Sophie Huiberts @sophie.huiberts.me · 12/10/2026This seems very permissive. Is one allowed to read an LLM-generated script? 340 4
Sophie Huiberts @sophie.huiberts.me · 11/10/2026This is my new favorite mnemonic to remember the order of matrix indices. Matrix-vector product to turn your meal plan 🥗🥙🍝🥫 into a shopping list 🍅🥬🧅🧆 1200 20
Sophie Huiberts @sophie.huiberts.me · 10/10/2026Seeing 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? 2110 11
Sophie Huiberts @sophie.huiberts.me · 08/10/2026lmao this commit message to describe 3 retractions 1702 70
Sophie Huiberts @sophie.huiberts.me · 29/09/2026Does anyone consider this a 'diverse sample balanced across institutions'? Asking sincerely 2150 15
Sophie Huiberts @sophie.huiberts.me · 07/09/2026Two decades since the original mistake, this is still how the page is. I wonder how many students are confused by this error every semester. 140 4
Sophie Huiberts @sophie.huiberts.me · 07/09/2026In 2023, the error propagated onwards into the page introduction. 110 1
Sophie Huiberts @sophie.huiberts.me · 07/09/2026In August 2005, a Wikipedia editor wrote a section called 'standard form' on the page Linear Programming. This editor got the forms confused and actually described the canonical form. 110 1
Sophie Huiberts @sophie.huiberts.me · 07/09/2026In Linear Programming we have two notational conventions, helpfully called the 'standard form' and the 'canonical form'. As you can see, this is confusing 4210 21
Sophie Huiberts @sophie.huiberts.me · 06/09/2026In 1986, UC Berkeley paid $15,000 annually for the linear programming software MPSX archive.org/details/CAT8... 1131 13
Sophie Huiberts @sophie.huiberts.me · 06/09/2026related: if anyone knows where to find this proceedings i would be very grateful! 210 2
Sophie Huiberts @sophie.huiberts.me · 06/09/2026In the 80s, the USDA had a mainframe computer and was solving linear programming problems for a frustratingly expensive $50-$500 a piece. 10.1287/opre.39.1.13 2151 15
Sophie Huiberts @sophie.huiberts.me · 20/06/2026me: wolfram alpha can you convert a quart to liters? wa: its 827 km from quart (italy) to litér (hungary). that takes 56 minutes by aircraft or 4ms for light in a fiber optic cable 😊 131 3
Sophie Huiberts @sophie.huiberts.me · 13/06/2026I made some animations for our next STOC video. Here is a preview of one! 3311 31
Sophie Huiberts @sophie.huiberts.me · 02/06/2026Today at #SIAMOP26, Alex Black will explain why the simplex method runs in polynomial time! Session MS8 at 10:30 080 8
Sophie Huiberts @sophie.huiberts.me · 30/05/2026Recently I've been obsessed with this 1994 paper by Hooker. Among other things, he explains the (largely unexamined) empirical content of NP-completeness theory. More theorists need to read it 0152 15
Sophie Huiberts @sophie.huiberts.me · 06/05/2026Letsgoo it holds in general! We have two pictured results for isotropic convex bodies: compact convex sets with center of mass at 0 and volume 1 Here, L_K is the isotropic constant: the covariance matrix of K is L_K times the identity matrix. Deep + recent theorems say that L_K is indeed constant! 140 4
Sophie Huiberts @sophie.huiberts.me · 06/05/2026For balls in R^n of volume 1, the radius is like √n. Surprisingly, most of the mass can be found inside a slab of width 1 around the equator. Only today did I realize that the all the cool concentration inequalities (Azuma, Hoeffding, McDiarmid) prove the same fact for the cube! 1161 16
Sophie Huiberts @sophie.huiberts.me · 11/04/2026Invited speakers announced! The second MIP European Workshop will take place at Sapienza University of Rome, Italy, from October 19–21, 2026. I hope to see many of you there! www.mixedinteger.org/EUROMIP/2026... 071 7
Sophie Huiberts @sophie.huiberts.me · 07/04/2026Hot take but the importance we attach to the Millennium Prize Problems should decrease with inflation 0141 14
Sophie Huiberts @sophie.huiberts.me · 13/03/2026In the LP relaxation we drop the integrality constraint for our variables, and immediately we see tons of fractionality. Using twelve 80% fractional wall pieces around the grid boundary and two 20% wall pieces right next to the horse, the LP can make almost the entire field be 80% enclosed 111 1
Sophie Huiberts @sophie.huiberts.me · 13/03/2026Where there is a ILP, there is a linear programming (LP) relaxation. To visualize it, first I adapted Buliamti's code to plot the ILP solution. It uses exactly 10 walls, as required 110 1
Sophie Huiberts @sophie.huiberts.me · 13/03/2026You've probably heard of hit game enclose.horse by now. Place 10 segments of wall to make the horse unable to escape, while enclosing maximal surface area. Obviously in NP, unclear if its NP-complete. Time for some mathematical optimization 🧵 1103 10
Sophie Huiberts @sophie.huiberts.me · 16/12/2025The numbers look impressive at first glance. Though we should be wary that perhaps for each paper, only the most enthusiastic author responded 100 1
Sophie Huiberts @sophie.huiberts.me · 01/12/2025or this one, mixing 3D printing and on-screen drawing. it this any good? i know its scuffed but i'm asking if it is fun scuffed or unfun scuffed 240 4
Sophie Huiberts @sophie.huiberts.me · 01/12/2025trying some new animation approaches for the FOCS video. like this one, where i screengrab blender to explain the simplex method (preview is 3x speed) 140 4
Sophie Huiberts @sophie.huiberts.me · 22/11/2025Next video I am considering this. Not super happy with it yet 200 2
Sophie Huiberts @sophie.huiberts.me · 27/10/2025Big shout-out to all the software devs who were willing to speak with us. This project could not have happened without both the open-source and commercial solver's support, and Laurent Poirrier of Bocconi University 1200 20
Sophie Huiberts @sophie.huiberts.me · 11/10/2025found an IRL linear program, took some pictures 4374 37
Sophie Huiberts @sophie.huiberts.me · 23/09/2025Gurobi published their 2025 State of Mathematical Optimization report. Always a nice read! www.gurobi.com/resources/re... 071 7
Sophie Huiberts @sophie.huiberts.me · 02/08/2025GHF brags about delivering 1.3M meals per day. Mathematicians have a moral responsibility to speak out about the fact that 1.3M <<< 6.6M 273 7
Sophie Huiberts @sophie.huiberts.me · 27/07/2025This story of the discovery of Alison Doig the person is wild pursuit.unimelb.edu.au/articles/cel... 1175 17
Sophie Huiberts @sophie.huiberts.me · 22/07/2025"I have been working on this topic for more than 4 decades. There is no case since WW2 of starvation that has been so minutely designed and controlled" 071 7
Sophie Huiberts @sophie.huiberts.me · 16/07/2025omg thats such bullshit >:( i recently had a fight about political ideology in an article i wrote. there, i got a lot of mileage out of pretending i didnt understand what they meant for example, i'd say that the marked words were political (oopsie! my b!) but the rest are politically neutral facts 010 1
Sophie Huiberts @sophie.huiberts.me · 04/07/2025Numbers and data remain a big part of the propaganda, but anyone should recognize that the numbers aren't good. 58 million meals over 5 weeks makes less than 0.8 meal per day per person Besides being a tool of forced displacement, it's also 4x too little food to avoid malnutrition 010 1
Sophie Huiberts @sophie.huiberts.me · 03/07/2025That's it for this #EuroMIP25. Thanks everyone for making this workshop a success 0221 22
![screenshot from paper. NEEDED: AN EMPIRICAL SCIENCE OF ALGORITHMS
J. N. HOOKER
Carnegie Mellon University, Pittsburgh, Pennsylvania
(Received February 1993; revision received October 1993; accepted December 1993)
Deductive algorithmic science has reached a high level of sophistication, but its worst-case and average-case results seldom tell us how well an algorithm is actually going to work in practice. I argue that an empirical science of algorithms is a viable alternative. [text continues, i suggest grabbing the paper instead of reading the screnshot]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:tqd5nxeev5uh47nfjrpxhcci/bafkreieeeerrf2oux2h3z3ju5wdbw3tsz6l6rrvlynz47qwmq5fnzptl2i)
![Screenshot of text from Geometry of Isotropic Convex Bodies (2014) by Brazitikos, Giannopoulus, Valettas and Vritsiou.
Proposition 3.2.12. There exists an absolute constant C > 0 such that for
every isotropic convex body K in R^n and every θ ∈ S^{n−1} one has
[nonsense],
and hence for every t > 0
|{x ∈ K : |langle x, θ rangle| >= C L_K t}| <= 2e^{−t}.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:tqd5nxeev5uh47nfjrpxhcci/bafkreihmghwi7fl3m3xmfptvs6weta6c2aiw2tynttrq4tadsdix7l7ama)
![Picture of text. "Hello everyone, welcome to the video. Today I will tell you about this recent paper I wrote with Eleon Bach. My name is Sophie Huiberts, and this is Optimal Smoothed Analysis of the Simplex Method.
[jump to title card + music]"
Another line of text is cut off near its top, being unreadable but suggesting that the script goes on below the clipping](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:tqd5nxeev5uh47nfjrpxhcci/bafkreiddoptduuyzqi5xrij5tutxqcwaaknll24qkb3yesbbq4v23yp5ta)
