Lionel

They're not sociologists / economists. Even if they talked about wealth inequality you should discount everything they'd say.

arxiv version updated — OA paper at www.sciencedirect.com/science/arti...

Read a bunch of his papers on network applications to financial correlation matrices. Great stuff.

📣 Announcing 𝗗𝗶𝗲𝗴𝗼 𝗚𝗮𝗿𝗹𝗮𝘀𝗰𝗵𝗲𝗹𝗹𝗶 as Invited Speaker at NetSci Conference 2026!
Professor at the IMT School for Advanced Studies in Lucca & at Leiden University’s Lorentz Institute for Theoretical Physics, his research combines statistical physics, information theory, & random-graph modeling.

Code is liability. AI allows us to create tremendous amounts of liability very quickly

Peut-on compresser la parole humaine pour les IA sans perdre ses nuances? FocalCodec montre que c’est possible.
En simplifiant la manière dont les tokens sont appris, ce modèle ouvre la voie à des LLM multimodaux plus performants.
À lire ici mila.quebec/fr/article/f...

arsenal often performs well according to stats that end up not really mattering to who wins the trophies.

what kind of delusion is this

great. can I ask what did you ask it to do?

The software complexity of nations Despite the growing importance of the digital sector, research on economic complexity and its implications continues to rely mostly on administrative …

🚨 New paper in Research Policy: The software complexity of nations
Economic complexity through open-source software contributions — and what this reveals about countries beyond administrative data.
www.sciencedirect.com/science/arti...

We’re thrilled to welcome 𝗠𝗲𝗹𝗮𝗻𝗶𝗲 𝗪𝗲𝗯𝗲𝗿 as a School Speaker at NetSci Conference 2026! An Assistant Professor at Harvard, she leads the Geometric Machine Learning Group, developing efficient, theory-backed ML methods.
Meet her & all the speakers 👉 www.netsci2026.com/speakers

a frog wearing a purple collar and a tag ALT: a frog wearing a purple collar and a tag

All glory to the hypnotoad!

That's one way to corner a market and gain first mover advantage.

I've stuck my head in the sand a lot when it comes to supervised learning, neural networks and all of that but turns out the people at UCL were working on very cool extension of Information Filtering Networks (IFNs) to neural network architectures.

Your typical knn graph optimizes finding top k nearest neighbors, while the TMFG requires building triangles and has a planarity condition. If your use case benefits from accounting for these local dynamics in the graph IFNs are probably the way to go.

For a long time I was puzzled as to why use Minimum Spanning Trees, and other Planar Maximally Filtered Graphs but I'm starting to think that it's not for their ability to capture nearest neighbor but more in the topological features of the graph construction

Information filtering networks: theoretical foundations, generative algorithms, and real-world applications Information filtering networks: theoretical foundations, generative algorithms, and real-world applications, Aste, Tomaso

(re)-reading "Information filtering networks: theoretical foundations, generative algorithms, and real-world applications" iopscience.iop.org/article/10.1...

Disentangling the Boltzmann brain hypothesis: Memory, entropy, and time We trust our memories because they feel natural, and we trust time because it seems to flow in only one direction. Physics, however, allows for stranger possibilities that challenge our intuition. In...

In a new paper, SFI researchers examine the Boltzmann brain, a thought experiment that raises fundamental questions about memory, entropy, and time. The work clarifies how arguments for or against these ideas depend on assumptions about the past that are not fixed by physical laws alone.

Tried @repl.it's agentic coding system: Agent disappeared for nearly an hour without a *single update* or *any code* to show, then halts with "You've reached your usage limit." 🤷‍♂️👎
PS- Says I'm locked out til *tomorrow*? rofl

You know how in french we say "tomber enceinte"(fall pregnant) cause it must have happened by accident 🤷🏿‍♂️

isn't it linked to ovulation cycle?

Sending job applications doesn't seem to work very much these days. Even after meeting up with people at conferences, applications can still slip through (i.e. rejection w/o phone screen), you definitely need to go the extra mile (i.e. reaching out to recruiters directly) or else nothing happens.

Missed this because I left this place a while ago but glad to see you're still in the game!

saw this a couple months ago, are you cherry picking or is this truly still not solvable at this point?

truly frightening stuff

easier to get funded

GitHub - thosvarley/syntropy: A python package for information-theoretic analysis of discrete and continuous data. A python package for information-theoretic analysis of discrete and continuous data. - thosvarley/syntropy

Introducing a new scientific computing library: syntropy.
Syntropy is a comprehensive package for information theory, aimed at both theoreticians and data analysts working on discrete, continuous, and mixed data.
1/N
github.com/thosvarley/s...

We’re excited to welcome Mirta Galesic as our next OSoMe Awesome Speaker!

🗓 Wednesday, Jan 28, 2026
⏰ 12:00–1:00 PM ET
🎙 Dynamics of belief networks

Register: iu.zoom.us/meeting/regi...

Guessing Games In my last post I discussed the case of a Polymarket trader who placed a series of well-timed bets on the ouster of Nicolás Maduro and made a profit of over $400,000 (on a $32,000 investment) in just ...

New post on prediction markets, spoofing, and forecasting accuracy:

open.substack.com/pub/rajivset...