iiinova — where your digital self keeps evolving

Carles Sáez @csaezmath.bsky.social · 22h

I'm so glad I leave academic math... Now I can just learn whatever I want on my own, and using whatever tools I like

Carles Sáez @csaezmath.bsky.social · 02/10/2026

I've never been to Copenhagen, but I once went to a summer school in Paris where the accommodation was arranged by the organization and the bathroom was exactly like this. Horrible experience

010

Reposted by Carles Sáez

Sridhar Ramesh @radishharmers.bsky.social · 30/09/2026

As ever, it was a land dispute. One pilot wanted to and the other didn't.

1177859

Carles Sáez @csaezmath.bsky.social · 23/09/2026

I guess this is one of those usa things, but congrats!

110

Reposted by Carles Sáez

Emily M. Bender @emilymbender.bsky.social · 20/08/2026

This sign is a great example for helping to explain why chatbots present such a potent illusion. A probably not short thread 🧵>>

Photo of a sign with a doorbell attached. The printed sign reads:

CURRENT GUGGENHEIM ACCES BY HUSKY CARD ONLY
[Logos for Amazon, FedEx, UPS]
DELIVERIES/VISITORS
PLEASE RING THIS DOORBELL

Delivery hours M-F 9:00am-(obscured)
University of Washington

Taped next to the doorbell is a hand written note that reads:
yes! it rang! Pelase wait 30 seconds before pressing again.
thanks!

10461190

Reposted by Carles Sáez

Michael Perrinow @michaelperrinow.bsky.social · 19/08/2026

Los billonarios vieron Mad Max Fury Road y decidieron que querían ser Immortan Joe.

813340

Reposted by Carles Sáez

Sridhar Ramesh @radishharmers.bsky.social · 09/07/2026

"Why don't you do something useful instead of comedy/abstract math/whatever?". The world isn't just for surgeons. Medicine is valuable not as an end in itself, but as a means to aid people enjoy their lives, enjoyment which includes doing whatever they want, such as comedy/abstract math/whatever.

618712

Reposted by Carles Sáez

Sridhar Ramesh @radishharmers.bsky.social · 06/08/2026

I'm not interested in math because it's useful. I'm interested in math because it's interesting.

121577

Reposted by Carles Sáez

Sridhar Ramesh @radishharmers.bsky.social · 06/08/2026

ChatGPT isn't sentient, according to the widely-accepted argument of John Searle. It speaks English, whereas a sentient being would speak Chinese. DeepSeek has been a breakthrough in this regard.

941343

Carles Sáez @csaezmath.bsky.social · 06/08/2026

And thus an ultrafinitist is born

010

Reposted by Carles Sáez

singular locus sarah @scgriffith.bsky.social · 05/08/2026

everything else aside, what kind of freak loves work and celebrates work

1212

Reposted by Carles Sáez

Andrés Trasado @andrestrasado.bsky.social · 05/08/2026

¿Os acordáis de Llados y sus cursos y los mindundis que pagaban esos cursos para hacer luego ellos la copia Temu de esos cursos (mentalidad tiburón con problemas)? ¿Y luego los mindundis normales que solo pagaban los cursos porque, Bro, eso es vida? Pues el ecosistema de la IA ya está en ese punto

59715

Carles Sáez @csaezmath.bsky.social · 30/07/2026

4. (Some) physical books are art pieces on their own (cover design, choice of book format, ...). Destroying them is like destroying a painting.

020

Carles Sáez @csaezmath.bsky.social · 30/07/2026

Why would we need an imperfect intermediary? I can only think of bad faith arguments for that. 3. Books are not only "information"!!! A lot of them (novels, etc.) are artistic works trying to convey beauty or other emotions but not directly "information" 2/

150

Carles Sáez @csaezmath.bsky.social · 30/07/2026

There are SO MANY things wrong with that... 1. An AI model is a lossy compression mechanism, so in general it'll be impossible to recover the exact document (and even if it's possible, you'll never be sure) 2. The delivery of information was and has always been directed to HUMANS. 1/

110

Carles Sáez @csaezmath.bsky.social · 30/07/2026

"Once that information has been extracted and encoded into an AI model, the delivery mechanism has served its purpose" Statements dreamed up by the utterly deranged

142

Reposted by Carles Sáez

Andrés Trasado @andrestrasado.bsky.social · 08/07/2026

Os juro que había escrito este bluit sin saber nada de esto

18514176

Reposted by Carles Sáez

Garth Marenghi's Catbus @catbus.phd · 26/06/2026

ok this one got me

reflecting pool full of green slime and barrels and you're carrying a crowbar in the style of Half Life

21112892437

Reposted by Carles Sáez

John C. Baez @johncarlosbaez.bsky.social · 12/06/2026

If I could, I would develop a whole theory linking differential geometry and logic just so I could write a textbook called Truth Tables and Lie Algebras.

68113

Carles Sáez @csaezmath.bsky.social · 31/05/2026

An example of why basic science is essential:

020

Reposted by Carles Sáez

Sam Power @spmontecarlo.bsky.social · 30/05/2026

With friends at the University of Warwick (in particular, Rocco Caprio and @adriencorenflos.bsky.social), we've recently arXived some work (arxiv.org/abs/2605.30253) on a method for approximate inference known as "Coordinate Ascent Variational Inference", or "CAVI" for short. Let me explain:

2319

Carles Sáez @csaezmath.bsky.social · 29/05/2026

Unpopular(?) opinion: I agree with the university that retired people should not give examinable courses. This course should be given by someone new who has a contract with the uni

130

Carles Sáez @csaezmath.bsky.social · 27/05/2026

No es mi experiencia en absoluto... Por desgracia mi padre ha pasado muuuuucho tiempo en el hospital los últimos años, y el trato que he visto por parte del personal sanitario siempre ha sido impecable

110

Reposted by Carles Sáez

Dr David S Anderson @dsaarchaeology.bsky.social · 27/05/2026

As an anthropologist, I endorse this message!

273207196673

Carles Sáez @csaezmath.bsky.social · 26/05/2026

Bad times for science, particularly for all that science which doesn't immediately help very rich people to get even richer

021

Reposted by Carles Sáez

Barbara Fantechi @barbarafantechi.bsky.social · 26/05/2026

How many research projects can you defund on the basis that they have "been determined to be of absolutely no value to the [country] taxpayer, and therefore it was being specifically terminated"? #mathsky

4146

Reposted by Carles Sáez

Fandogamia Editorial @fandogamia.com · 26/05/2026

ESTO OCURRE JUSTO LA SEMANA EN LA QUE SALE ESTE TEBEO A LA VENTA???? no puede serrrrrrrrrr

Portada de BEATA Y CALISTO, ABOGADOS DE CRISTO.

7211110

Reposted by Carles Sáez

Coach Finstock @coachfinstock.bsky.social · 25/05/2026

Converting to Catholicism so I can tell my bosses using unregulated AI is against my religion and I will sue them like that cake lady who hated gay people if they make me use it

104112172681

Reposted by Carles Sáez

Nvidie @dervine7.bsky.social · 25/05/2026

updating your committee on your dissertation progress

131015139

Reposted by Carles Sáez

Barbara Fantechi @barbarafantechi.bsky.social · 16/05/2026

So the best surfaces (that's how we call 4 manifolds here) are those of general type. The super best have the smallest possible invariants, and are called Godeaux. Godeux surfaces have finite cyclic fundamental groups of order at most 5. 1/

152

Reposted by Carles Sáez

JudithFlanders @judithflanders.co.uk · 14/05/2026

Really, I can’t see any problem with historians being replaced by AI. It’ll all be fi— . Oh.

0147

Reposted by Carles Sáez

Vicent Navarro @vicentmath.bsky.social · 03/05/2026

Hola!! Soy Vicent :) algunos me conoceréis de Twitter. Por desgracia perdí la cuenta porque me la hackearon y me la tumbaron. Me ayudáis a recuperar seguidores? :)

032

Carles Sáez @csaezmath.bsky.social · 30/04/2026

Unpopular(?) opinion: la demografía mola

100

Reposted by Carles Sáez

Eva Cid @evcdmr.bsky.social · 22/04/2026

Se habla mal y se escribe mal y parte de la culpa radica en consumir acríticamente movidas gringas sin traducir o traducidas mal. Leed cosas en vuestra lengua materna escritas por personas que lleven ya un tiempo muertas, parece que no pero es fundamental.

18549206

Carles Sáez @csaezmath.bsky.social · 15/04/2026

Aquí os espero, en el INE no se vive mal

110

Carles Sáez @csaezmath.bsky.social · 13/04/2026

I agree, I was just paraphrasing the original bluit

010

Carles Sáez @csaezmath.bsky.social · 13/04/2026

* Modules are a special case of sets because every module is just a set with extra structure * Modules are a generalization of sets because every set is an F_1-module

140

Carles Sáez @csaezmath.bsky.social · 12/04/2026

Well, it's a great book, so it can be a good idea! But, since the book is 400 years old, it will be quite difficult to read even if you achieve a good level of modern Spanish (Spanish has evolved quite a bit since the XVIIth century!)

110

Carles Sáez @csaezmath.bsky.social · 11/04/2026

Yo aluciné en colores cuando me leí el Quijote. ¿Humor absurdo, sátira, crítica social, ...? Está todo allí. ¿Todos esos juegos metanarrativos que parecen tan modernos? Cervantes ya los hizo antes en el Quijote.

110

Reposted by Carles Sáez

Fleibur @drfleibur.bsky.social · 07/04/2026

Finalmenchi. Finalmenchi a mis 34 años me voy a leer El Señor de los Anillos de la gente que le dice a Reverte "Don Arturo".

33783192

Reposted by Carles Sáez

Sridhar Ramesh @radishharmers.bsky.social · 29/03/2026

The LLM takeover has made everything in computing so loosey-goosey. "Maybe it will work. Empirically, there's a good probability...". This is not what interested me about computing. Of course, sometimes it does work, which is magic. I am just a curmudgeon who doesn't like magic.

626315

Reposted by Carles Sáez

Sam Power @spmontecarlo.bsky.social · 21/03/2026

Some more-or-less equivalent facts about probabilities: 1) for random variables of a given mean and variance, the Chebyshev-Cantelli inequality provides an upper bound on tail probabilities 2) for random variables of a given variance, the mean and quantile (at a given level) can't differ too much

1111

Reposted by Carles Sáez

ꭓle Ormsby 𓆙 @e-infinity.space · 21/03/2026

Gonna get this printed on a hat and wear it to conferences. 🧮

$Born to Prove S \simeq K(FinSet) Forced to Compute [mathematical expression for the Novikov spectral sequence]$

2324