchers produce disruptive work less often, but when they do, it persists longer: later research consolidates around it rather than superseding it. The mechanism seems to be breadth of prior knowledge enabling a kind of disruption that depth-over-time produces and that youth, with its narrower base, typically can’t.
Read on Physics World →Moonshot AI’s Kimi K3.1 Pro — 104 billion active parameters from a 2.8 trillion-parameter mixture-of-experts — benchmarks at 55 on Artificial Analysis this week, ahead of Fable 5.1 and within three points of Claude Opus 5.5 (58). It’s the first time an open-weights model has scored this close to the closed frontier. Open weights release is promised by November 4; a Highspeed API variant runs 2–3× faster at the same price.
swyx’s breakdown covers the architecture and what this does to the competitive landscape. The prediction from two months ago — Fable-class open weights by year end — arrived early.
Read on Latent Space →Architect Peter Morris designed his own home in Gospel Oak, London: marshmallow-pink render, sweeping arches, turquoise detailing, Miami by way of North London. Two offset residences sharing a PostModern exuberance under one roof. Polarising by intent, which is most of the point.
View on Dezeen →DNA extracted from a 35,000-year-old woman in southwest France shows ancestry shared across Europe, Central Asia, and Siberia — a far broader network than expected for hunter-gatherers that far predates the agricultural migrations usually invoked to explain population mixing. Genetic connectivity and physical distance were decoupled much earlier than the standard model assumed.
Read on Nautilus →California’s 2026 Billionaire Tax Act — a one-time 5% levy on net worth above $1B — was projected to raise $100B. A new NBER paper estimates it will collect roughly $30B. Publicly confirmed departures have already eliminated ~30% of the wealth tax base; a litigation-survival discount reduces the rest.
The real number is the net present value: California’s billionaires pay an estimated $3.3–5.8B/year in ordinary state income taxes. The permanent loss of those streams from departures more than offsets the one-time wealth tax revenue, yielding an NPV of roughly negative $38.9B. The paper validates this against the wealth-tax elasticity literature: behavioral responses at the top of the wealth distribution are large, and treating billionaires as immobile is a persistent modeling error.
Read the paper →Elizabeth Oldfield and her husband moved into a communal house expecting to diagnose something broken in modern isolation. What they found was something broken in themselves. Living communally collapses the gap between who you think you are and how others actually experience you — a gap that atomised life lets you maintain indefinitely. A quiet, precise essay about the cost of self-images that never get tested.
The math underneath the current AI chip restrictions debate: unilateral US export controls on controlled goods are 28% offset by third-country substitution. Multilateral controls through export control regimes (Wassenaar and equivalents) reduce that offset to 7%, making them roughly 7× more effective at actually degrading adversary military capability — for similar civilian cost. As the US debates tightening restrictions on advanced AI chips and model weights, the coalition structure matters more than the restriction level.
Willison prompted Claude Opus 5.5 High to “build an artifact where you open a photo and it rapidly paints a version of that photo using a canvas element and simulated paintbrush.” The result uses Sobel edge detection with no external libraries and produces something genuinely satisfying. What he doesn’t say: this used to require a creative coding specialist and a week of iteration. Now it’s a one-sentence spec and two minutes.
The strongest pieces from this week’s editions. You may have read some; these are the ones worth going back for.
Human experts have now unpacked the AI-generated proof of the quasi-Riemann hypothesis and found it sound. “The most significant AI-for-math achievement so far, by quite a margin.” The best account of what this result actually means and what it doesn’t.
A levitated nanoparticle suspended in near-perfect vacuum can now resolve individual noble gas atom collisions — single-atom sensitivity in a force measurement. The kind of levitated quantum sensing result that hits differently when you know the field.
China’s MOFCOM and the US DHS both moved to regulate open-weight AI in the same week — building parallel regulatory frameworks in real time. Nathan Lambert’s breakdown is the clearest read of what the regulatory documents actually say versus what coverage implied.