links 10/13/26: https://roamresearch.com/#/app/srcpublic/page/10-13-2026

• • https://en.wikipedia.org/wiki/Leontyne_Price impressed by how many people and institutions were involved in giving her her start. it's a long list of mentors and colleagues who recognized her talent, devoted family members, black community institutions, etc. would we do as well today at recognizing and supporting an extraordinary talent who faced discrimination? maybe better in some ways, maybe worse in others.
    • i was looking to see if there was any commentary about how odd her voice sounds in the Aïda video I was watching from the 70s -- almost bell-like, no vibrato at all -- but no mention of it. I think it might be bad recording quality; she sounds normal in other things.
  • https://en.wikipedia.org/wiki/Jenks_natural_breaks_optimization algorithm for data clustering. Claude suggested it for partitioning a grayscale image into tone bands
  • https://www.cs.ucdavis.edu/~ma/SIGGRAPH02/course23/notes/papers/Winkenbach.pdf classic 1994 paper laying out the concepts of how to generate computer graphics images that resemble pen-and-ink drawings. Recommended by Claude.
  • https://whitehotmagazine.com/articles/path-carved-through-layers-time/7765 Jiang Miao has a strikingly original visual vocabulary. turns out she carves into paint layers with a knife. Art inspired by Taoist symbolism.
  • https://www.gutenberg.org/ebooks/3543 Heartbreak House, by George Bernard Shaw. I need to reread this. he's such a master.
  • https://petson.livejournal.com/200935.html Diana Wynne Jones on heroism in fiction.
  • https://venkateshrao.com/blyg/f/5sg927h6j0zabw1nt708xhn6y4/ Venkatesh Rao on cryptography
  • https://www.motrpac.org/ big ten-year study trying to find the molecular mechanisms by which exercise affects the body, acutely and chronically
  • https://en.wikipedia.org/wiki/Type_2_diabetes in type 2 diabetes, the beta cells are eventually destroyed
  • https://arpa-h.gov/sites/default/files/2026-08/Heilmeier-Questions-Aug26.pdf ARPA-H's guide to writing up a program succinctly -- the "Heilmeier Catechism", explained
  • https://karagila.org/2026/openai-pp/ mathematician Asaf Karagila about why he isn't bothering to evaluate AI-generated proofs. They're very hard to read and OpenAI has not bothered to communicate them clearly.
    • also as he says here: https://karagila.org/2026/lean/ a Lean certificate does not guarantee a correct proof. The formalization into Lean may not be a fair representation of the theorem, and there are known to be bugs in the Lean kernel itself. How do we know the bot hasn't found an exploit?
      • when I've seen Twitter commentary by non-lab-employee mathematicians who are Lean experts, which iiuc Karagila is not, they generally seem to think it's implausible that the proofs are just exploiting Lean bugs, for reasons too technical for me to understand. But Karagila is correct that this is in principle possible; a Lean certificate is certainly not a guarantee that the theorem was indeed proved.