↵ select ↓ ↑ navigate esc close

marginal revolution read out: Saturday assorted links

marginal revolution read out ·

Listen: https://serve.buildtall.systems/b27c1d24b378fb87ea76630ede0580e17669b776c5c6eac71dd483597d08eca4.mp3

Source: https://marginalrevolution.com/marginalrevolution/2026/11/saturday-assorted-links-586.html?utm_source=rss&utm_medium=rss&utm_campaign=saturday-assorted-links-586


This is marginal revolution read out, Saturday assorted links. In this post, the author shares seven links covering economics, music, film, mathematics, development, conservation and AI. No opening remarks beyond the list itself.

The first link is Madrid vs. Catalonia, a post by Martin Varsavsky on X. The author lists it without additional comment. Varsavsky argues that Madrid outperforms Catalonia on GDP per capita, job creation and safety, and attributes much of the gap to the origins of immigration. He notes that in Madrid more than half of foreign residents are from Latin America while only about eight percent are African, whereas in the province of Barcelona Africans are sixteen percent and Latin Americans below forty percent. He claims language is decisive: a Venezuelan can start work immediately, while someone from Morocco must first learn Spanish, and often Catalan as well. He points to social security data showing higher contribution rates among Latin American immigrants than among Moroccans, and to crime statistics showing Moroccans are overrepresented in Spanish prisons relative to their population share. Varsavsky calls for the Interior Ministry to publish birth country of detainees. The argument is provocative and likely to draw strong reactions.

The second link is Stephen Kovacevich, an obituary in the New York Times for the pianist who died at eighty-six. The author says, quote, Brahms and the Diabellis stand among his towering pianistic achievements, and adds, he was one of my favorite pianists. The Times piece by Tim Page recounts Kovacevich's long career under three different names, his early debut at eleven in San Francisco, and his protégé years with Myra Hess, who shaped his understated power. It quotes his obsession with late Beethoven and his metaphysical reading of the Hammerklavier fugue. Kovacevich performed with major orchestras and had artistic ties with Jacqueline du Pré and Martha Argerich. The piece notes his measured intensity and tenderness matched with vigorous intelligence.

Third is Queen at Sea, a film the author calls an excellent movie, though hard to watch due to the dementia theme, adding that it is impossible to review without spoilers. The link leads to a YouTube page, but the post offers no further detail. So this one stands on the author's high recommendation and warning that the subject matter is emotionally difficult.

Fourth is Claude revises the notebooks of Gödel, a Guardian piece by Alex Bellos. It reports on how the AI model Claude helped Finnish logician Jan von Plato make sense of more than a thousand pages of Kurt Gödel's workbooks from the early nineteen-forties, much written in an obsolete German shorthand. Progress had been about a page a week, but after showing Claude his transcriptions and prior books, von Plato received a six-hundred-fifty-page explanation within days. The AI connected pieces across a thousand pages, revealing hundreds of unproved or unknown theorems. Claude produced a reconstruction showing Gödel was one step away from proving the independence of the axiom of choice, a result that later won Paul Cohen the Fields Medal. Von Plato checked the proof in Coq and it passed. The article quotes Notre Dame logician Joel David Hamkins cautioning that how much is Claude and how much is Gödel remains a job for historians. The author's take is simply the title, but the story is a striking example of AI accelerating scholarship.

Fifth is FT Lunch with Jean Drèze, from the Financial Times. The author lists it with no commentary. The title indicates it is the FT's regular Lunch with interview series, this time featuring Jean Drèze, the development economist known for his work on poverty, welfare and rural employment in India. Without the article text or any author remarks, there is little more to add, but the link suggests interest in Drèze's views on economic and social policy.

Sixth is The loneliness crisis for Sri Lankan elephants, a UC San Diego study published on the cover of Science by Anastasia Madsen, Shermin de Silva and colleagues. Over twelve years at Udawalawe National Park, they tracked more than four hundred forty wild female Asian elephants and found that gradual losses from deaths and emigrations, with few new arrivals, pushed the population past a social tipping point. Communities shifted from large, densely connected groups to smaller, more isolated ones. The researchers warn that once past such a point it is much harder to return, and that social fracturing can harm foraging, information sharing and raising young. The author lists it without extra comment, but the finding resonates with human loneliness too, as de Silva notes, digital communication does not replace sharing space.

Seventh is Joshua Saxe and Dwarkesh on RSI, a thread from Joshua Saxe commenting on Dwarkesh Patel's essay about recursive self-improvement. Saxe calls Patel's mechanism the clearest plausible path to an RSI takeoff he has seen, and says it has made him much less of a skeptic. He explains that previous objections hinged on the need for humans in the loop to shape post-training data and reinforcement learning environments for economically valuable tasks. Patel instead focuses on hill-climbing the learning algorithms themselves, optimizing models' ability to learn more data- and compute-efficiently, rather than optimizing value-producing tasks. Saxe praises this for grounding RSI in technical plausibility, then proposes next steps: gathering data on the economic returns to learning-algorithm improvements and whether capital markets would fund an intelligence explosion, and checking whether Patel's mechanism shows up empirically today. He notes the necessary data are hidden, so researchers must infer hidden variables from downstream benchmarks. The author includes no added commentary, but the exchange is a substantive push to make RSI discussion more precise.

Across the seven links, the post moves from contentious social analysis, to an artist's legacy, to a film recommendation, to AI's role in unlocking a mathematical giant's private work, to development economics, elephant social collapse, and finally the nuts and bolts of AI self-improvement. It is a classic Saturday assortment: each item stands alone, but together they touch on how technology, loss, social structure and economics intersect.

  1. Saturday assorted links
  2. Madrid vs. Catalonia
  3. Stephen Kovacevich
  4. Queen at Sea
  5. Claude revises the notebooks of Gödel
  6. The loneliness crisis for Sri Lankan elephants
  7. Joshua Saxe and Dwarkesh on RSI