Sign in

Sophie Huiberts

@sophie.huiberts.me
1.5K followers 458 following 817 posts

CNRS researcher in linear programming

PostsRepliesMedia
Reposted by Sophie Huiberts
Dulwich Quantum Computing @dulwichquantum.bsky.social · 5h
OpenAI: I proved that only 3 MUBs are possible in dimension 6. github.com/openai/math/... Me: Great, but where are the scripts your solution relies on? - seed.cpp - covout.cpp - ub.cpp - patchx.cpp - atlas.cpp - stageB.cpp - stageC.cpp - proof_run.py - ... OpenAI: Ups, forgot to attach...
162
Sophie Huiberts @sophie.huiberts.me · 40m
This is my new favorite mnemonic to remember the order of matrix indices. Matrix-vector product to turn your meal plan 🥗🥙🍝🥫 into a shopping list 🍅🥬🧅🧆
Text on top: How many ingredients are needed to cook meals?
Below is a matrix-vector product. The matrix has entries A subscript ingredient emoji comma meal emoji. The vector has entries falafel sandwich emoji, salad emoji, pasta marinara emoji, can of soup emoji. After the equality sign we find a right-hand vector with entries leafy green emoji emoji, onion emoji, falafel balls emoji.

From the word "ingredients" in the title goes an arrow to the ingredient vector, and from the word "meals" goes an arrow to the meal vector."The units and indices line up: A subscript tomato comma soup is measured in tomatoes per can. Ingredient (row) goes first. Meal (column) goes second.
040
Sophie Huiberts @sophie.huiberts.me · 17h
speaking as someone intimidated by stack exchange: how would debate there work? i thought they only allow questions and answers
120
Reposted by Sophie Huiberts
Clément Canonne @ccanonne.github.io · 18h
I wrote a blog post on "TCS, AI, and the Adults in the Room," addressed to the senior researchers and established academics in our field: minorfree.github.io/Clement/ Thanks to @hunglv.bsky.social for lending me his blog!
56024
Reposted by Sophie Huiberts
Barbara Fantechi @barbarafantechi.bsky.social · 19h
I wasn't aware, or possibly was aware and forgot, of the marvellous "video of research talks journal" Mathematical Discourse. My deepest thanks to all the editors. www.mathematicaldiscourse.org
mathematicaldiscourse.org
Mathematical Discourse - Mathematical Discourse
A peer-reviewed video journal for mathematical research talks.
2135
Sophie Huiberts @sophie.huiberts.me · 22h
Seeing this metaphor of 'conjecture as lighthouse' a lot lately. I don't like it, I thought lighthouses are a thing you should avoid (at risk of sinking and drowning). Can any lighthouse expert weigh in on this?
picture of text. "A mathematical question is much more than a theorem waiting to be proved; it is a source of momentum. First and foremost, it is a lighthouse in the night: it illuminates and guides mathematicians in their wanderings, both esthetic and scientific. "
0100
Sophie Huiberts @sophie.huiberts.me · 10/10/2026
I don't know what anyone supposed to get from this information? Since OAI is so bad at citing?
120
Reposted by Sophie Huiberts
Edward Vogel @mathartforall.bsky.social · 10/10/2026
"At minimum you need to explain what you did, including which steps required new ideas and which steps are already well-known. More substantively, you are expected to nurture the literature." - Words to thrive by
081
Reposted by Sophie Huiberts
Hung Le @hunglv.bsky.social · 09/10/2026
The problem with big names praising "breakthroughs" by OpenAI is that it gives OpenAI a pass on mostly sloppy and sometimes incorrect proofs, treating them as "machine things" that we should accept as a "small price". Would love to see them read the proofs and then write their understandings.
2151
Reposted by Sophie Huiberts
Anuj Dawar अनुज दावर انُج داور @anujdawar.bsky.social · 09/10/2026
Hearing anecdotally of other such stories. Then there's this. curticapean.net/math-notes/
curticapean.net
Timeline of Math Results
A mathematics result catalogue with discussions and synchronized timelines.
051
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
I Wrote about the responsibility we carry to each other.
sophie.huiberts.me
Community and Company
My community is under attack by commercial interests.
15720
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
lmao this commit message to describe 3 retractions
github screenshot of a commit. "Update manuscripts and Lean formalizations
dr-openai committed 7 hours ago"
1672
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
True, if a non-expert outside the walls of OpenAI did the same then that would be just as bad. But I just don't think that that justifies the non-experts inside OpenAI doing it.
010
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
But we don't need OpenAI to publish this, no? Probably the same results were already being prompted by other more responsible researchers. Those people would have written a good paper immediately, without first putting out sloppy or faulty manuscripts. Example:
220
Reposted by Sophie Huiberts
Ragnar {Groot Koerkamp} @curiouscoding.nl · 08/10/2026
Turns out 3 of the papers have been retracted based on a sign error in one of them. github.com/openai/math/...
3 withdrawals based on a sign error in a paper on the Algebraicity of Weill classes on split abelian eightfolds.
4507
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
Everybody complains about the abysmal quality of the writeups. So OpenAI is making the state of the literature actively worse. I would not call contributing. Consider: they say it took them 3 hours of compute per problem. Why publish that? Its not like nobody else would have found it soon anyway
130
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
i'm ready to believe that the results are real and independent. but i think it's misplaced for the supplier of this software to interfere with its (potential) customers. say what you will about people writing papers with LLM generated results, its preferable over this irresponsible dump of trash
130
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
no idea 🤷‍♀️
010
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
*subset sum
010
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
oops lmao yeah i meant to write subset sum
120
Sophie Huiberts @sophie.huiberts.me · 08/10/2026
OAI scooped a customer on 3SUM arxiv.org/abs/2610.10446
arxiv.org
Subset Sum via Partial Match
We show that the subset sum problem can be solved in time $O(2^{0.499999n})$, breaking the $2^{n/2}$ meet-in-the-middle barrier. Our approach builds on the representation technique framework of Rand...
6171
Reposted by Sophie Huiberts
Anastasia @hamelinny.bsky.social · 07/10/2026
when I said I like the aesthetic of research in Fallen London, it was more like we should dress goth, less like we should be deciphering unintelligible scripts that damage our mental health
051
Sophie Huiberts @sophie.huiberts.me · 07/10/2026
its hard to understand people who use LLMs 'to improve the writing' because what is this garbage 😭
030
Reposted by Sophie Huiberts
Michael Barany @mbarany.com · 07/10/2026
importantly, most theorems in pure mathematics are exhaust fumes from other things (training, trying to build collective understanding, maintaining systems of employment and recognition, etc), so this is a different kind of exhaust but it's not unrelated
071
Reposted by Sophie Huiberts
Sophie Huiberts @sophie.huiberts.me · 01/10/2026
A script kiddie has claimed a new matrix multiplication constant improvement 😵‍💫
0132
Sophie Huiberts @sophie.huiberts.me · 06/10/2026
The text Taylor cites is titled "How the result was originally found" but it is actually a financial disclosure. Of an undisclosed sum of money directly into the authors' pockets.
190
Sophie Huiberts @sophie.huiberts.me · 06/10/2026
On the other hand, later the authors write "The authors’ contribution to this publication was done in their individual capacities, not in connection with their duties or responsibilities to Columbia University or to MIT." Despite those being their listed affiliations on the top of the manuscript.
190
Sophie Huiberts @sophie.huiberts.me · 06/10/2026
I am not sure how to feel about this. One the one hand, if Anthropic is asking researchers to look at their pdfs then yeah compensation seems justified
1143
Sophie Huiberts @sophie.huiberts.me · 05/10/2026
dolphins jump out of beautiful water. a rainbow is above them, and in the sky are clouds shaped like hearts and also hot air balloons. the image is captioned with "how it feels to spread misinformation"
040
Sophie Huiberts @sophie.huiberts.me · 03/10/2026
ohh where is this?
111
Reposted by Sophie Huiberts
Lee Hurley @hleehurley.com · 03/10/2026
I implore all cis people to read this.
17990417
Sophie Huiberts @sophie.huiberts.me · 02/10/2026
This seems like a strange change? 1. STOC makes arXiv posting mandatory because of submissions increase 2. arXiv tightens its rate limit because of submissions increase 3. STOC abandonds arXiv posting because of said rate limit? Not sure what I am not understanding here.
220
Sophie Huiberts @sophie.huiberts.me · 01/10/2026
A script kiddie has claimed a new matrix multiplication constant improvement 😵‍💫
0132
Sophie Huiberts @sophie.huiberts.me · 29/09/2026
simons.berkeley.edu/ai-tcs-worki...
010
Sophie Huiberts @sophie.huiberts.me · 29/09/2026
Indeed. That Simons organizes where they are at, makes sense. But for anyone to claim they can support a diverse sample of visitors to fascist America is both laughable and embarrassing.
041
Sophie Huiberts @sophie.huiberts.me · 29/09/2026
I would say, both organizing on short notice and organizing in the USA are choices. And they are choices with consequences, the smallest of which is that all claims of diversity are moot
180
Sophie Huiberts @sophie.huiberts.me · 29/09/2026
Does anyone consider this a 'diverse sample balanced across institutions'? Asking sincerely
List of people, with names redacted.

****** (Columbia University), ****** (Purdue University, Google), ****** (Stanford University), ****** (Harvard University), ****** (Rutgers University), ****** (Apple), ****** (UCSB / UCLA), ****** (Ohio State University), ****** (IT University of Copenhagen), ****** (University of California, Irvine), ****** (New York University), ****** (University of Illinois Urbana-Champaign), ****** (Princeton University), ****** (Massachusetts Institute of Technology), ****** (University of California, San Diego), ****** (Institute for Advanced Study), ****** (University of Chicago), ****** (Anthropic / on leave from UC Berkeley), ****** (University of Washington), ****** (OpenAI / on leave from UC Berkeley), ****** (Massachusetts Institute of Technology), ****** (University of Illinois Chicago), ****** (Institute for Advanced Study / a16z crypto), ****** (Columbia University), Min Jae Song (University of Chicago), ****** (University of California, San Diego), ****** (Google / on leave from Carnegie Mellon University), ****** (UC Berkeley)
2150
Sophie Huiberts @sophie.huiberts.me · 28/09/2026
The most important principle when teaching LP is to not cover any theoretical nonsense
sophie.huiberts.me
Principles for Teaching Linear Programming
4223
Sophie Huiberts @sophie.huiberts.me · 28/09/2026
This is more intuitive for the (first-order) Peano axioms. Lots of stuff is independent of Peano's axioms, but if you choose the 'wrong' axiom to add then it no longer describes the 'standard model' of PA (the natural numbers). Similarly, set theorists can have opinions about sets and CH.
130
Sophie Huiberts @sophie.huiberts.me · 28/09/2026
People have beliefs about what sets are which are not captured in the ZF axioms. For example, the value of the the 748'th Busy Beaver number BB(748) is independent of ZF. There is a 'correct' value of BB(748), but ZF cannot prove that.
120
Sophie Huiberts @sophie.huiberts.me · 24/09/2026
querty? that sounds even scarier than azerty
110
Sophie Huiberts @sophie.huiberts.me · 23/09/2026
Tarik is one of the brightest thinkers in our field. Each and every time it is a joy to read what he writes.
proofsandprompts.com
Navigating the AI crisis: a humble guide for students
Tarik Aougab, Professor at Haverford College The premise for this post is simple: no matter what one thinks the central goals or purposes of mathematics are, the achievement of those goals by human…
0111
Reposted by Sophie Huiberts
Huck Bennett @huckbennett.bsky.social · 22/09/2026
Wow does the post title bury the lede! I particularly like the pictured comment.
35715
Sophie Huiberts @sophie.huiberts.me · 20/09/2026
I make it a point to never use or teach any form of tableau 😅 imo it is confusing because treating variables and slacks interchangeably should be a type error. But I do really love the idea of gamifying the simplex method of course :)
111
Sophie Huiberts @sophie.huiberts.me · 20/09/2026
first time i feel sad for not understanding tableaus
111
Reposted by Sophie Huiberts
Clément Canonne @ccanonne.github.io · 19/09/2026
"The proof [...] was obtained in a conversation with ChatGPT-6 Astra on Tuesday, September 15, 2026 at 1:02 AM PDT. We then prepared this manuscript for public release, with the intent of uploading it on the morning of Thursday, September 17, 2026" Shameless, and shameful. arxiv.org/abs/2609.207...
arxiv.org
The Strong Secretary Conjecture is True for Linear Matroids
We prove a $1/e$ guarantee for the matroid secretary problem on linear matroids, therefore settling the strong secretary conjecture in this class of matroids. The result holds both when the matroid is...
2459
Sophie Huiberts @sophie.huiberts.me · 16/09/2026
Haha yeah its definitely tough! Do hop into the speedrunning discord if you want any resources or advice for your first clear :)
110
Sophie Huiberts @sophie.huiberts.me · 15/09/2026
I am using Typst to write my lecture notes this semester. It is a very pleasant experience.
1190
Sophie Huiberts @sophie.huiberts.me · 13/09/2026
hmm okay. so does that mean you don't expect the introduction of optimizing software to really affect job satisfaction among planners?
100
Sophie Huiberts @sophie.huiberts.me · 13/09/2026
when scheduling optimization software enters a business, how do the workers in the planning office feel about that? do they feel like the "puzzle solving" part of their job is lost? #orms
141