In dogana con gioielli nascosti dentro un pane
Nell’ambito di un controllo al valico di St. Margrethen, le guardie di confine hanno trovato la presunta refurtiva di un colpo messo a segno nel canton Zurigo
https://www.rsi.ch/s/3301967

The nightline signs of the Viennese public transport are really cute

# Akkoma # Statistics 2025-11-24 05:00 CET
Number of active instances: 652
Number of users: 21 211
Number of statuses: 8 681 678
Number of users last 4h: -53
Number of statuses last 4h: 849

# Fediverse

RT @burkov
This paper shows how to beat models with 671 billion parameters using just 7 million parameters on hard puzzle tasks. On the ARC-AGI reasoning benchmark, this tiny model achieves 45% accuracy compared to 15.8% for DeepSeek R1 and 34.5% for OpenAI's o3-mini. On extremely difficult Sudoku puzzles with only 1,000 training examples, it reaches 87% accuracy where the previous best method managed 55%.

The method treats problem-solving as iterative refinement. The network maintains a current answer and a separate "reasoning" state, then repeatedly updates both: reason about the problem, propose a better answer, reason again, improve further. A Sudoku solver might start with a partially filled grid, update some cells, then reason again about the new constraints—gradually correcting mistakes across up to 16 refinement cycles.

The core insight is that large models fail because they generate answers in one pass, so early mistakes cascade. Breaking the problem into iterative steps, even with far less computational power, proves more effective. Training takes hours on a single GPU instead of weeks on clusters, and the model fits in 28MB instead of terabytes.

Let the article talk to you on ChapterPal: https://chapterpal.com/s/e2942994/less-is-more-recursive-reasoning-with-tiny-networks

Download the PDF: https://arxiv.org/pdf/2510.04871

"The withdrawal of the US from UNESCO and US legislation being increasingly at odds with Open Science values raises concerns regarding Open Science infrastructure in the US. While much is still unclear, our authors argue that the implications for Open Science could be profound."
https://www.leidenmadtrics.nl/articles/the-withdrawal-of-the-us-from-unesco-what-does-this-mean-for-open-science

RE: https://floss.social/@alcinnz/115599587720338043

“Stop rushing to fix bugs for people without a support contract. Patch security flaws at a more leisurely pace unless someone is willing to pay for greater urgency. Take your time and enjoy your hobby more, since that is what unpaid software maintenance is.”

Are you in Amsterdam? I'll be giving a keynote next week! "We Can Change the Defaults: Building Networks of Consent and Spaces of Joy in the Ruins of Social Media" https://spui25.nl/programma/we-can-change-the-defaults-building-networks-of-consent-and-spaces-of-joy-in-the-ruins-of-social-media/

How far will it go this time?
This post from #Mastodon can reach the whole #Fediverse, #Bluesky & more.
That's #Sharkey, #Misskey, #Pixelfed, #Mbin, #Lemmy, #Friendica, #Hometown, #Akkoma…

If you see it, please share it.

#SocialMedia @fediverse

First results 👇🏼

Le piège des newsletters payantes se referme : https://danslesalgorithmes.net/stream/le-piege-des-newsletters-payantes-se-referme/

I search: garnet ant ring
Search "corrects" it: garnet and ring

I try: garnet "ant" ring
Search "corrects" it: garnet "and" ring

I try: garnet "ant" ant ring
Search "corrects" it: garnet "aunt" and ring

I try: garnet "ant" ant insect ant ring
Search "corrects" it: garnet "aunt" and insect ant ring

This is a wonderful technology that makes search so much easier. This was on etsy.

New in our electronic #collection:

➡️ Human Rights in the Digital Domain. Core Questions

🔗 https://doi.org/10.1017/9781009606295

#HumanRights #SocialMedia #digitalization #legislation #privacy #governance #ethics #education #OpenAccess #OpenScience #books #bookstodon

I hate ‘self-taught’. You’re not self-taught, you’re community taught. Self-taught is erasing all the work that allowed you to learn from freely available materials and people.

It’s a poor phrase for the idea that you are degreeless. And there’s nothing wrong with skipping formal education.

"Linux is too complicated for anyone"

Windows users:

#Linux #Microsoft #FuckAI #Windows11

the amount of detail that went into this https://youtu.be/_8QLA43Y-uM

Just in case: I don't want to needlessly trashtalk a free, volonteer-operated festival. You're not even remotely as bad as Angoulème!

FOSDEM is great! It's just me me having issues with large crowds :D

Thanks for all the work you put into organizing this! ♥️

Il genetista Angelo Abbondandolo spiega cosa c'è di sbagliato in dottrine quali eugenetica e #darwinismo sociale dal punto di vista della #scienza e quanto sono distanti dalla teoria evolutiva di #Darwin.
Per Nessun Dogma, progetto editoriale Uaar 👇
https://nessundogma.it/libro/i-figli-illegittimi-di-darwin/