aanet

> After retiring from the Baltimore Ravens, John Urschel said he looked forward to being known as "that mathematician who used to play football."

> A paper called “On the Growth Factor of Random Matrices” would not normally attract headlines. But when it was posted on social media this week by another researcher in the field, online observers took note for an unusual reason: The author, John Urschel, had a previous life as a professional football player for the Baltimore Ravens.

> Urschel, 35, is now a tenure-track professor at the Massachusetts Institute of Technology. He quit football in 2017, partly over concerns about sustaining the kind of traumatic, degenerative brain injuries that have scarred a generation of players in the National Football League. He went on to finish his PhD in mathematics, wrote a memoir and now says he’s enjoying life as a regular mathematician in the academic sanctuary of Cambridge, Massachusetts, far from the spotlight.

His paper: https://arxiv.org/abs/2610.06785

> Including Microsoft, Cap Gemini, Tata, Infosys, Wipro, HCL, ...

Apropos of 2026 Nobel Prize in Physics...

> A vast telescope buried beneath Antarctica has captured high-energy neutrinos from the Galactic Centre, ushering in a new era for observing the Universe.

> In June, researchers released what might be the most unusual portrait of the Milky Way yet. For the first time, astronomers have charted our Galaxy using something other than light rays — a finding that heralds a new way of studying what goes on in the busy centre of the Milky Way.

> The data came from the IceCube observatory at the South Pole, which detects neutrinos, the lightest, most pervasive of the Universe’s elementary particles. The observatory is a vast array of 5,160 sensors, buried more than 1.5 kilometres below the surface of the ice sheet. Over the course of ten years, the array detected neutrinos one at a time and measured the direction they came from, which allowed researchers to chart their origin to the centre of the Milky Way.

> The first neutrino map of our Galaxy is a milestone for the nascent field of neutrino astronomy, say researchers. And it’s just the start. Construction is under way on several neutrino-detecting arrays, at sites from the Mediterranean Sea to Siberia.

If you loved Thinking In Systems: A Primer, you may want to read Business Dynamics by John Sterman [1], one of the proponents and pioneers (after Jay Forrester) of systems dynamics. It's a textbook, but even so it is readable and engaging.

Systems thinking changes the way people think about solutions, often for the better (but not always). There's a tendency to think: "uh... it's the system, nothing's gonna change unless we change the _system_ per se". And that's correct. But individual actions also change systemic outputs. So. Case in point: climate change and all that...

[fun fact, I took a class with Sterman at MIT. He's fantastic, very engaging!]

[1] https://mitsloan.mit.edu/faculty/directory/john-d-sterman

Last night, I dreamt I went to Manderley again...

I first read _Rebecca_ as a teen, years ago. Still recall the thrill of the book. Few writers wrote so engagingly as Du Maurier. Hitchcock's movie was fine, but the book is the real deal.

Archive: https://archive.is/20260927163721/https://fortune.com/2026/0...

It me. 100% me.

I, too, like going to (some) parties, but more often I don't. I like hosting, but only small groups of people (3-6ish), and I cannot handle more than dozen at my place, though I don't mind going to parties where there are more. It's just that I end up talking with the handful of people I know.

Planning parties is, well, exhausting. Hosting them is exhausting. I'd much rather talk, chat with, go to concerts with, workout with, hike with people, than go to parties or host them.

When it comes to parties, I want to be invited, but don't want to go.

If there's one club I belong to, it's probably the Groucho Marx club. "I woudn't want to belong to any club that would have me as a member."

¯\_(ツ)_/¯

This is brilliant

Also: >> "Some Anthropic veterans are reportedly buying remote land in case "AI goes awry" Here: https://the-decoder.com/some-anthropic-veterans-are-reported...

This is where the sea-steading guys meet the EA/"AI Safety" cult

This makes me unnecessarily happy and giddy.

I love Fourier series/transforms/DFTs irrationally.

This just makes my heart leap with joy to see creative output with it.

More, from NYT: https://www.nytimes.com/2026/09/22/world/americas/praxis-uru...

Archive: https://archive.is/20260924060037/https://www.nytimes.com/20...

> Mr. Brown’s plan is part of a broader drive among Silicon Valley entrepreneurs to build tech-friendly havens run like for-profit corporations. The movement is inspired by the concept of a “network state,” first popularized by Balaji Srinivasan, the former chief technology officer of the crypto exchange Coinbase.

> In Mr. Srinivasan’s vision, online communities bound by shared values eventually crowd-fund physical territory, form mini nations with their own laws and secure diplomatic recognition from existing states.

> Critics describe attempts to build such a state as elitist efforts by tech moguls to secede from society or unfeasible pipe dreams. But supporters — including high-profile venture capitalists like Marc Andreessen and Mr. Thiel — say they could be essential laboratories for governance that could help cut through bureaucratic red tape to accelerate technological innovation.

> Praxis, a technology-focused community backed by Apollo Projects, a fund associated with Sam Altman, has selected Uruguay for its physical development. The project involves a projected investment of $1 billion in the Colonia development near Colonia del Sacramento. The first residents are expected in mid 2027 while construction of the larger community continues over the following years. Plans include private residences, a resort and spaces for art, music and other cultural activities. Praxis presents the project as an experiment in how communities could live and organize in a future shaped by AI.

It's the same home-steading guys...

My eyes!!!

> Despite some unfortunate and highly publicised accidents, data suggests the best AVs are already a lot safer than human drivers. Nearly 1.2mn people die globally every year in road crashes; human error is a contributing factor in the vast majority. An independent analysis found driverless vehicles from Alphabet’s Waymo in four US cities were involved in 68 per cent fewer crashes per vehicle mile than human drivers — though this is human drivers generally, not just professional ones. Driverless vehicles can struggle with “edge cases” ranging from snow to traffic light failures. But they may become better at handling such situations over time as the technology evolves. Rather than relying on their training on high-definition maps, the latest “vision-language-action” models, developed by companies including Wayve, use cameras paired with language-based AI to improve their ability to interpret unfamiliar situations.

> For now, robotaxis are costlier to operate due to development costs and extra kit. But Goldman Sachs says operational and scale efficiencies could cut costs sharply, and projects a $415bn global robotaxi market by 2035. As costs fall, driverless vehicles could start to displace driver-operated taxis and, in big cities, even private car ownership (the RAC found the average personal car in England is parked for 96 per cent of the time). AV evangelists suggest that could open the way, longer term, to their vision of “shared mobility”, easing congestion and cutting emissions — provided cheap AV rides do not encourage more journeys and lots of unnecessary miles by loitering vehicles.

Goldman Sachs Report: https://www.goldmansachs.com/insights/articles/robotaxis-to-...

> Losses after the Sept. 11 terror attacks put HRT’s Jason Carroll on a new path

Archive: https://archive.is/20260910112856/https://www.bloomberg.com/...

Fantastic project. Love the bells and whistles. I could spend all day here.

Would love to read a bit on how this was implemented (your stack, techniques, etc).

Thanks!

> And there is essentially no discussion of music theory in other cultures (middle eastern, african, east asian, indian)...

To me, that above ^ looks like the biggest gap.

To be fair, the book doesn't claim to demonstrate theory of the "world music" (sic, a phrase as bad as it sounds).

My (brief and shallow) understanding is that most scholars of music in the western world just do not have enough knowledge to talk intelligently about non-western genres. Exceptions exist, of course, but they are quite rare.

[Other reason is that there is a lot of gate-keeping. I've seen that in Hindustani/Carnatic classical studies, FWIW, which is one my areas of interest]

^^ THIS

Hindustani (and Carnatic) music indeed do not denote pitches for a very sound (pun intended) reason: They let singers choose the root note (Sa), and thereupon the melody is built. Most (all?) of Hindustani music is melody-based, not harmonic based (like Western music). And the melodies are themselves created within the Raga framework. Raga is like a mode (but not exactly that), and is hard to explain within the constraints of Western music theory precisely because that concept does not exist in western music per se. Which is why most Indian classical theorists call a raga a "melodic framework", and leave it at that (or expound on it, as the case may be).

A great reference for Indian classical music is [1] and and more recent one [2]

[1] The Ragas of North Indian Music, N.A. Jairazbhoy, https://archive.org/details/dli.ministry.26725

[2] Rāgs Around the Clock, A Handbook for North Indian Classical Music, with Online Recordings in the Khayāl Style, https://www.openbookpublishers.com/books/10.11647/obp.0313#r...

There are several music samples in there...

Example: https://musictheory.pugetsound.edu/mt21c/DiatonicChordsInMaj...

Book: Music Theory for the 21st-Century Classroom

by Robert Hutchinson

Let’s not forget the new new wunderkind who was hailed as the new financial genius and whose AI fund got trash sold to Citadel for 20% of its value (after rising 450% of course)

> The Telluride Film Festival turned its mystery Sunday slot into a phone-free event, unveiling Nathan Fielder’s surprise Elizabeth Holmes documentary “You Can See Everything.” Filmed largely in secret over three years, the nearly three-hour doc is an inside view of Holmes’ final weeks before going to prison, then mutates into reenactments and mind-bending twists that meld reality and performance.

> Dan Ariely, professor of business administration in the Fuqua School of Business, is once again under scrutiny after an article released Monday found evidence that he likely tampered with or fabricated data in an influential 2002 research paper on procrastination.

> The findings — published in Data Colada, a blog run by three esteemed behavioral science professors that investigates and replicates academic research — identified four key “red flags” in the study’s data, which suggested the results were not possible without interference.

> “We are unable to generate a benign explanation for all of the anomalies presented here,” the professors wrote. “Based on this evidence, we believe the data for Study 2 of Ariely and Wertenbroch (2002) were severely tampered with or fabricated to produce the desired results.”

"Three California college students had to be rescued from Mount Shasta on Monday during a poorly planned summit climb in which they overshot their itinerary, meandered off-route and ran out of food and water."

and

"Then in August, while hiking in Shasta's popular Bunny Flat area, a man stopped rangers to ask for directions. He showed them a print-out map he'd been following and explained that he'd prompted ChatGPT to produce it for him.

"It was terrible — terrible — not even close to accurate," Meyers said. "It had route lines and place names that were totally off — just ridiculous. We were like, dude, you gotta throw that away." "

Sorry I LOL'd. Too funny.

More serious though:

"On Saturday, the trio of climbers, all from California, left from the Clear Creek Trailhead on Shasta's southeastern side with a plan to camp partway up before making a summit push early Sunday morning, according to a forest service report. They'd asked ChatGPT for specific guidance, researched the climb on YouTube, and planned to use the AllTrails phone app to navigate the giant peak, the report said.

Clear Creek is a popular late-summer route because, when it melts out, it offers climbers a non-technical way up the peak — it can be done in hiking boots without ice axes or crampons. However, it's an 8-mile journey that gains 6,500 feet in elevation, and guides caution that it is not to be taken lightly.

The climbers reached the summit Sunday evening — far later than the mid-afternoon turnaround time recommended by Shasta climbing experts. They'd clumsily navigated the mountain's rugged volcanic terrain by following the line of a digital map in the AllTrails phone app, misreading it as the way of an established trail."

On Krebs: https://krebsonsecurity.com/2026/09/fbi-probes-service-selli...

I hope the German workers (can) negotiate to keep the 35 hour/week arrangement.

(I'm not German, nor do I work there).

and I hope somehow we are able to bring a less than 40 hour/week arrangement in the US.

Workers in the US work, by far, the most of any other country. And yet, only a relative few are able to afford quality medical care, and decent quality of life. (The workers in the tech sectors like Bay area, NYC, etc. are, IMHO, an outlier with respect to pay packages compared to the rest of the country.)

It's a devil's bargain: you either work your a* off, 60-80 hours/week, and hope to get paid $$$$ one day; or you work "reasonable" hours and get paid for a shitty qualify of life.

From the abstract:

"Here, we introduce the Concept-Wrapper Network (CW-Net), a method for faithfully explaining the behaviour of machine-learning-based planners that causally grounds their reasoning in human- interpretable concepts without sacrificing performance. We deploy CW-Net on a real self-driving car and show that the resulting explanations improve the human driver’s mental model of the vehicle, allowing them to better predict its behaviour, particularly in surprising situations. This demonstrates that explainable deep learning integrated into self-driving cars can be both understandable and useful in a realistic deployment setting. We anticipate our method could be applied to other safety-critical systems, such as autonomous drones and robotic surgeons, as well as to other architectures, such as end-to-end learning systems and vision–language–action models. "

"CW-Net grounds the reasoning of a black-box ML planner in human-interpretable concepts, such as ‘Approaching stopped vehicle’ or ‘Close to cyclist’. This method is rooted in case-based reasoning, a classical artificial intelligence (AI) approach31,32,33,34 inspired by cognitive models of human reasoning and memory35. CW-Net can be applied to arbitrary pretrained deep neural networks, does not require retraining from scratch and does not degrade the performance of the original black-box ML planner. As the inferred concepts are the sole input to the final decision-making module of CW-Net and therefore directly determine AV behaviour, we refer to them as causally faithful. Notably, our approach contrasts with popular post hoc explanation methods36,37, which are applied to pretrained models and do not, by construction, guarantee faithfulness to the model decision process"

My 2c: GoPro's fall started years ago when they decided to be a media company (like Youtube) and spent a lot of time + effort both recruiting media focused talent, instead of on devices as such.

Also podcast: AI is the First Bubble with Every Ingredient All At Once - https://www.themebfabershow.com/episodes/WrD9xoqhwiy

This is fckn brilliant!!

I’m bookmarking it for sharing