11011110
About All posts
« The hyperbolic spiral as a trisectrix

Linkage

Oct 31, 2025

  • Well-meaning but misguided editors are using AI to translate Wikipedia into vulnerable languages (\(\mathbb{M}\)). The result is to accelerate the deterioration of online content in these languages, sending them into a “doom spiral”.

  • Image searches for Diwali lanterns turn up cubocathedral lamp shades. Robin Houston wonders why.

  • Cyanotypes of Miura-ori folds, Elecia White. That is, the cyanotype paper was itself folded into a Miura-ori, exposed in folded form, developed, and flattened.

  • SOSA 2026 accepted papers (\(\mathbb{M}\)). From the Symposium on Simplicity in Algorithms to be held this January in Vancouver.

  • Ruth Asawa retrospective at MoMA in New York (\(\mathbb{M}\)). A lot of her art consists of abstract geometric forms, in both sculptural wireframes and flat pieces.

  • Don’t bother reading this arXiv/cs.CC preprint claiming to use Lean to prove \(\mathsf{P}=\mathsf{NP}\) (\(\mathbb{M}\)). In the discussion link and a later thread, the comments point out that it is the same old “nondeterministic programs are nondeterministic and there can be no other way than running the program” nonsense, the two different Lean “complete proofs” claimed in the paper both do not compile, the repository containing a “full proof” linked in the paper does not exist, at least one claimed theorem is obviously false and two more contradict each other, many proofs do not prove what they are claimed to prove, and the underlying definitions do not work. So anyway, now we can add “giving us good reasons not to trust claims of computer-verified proofs” to the benefits of AI.

  • If you use Firefox (as I do after Chrome disabled uBlock Origin) you might find helpful these tips for disabling its AI features (\(\mathbb{M}\), see also).

  • An apparent contradiction between demands that universities not look at or consider national origins of student applicants and that they limit students from other nations than the US to a small fraction of admissions (\(\mathbb{M}\)). Far from the stupidest or most destructive thing the current US administration has done, but symptomatic of their active avoidance of thought and care.

  • Poetry of repetition: constructing verse through combinatorial design theory (\(\mathbb{M}\)). A curious arXiv preprint by Ajani De Vas Gunasekara and Miriam Wei Wei Lo, found by Chris Purcell using arXivist.com, a site for generating personalized announcements of interesting new arXiv preprints, on an otherwise slow day.

  • Python Software Foundation refuses a $1.5M grant from the National Science Foundation (\(\mathbb{M}\)), because of conditions that to accept the money it should renounce all efforts at improving diversity, equity, and inclusiveness. The PSF chose principle over cash and set an excellent example for us all.

  • SpiderMonkey replaces GraphViz by a custom graph layout algorithm specialized for visualizing the block structure of computer programs (\(\mathbb{M}\), via). They cite as reasons for the change two problems with GraphViz-based layout: First, because GraphViz is based purely on the graph structure of the code blocks and not on the logical structure of the code, its placement of nodes makes it hard to understand the nesting, sequential ordering, and loop structure of code blocks. And second, because it allows big rearrangements in order to try to optimize the layout, small changes to the code can cause big changes to its layout. Instead, their custom algorithm simplifies the usual Sugiyama-style layered graph drawing framework by using its understanding of code structure to break cycles (because this layout needs its graph to be a DAG) and to assign blocks to layers, and by skipping the part of the framework that permutes blocks within each layer (as counterproductive for understandability and stability).

  • AI and the power of nonuniform circuits (\(\mathbb{M}\)). Lance Fortnow sees LLM training as analogous to the advice used to differentiate uniform from non-uniform complexity classes. The successes of LLMs hint that computational devices with this sort of advice might be thought of as surprisingly powerful rather than merely treating the advice as a technicality to be worked around or an awkward way of providing randomness.

  • Dave Richeson makes giant cardioid string art with the help of some volunteers remaining on his campus over a break.

  • UCLA math department TA, grader cuts spark concern over student learning, support (\(\mathbb{M}\)). The same thing is, I think, happening across UC campuses and departments. A teaching assistant union contract from a couple years ago (now under renegotiation) roughly doubled the cost of hiring a TA, but the state and the central UC administration did not provide any more money to pay for this. The predictable outcome is that the number of TAs per department and per class (hired separately each term on a term-by-term basis) have been dramatically slashed. The undergraduate students who depend on teaching assistants to help them with their coursework don’t get as much help, undergraduate graders are not being hired at all any more, and graduate students who used to depend on TA jobs to fund their graduate education are going unsupported.

  • David Eppstein

Geometry, graphs, algorithms, and more