Greg Bodwin @gbodwin.bsky.social · 23/09/2026OpenAI be like, source: it was revealed to me in a DRM 050
Reposted by Greg BodwinClément Canonne @ccanonne.github.io · 08/09/2026Well, this Navier-Stokes affair did blow up in finite time 843862
Reposted by Greg Bodwinsingular locus sarah @scgriffith.bsky.social · 30/08/2026i haven't seen it mentioned since i started posting here so i'm reupping "the ideal mathematician" by david and hersh personalpages.manchester.ac.uk/staff/hung.b... 0183
Reposted by Greg BodwinGautam Kamath @gautamkamath.com · 05/07/2026When someone absolutely demands a combinatorial construction for linear-sized cut sparsifiers 0171
Greg Bodwin @gbodwin.bsky.social · 01/06/2026RIP Fermat you would have loved \usepackage[margin=0.5\textwidth]{geometry} 081
Reposted by Greg BodwinThatchaphol Saranurak @eigx.bsky.social · 06/04/2026I poured my soul into building this course last fall: 📚 Graph Algorithms via Graph Decomposition 📚 Graph decomposition has been a powerful framework in graph algorithms for over 20 years, but the literature is scattered and technical. Thus, I tried to organize part of it into one coherent story. 3579
Greg Bodwin @gbodwin.bsky.social · 06/04/2026This paper actually does a bit more than that - e.g., it can approximate dist_{G\F}(s, t), not just report reachability. But the point of it, for me, is that it reveals that the F ⊂ E(G) setting is a truly different yet equally fundamental model. Now we all have to go try to understand it. 000
Greg Bodwin @gbodwin.bsky.social · 06/04/2026This paper finally settles that problem ... on the *upper* bounds side. There is actually a better solution (about n*sqrt(f) bits) than the famous Nagamochi-Ibaraki paper, if we require F ⊂ E(G)! This is extremely surprising. 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026For years, this "break" was considered a technical artifact of the lower bound proof method, and it was regarded an interesting but somewhat mechanical open problem to extend the lower bound to the more restricted F ⊂ E(G) setting. 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026Well, there's a small catch. In the lower bound, the edge set F used to distinguish H_1 and H_2 might include some edges that are not in H_1. So it breaks in the natural but more restricted model where we require the set F given in queries to be a subset of the edges of the input graph G. 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026So nf bits is the right space bound, and the problem has been closed for about 30 years. Right? 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026(...) Let F be all the other edges incident to u in H_2. Then F is an (s, t) cut in H_2, but not in H_1. Since there is a query that differentiates H_1 and H_2, we need to use different data structures for them. So all 2^{nf} subgraphs of G need different bit representations, so one uses nf bits. 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026There is also a standard lower bound showing that about nf bits of space are necessary. It goes like this: think of your favorite f-regular graph G. Consider any two subgraphs H_1, H_2 of G, and let (u, v) be an edge contained in H_1 but not H_2. (...) 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026There is a famous paper by Nagamochi and Ibaraki from 1992 that shows how to build such a data structure, for any n-node input graph, in about nf bits of space. 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026You have an undirected input graph G. You want to build a small-space data structure that can answer queries of the form: given two nodes (s, t) and a set of |F| <= f edges, is F a cut for (s, t)? That is, are there any remaining s-t paths in G\F? 100
Greg Bodwin @gbodwin.bsky.social · 06/04/2026I really like this paper, one of my favorites in recent memory. Quick explainer of what's going on: 150
Greg Bodwin @gbodwin.bsky.social · 12/02/2026I am thinking of starting a cult that believes that fast max flow algorithms might someday be used to produce paperclips in a manner that will destroy humanity. It's unfair that the AI subfield has monopolized this. 020
Greg Bodwin @gbodwin.bsky.social · 06/02/2026Academia hack: when writing scathing rejections, also demand that the authors add citations to your enemies 050
Greg Bodwin @gbodwin.bsky.social · 07/12/2025Motion to bring medieval kerning back in our papers 0280
Reposted by Greg BodwinarXiv cs.DS Data Structures and Algorithms @csds-bot.bsky.social · 28/10/2025Kevin Pratt, Yahel Uffenheimer, Omri Weinstein: (Approximate) Matrix Multiplication via Convolutions arxiv.org/abs/2510.22193 arxiv.org/pdf/2510.22193 arxiv.org/html/2510.22193 063
Greg Bodwin @gbodwin.bsky.social · 28/10/2025this is a pretty accurate summary of my research area 0242
Greg Bodwin @gbodwin.bsky.social · 20/10/2025I always thought it seemed weird how fancy waiters in movies would be like "Excellent choice, sir" after someone orders off a fixed menu. Now ChatGPT does this, and I can confirm: it is indeed weird 020
Greg Bodwin @gbodwin.bsky.social · 03/10/2025Some UMich TCS lore: a (now-retired) professor once photoshopped these shocked Hilberts to express his surprise at increasingly complicated Hilbert's Hotel situations. They are now in use as all-purpose math reactions. Please enjoy 050
Reposted by Greg BodwinHung Le @hunglv.bsky.social · 25/08/2025Some questions on spanners in my talk at the Simons Institute. Since the talk, progress has been made on a few questions, but most are open. minorfree.github.io/SpannerQues/minorfree.github.ioSome Questions on Spanners | Rambling on Graphs 262
Greg Bodwin @gbodwin.bsky.social · 01/07/2025adding a "Do you like this personality 👍👎" box to my email signature 040
Greg Bodwin @gbodwin.bsky.social · 04/06/2025but what name could possibly be cooler or more metal than Chris Peikert's Theorem 100
Greg Bodwin @gbodwin.bsky.social · 04/06/2025Thinking about how the security people give their results metal names like "spectre" and "meltdown" and design cool logos for them. We should give that treatment to our theorems 391
Reposted by Greg BodwinMichael Dinitz @mdinitz.bsky.social · 31/05/2025Incredibly well deserved!! 062
Greg Bodwin @gbodwin.bsky.social · 25/05/2025Hot CS take: Big-O notation should have been defined to hide constant factor changes in the input variable, not the output function value 130
Greg Bodwin @gbodwin.bsky.social · 13/05/2025when we gonna stop with these human-written proofs, we were gifted with bots 000
Reposted by Greg Bodwin𝖬𝖺𝗁𝖽𝗂 𝖢𝗁𝖾𝗋𝖺𝗀𝗁𝖼𝗁𝗂 @mahdi.ch · 04/04/2025Huge congratulations to my amazing student Yeyuan Chen (+co-author Zihan Zhang of OSU advised by Zeyu Guo) for being awarded the STOC 2025 Best Student Paper Award! Their monumental result proves that explicit Reed-Solomon codes can correct more errors than previously known: arxiv.org/abs/2408.15925arxiv.orgExplicit Folded Reed-Solomon and Multiplicity Codes Achieve Relaxed Generalized Singleton BoundsIn this paper, we prove that explicit FRS codes and multiplicity codes achieve relaxed generalized Singleton bounds for list size $L\ge1.$ Specifically, we show the following: (1) FRS code of length $... 2477
Greg Bodwin @gbodwin.bsky.social · 14/03/2025Thanks, good nuance. Lately I've been thinking about how I write papers assuming that the reader is going sequentially ("I don't need to re-explain this nuance - I just talked about it a page ago") but I almost never read this way. I think I have room for improvement here. 010
Reposted by Greg BodwinChris Peikert @chrispeikert.bsky.social · 14/03/2025I find that this works, mostly, but requires the right amount of skepticism. Too much leads me to the “dual” of “fully trustful” reading, bogged down in doubting every little piece of ink, and not seeing the shapes or overall picture. 193
Greg Bodwin @gbodwin.bsky.social · 04/03/2025when my family asks me about the impact of my research 14813
Greg Bodwin @gbodwin.bsky.social · 23/02/2025frantically changing "graph theory" to "autonomous killer graph theory" in all my papers 1192
Reposted by Greg BodwinHuck Bennett @huckbennett.bsky.social · 21/02/2025This is terrible news, and part of the ongoing assault on science in the U.S. by the new administration. To honor Tracy Kimbrel and his service to the NSF's AF division, here's a short thread about a beautiful algorithm of his, joint with Rakesh Sinha (www.sciencedirect.com/science/arti...). 1/sciencedirect.comA probabilistic algorithm for verifying matrix products using O(n2) time and log2n + O(1) random bits 1116
Greg Bodwin @gbodwin.bsky.social · 06/02/2025I didn't mean me, of course. Everything I say is brilliant and beyond reproach. 040
Greg Bodwin @gbodwin.bsky.social · 06/02/2025A good sign for this adversarial reading is if you find yourself reading the paper in a non-linear order. A trustful reading usually proceeds line-by-line, but in a skeptical reading you constantly skip around to the part that you currently find most suspicious. 040
Greg Bodwin @gbodwin.bsky.social · 06/02/2025The alternative is to read with the view that the results could not possibly be correct, and the authors (those fools) could not possibly have proved it, and you will find the inevitable flaw. And then as you read you're slowly, begrudgingly, forced to admit that they were right after all. 171
Greg Bodwin @gbodwin.bsky.social · 06/02/2025In particular, trustful reading leads to the sense that you can verify each individual step of the proof, but you have no idea of the bigger picture, how you might have come up with it yourself, or how to extend the argument. 120
Greg Bodwin @gbodwin.bsky.social · 06/02/2025I think a common newcomer mistake is to read papers automatically trusting all the claims made by the authors, because the authors are smart and the paper was published and so the claims are probably correct. The claims *are* probably correct, but you should set this aside as a reader. 120
Greg Bodwin @gbodwin.bsky.social · 06/02/2025A proof is a logical argument written to convince a skeptical audience. A corollary is that the best way to read a proof is to roleplay as a skeptical audience. 2102
Greg Bodwin @gbodwin.bsky.social · 16/01/2025Guy who updates his beliefs towards frequentism after witnessing examples of it working 140
Reposted by Greg BodwinMare Pan @paarsec.com · 16/01/2025One of my favourite things anyone has ever said about art. You'll be missed, David. 🌹 26110563560
