iiinova — where your digital self keeps evolving

Mathieu Besançon @matbesancon.xyz · 08/10/2026

Same thing for academics: you're writing for that paper? Adds time. Working on the content, proofs, experiments? Adds time. Tweaking the latex template? Substrat time. Getting an LLM to mess up the template for you? Remove all the time budget.

120

Mathieu Besançon @matbesancon.xyz · 07/10/2026

Asking here in case: does anyone know a good jurist or lawyer in Germany specialized in citizenship, the precise requirements for applications? Not for me, and yes this is for an edge case of the standard procedures

000

Mathieu Besançon @matbesancon.xyz · 05/10/2026

I was surprised they ended up saying "hey fine use AI to review but give us your prompt if you do"

120

Mathieu Besançon @matbesancon.xyz · 30/09/2026

French tenured positions have been pretty forward-thinking on this: committees ask for candidates' top-k papers, with some members assigned to reading and evaluating said papers. k=5 at for junior (assistant professor) roles and 10 for senior (full prof)

000

Mathieu Besançon @matbesancon.xyz · 30/09/2026

I reviewed way too many papers this year. Frankly I would probably decline reviewing for a venue that explicitly puts the pressure of evaluating (and thus understanding) slop on the reviewers instead of the authors

050

Mathieu Besançon @matbesancon.xyz · 28/09/2026

I think this class would be a great entertainment & education piece to opt researchers as well. You had me at "This is a terrible situation that must be avoided at all costs."

020

Mathieu Besançon @matbesancon.xyz · 23/09/2026

Sure humankind has been bending fire and electricity, but I was today-year old when I learned that Nice has been bending chards into a dessert: fr.wikipedia.org/wiki/Tourte_...

fr.wikipedia.org

Tourte de blettes — Wikipédia

160

Mathieu Besançon @matbesancon.xyz · 19/09/2026

Je ne savais pas que Lean utilisait GMP (ou autre bibliothèque "numérique"

000

Mathieu Besançon @matbesancon.xyz · 15/09/2026

That would have been a highlight of the day! I just passed by both, had a change in Brussels but not even enough to get the waffle myself Another time!

010

Mathieu Besançon @matbesancon.xyz · 15/09/2026

And when I say I got Deutsche Bahned, I don't mean a cute little delay, I mean that I should be around Burgundy and I'm currently watching someone eat a Belgian waffle 🫠

240

Mathieu Besançon @matbesancon.xyz · 15/09/2026

@deutschebahn.com has Deutsche Bahned me again: insane delays but not enough to claim reimbursement, missed connection to an international train and left to fend for myself finding another connection Is it me or is network reliability getting *even* worse? Are there dashboards monitoring it?

010

Mathieu Besançon @matbesancon.xyz · 15/09/2026

C'est futé d'avoir laissé un bout de métal coupé dépasser du sol 🫪 Je sais pas si ça aurait été laissé tel quel sur la voirie motorisée

000

Mathieu Besançon @matbesancon.xyz · 07/09/2026

Les seuls trucs que j'utilise sur ces pages horribles c'est essayer de télécharger le PDF et essayer de trouver l'entrée bibtex, mais un chatbot c'était plus important 🫠

191

Mathieu Besançon @matbesancon.xyz · 04/09/2026

Oh yes that was clear sorry, it was just a spontaneous thought from seeing the location :)

110

Mathieu Besançon @matbesancon.xyz · 04/09/2026

Does CVPR have remote options similar to generic ML conferences? Especially since the venue is in the US

100

Mathieu Besançon @matbesancon.xyz · 03/09/2026

Le bug venait de l'interface ocaml ou de la shared lib appelée par celle-ci ? Je ne suis effectivement pas sûre que ça soit l'interface à Gurobi la plus scrutée

100

Mathieu Besançon @matbesancon.xyz · 29/08/2026

And you'd think that a 45 min correspondence is enough to be safe, you'd THINK

020

Mathieu Besançon @matbesancon.xyz · 29/08/2026

Deutsche Bahn advertising *relaxed* trips on the screen while I spend 4 hours estimating our real delay and plausibility of my correspondence because their updates are systematically over-optimistic. Did not miss this 🫣

030

Mathieu Besançon @matbesancon.xyz · 27/08/2026

I get that @arxiv.bsky.social is under massive submission loads, but not all rules give good filters. Our paper got flagged for repeated text. Without more info, we can only guess that it is about restated things (w/ thm-restate). But it has the .tex source, it could check the source directly no? 🤔

040

Mathieu Besançon @matbesancon.xyz · 24/08/2026

Well in that case it seems one of my trains is just... gone? Luckily there is a replacement, because my second train is a once-a-day cross-border that would have been pretty critical to miss

010

Mathieu Besançon @matbesancon.xyz · 24/08/2026

I like living risky but receiving the dreaded "Changes of your trip" email from Deutsche Bahn, that's too much

130

Mathieu Besançon @matbesancon.xyz · 13/08/2026

MPC is one of the flagship journals in optimization, it filled a crucial gap at its creation for the recognition of computational work and solver development in mathematical optimization

040

Mathieu Besançon @matbesancon.xyz · 13/08/2026

Non-humble brag: we got an "outstanding paper of the year" award of the Mathematical Programming Computation journal for our work on the Boscia framework! link.springer.com/journal/1253...

link.springer.com

Mathematical Programming Computation

Mathematical Programming Computation (MPC) publishes original research articles advancing the state of the art of practical computation in Mathematical ...

2150

Mathieu Besançon @matbesancon.xyz · 29/07/2026

Damn indeed. Except CORE ranks have for legacy reasons more of an aura to them

000

Mathieu Besançon @matbesancon.xyz · 29/07/2026

Whatever we think of CS conf rankings, they are used by hiring committees, etc. The sub-domain bias is pretty striking on CORE. The few integer optimization conferences are now all rank B I believe. Searching for computer vision gives me a bunch of A* ones.

180

Mathieu Besançon @matbesancon.xyz · 24/07/2026

My direct thought as well 🫠

010

Mathieu Besançon @matbesancon.xyz · 07/07/2026

So have we reached the "Freedom Waffles" stage on every US menu already?

020

Mathieu Besançon @matbesancon.xyz · 06/07/2026

@proton.me I haven't found a definitive answer on the various forums, is there a way to "send as" in proton from another email (a professional email)? For academics that have 3+ email addresses this is essential

010

Mathieu Besançon @matbesancon.xyz · 25/06/2026

For academics organizing workshops: EasyChair or ScienceConf? I could cope with making a website separate from the CFP if needed.

010

Mathieu Besançon @matbesancon.xyz · 21/06/2026

I know the intent is good but all papers having an appendix section "AI good and bad we just do the method" is not adding any value

010

Mathieu Besançon @matbesancon.xyz · 21/06/2026

The blanket requirement of ML conferences on ethics and impact statements, regardless of the contribution, and without any training in ethics, is really just clutter on papers and additional noise, with the same empty paragraphs slapped in all of them.

120

Mathieu Besançon @matbesancon.xyz · 11/06/2026

Please do a remix of Tom Cardy Business Man

010

Mathieu Besançon @matbesancon.xyz · 28/05/2026

I had never noticed americans using cheers, thought it was a British thing! So I should possibly deduce that I was always on a formal basis in any American exchange 🫥

100

Mathieu Besançon @matbesancon.xyz · 28/05/2026

That is exactly what I found in my previous email with that person 🥲

000

Mathieu Besançon @matbesancon.xyz · 28/05/2026

For people speaking languages with a distinction of formal and informal "you", are you also looking up through piles of emails to see if you're on an informal basis with someone?

130

Mathieu Besançon @matbesancon.xyz · 26/05/2026

I'll be at the jump.dev workshop and SIAM OP26 in Edinburgh! Come say hi to chat on computational optimization, constrained first-order methods, or hunt down the best specialty coffee and bouldering gyms in town!

jump.dev

JuMP

JuMP is a modeling language and supporting packages for mathematical optimization in Julia.

040

Mathieu Besançon @matbesancon.xyz · 22/05/2026

Des annonces / articles de presses ?

010

Mathieu Besançon @matbesancon.xyz · 17/05/2026

Mais justement "sa place est dans un musée"

010

Mathieu Besançon @matbesancon.xyz · 21/04/2026

Clairement c'est un usage très raisonnable, en le faisant chapitre par chapitre même, et en ne gardant qu'un peu d'ajustements cosmétiques à la fin

000

Mathieu Besançon @matbesancon.xyz · 21/04/2026

Oui, il y a un peu de travail pour les blocs multi-equations, si il n'y a pas d'agencements farfelus de diagrammes ou de Tikzeries, ça ne doit pas être affreux

100

Mathieu Besançon @matbesancon.xyz · 21/04/2026

Si les papiers et donc le contenu sont déjà là, le travail de formatage ne devrait pas être monstrueux non ?

100

Mathieu Besançon @matbesancon.xyz · 20/04/2026

Une des meilleures améliorations récentes du campus

100

Mathieu Besançon @matbesancon.xyz · 03/04/2026

Yeah sure AGI but when will train platforms apply reduction cards correctly on a trip?

020

Mathieu Besançon @matbesancon.xyz · 14/03/2026

La comparaison est douteuse: je ne vois pas la pertinence de mettre sur le même plan des militants exprimant des opinions et des personnes qui ont opéré au sein d'un gouvernement et ont pris des décisions arbitraires sur des programmes de recherche qui les dépassaient

100

Mathieu Besançon @matbesancon.xyz · 09/03/2026

Another missed opportunity after *check notes* the first and second Ukraine wars, Georgia, Irak...

030

Mathieu Besançon @matbesancon.xyz · 09/03/2026

Je pensais avoir un quartier de gauche, et j'ai regardé les affiches de ma rue 😭

000

Mathieu Besançon @matbesancon.xyz · 17/02/2026

Super curious to see the final slide deck if you make it public!

030