Abdoulaye Diack

Qualcomm’s Acquisition of Arduino Creates a New Vibe—AI and Signal Processing on the UNO Q Qualcomm buys Arduino—and a Dragonwing MPU and STMicro MCU now creates the latest board, Arduino UNO Q, with development support for Linux OS and vibe-coded AI solutions.

New vibe
www.electronicdesign.com/technologies...

A Nobel medal

BREAKING: The Nobel Prize in Physics has been awarded to John Clarke, Michel H. Devoret, and John M. Martinis "for the discovery of macroscopic quantum mechanical tunnelling and energy quantisation in an electric circuit"

Stay tuned for more.

"Lapse"
for(float i,z,d,h;i++<5e1;o+=vec4(3,z,i,1)/d)
{vec3 p=z*normalize(FC.rgb*2.-r.xyy),a;a.y++;p.z+=7.;a=mix(dot(a,p)*a,p,sin(h=length(p)-t))+cos(h)*cross(a,p);
for(d=0.;d++<9.;a+=sin(round(a*d)-t).zxy/d);z+=d=.1*length(a.xz);}
o=tanh(o/1e4);

To keep the tradition, the lineup is 🔥🔥🔥

Learning without training: The implicit dynamics of in-context learning One of the most striking features of Large Language Models (LLM) is their ability to learn in context. Namely at inference time an LLM is able to learn new patterns without any additional weight updat...

Interesting paper suggesting a mechanism for why in-context learning happens in LLMs.

They show that LLMs implicitly apply an internal low-rank weight update adjusted by the context. It’s cheap (due to the low-rank) but effective for adapting the model’s behavior.

#MLSky

arxiv.org/abs/2507.16003

Picture of a digitised index card.

Card catalogues aren't just a relic of the past - many institutions still rely on them because full migration is too expensive. VLMs could help change that.

I uploaded two new @hf.co datasets (~470K cards) for training/evaluating models to extract structured metadata from catalogue cards.

ModernVBERT: Towards Smaller Visual Document Retrievers Multimodal embedding models are gaining prevalence, notably for document retrieval as efficient alternatives to text-only pipelines. These models are typically built by finetuning large vision-languag...

ModernVBERT: Towards Smaller Visual Document Retrievers

Proposes a 250M-param vision-language encoder using bidirectional attention and MLM-based alignment that outperforms models up to 10x larger on document retrieval tasks.

📝 arxiv.org/abs/2510.01149
👨🏽‍💻 huggingface.co/ModernVBERT

China’s most infamous ghost town is now training ground for driverless trucks The empty boulevards that once symbolized the mainland’s property bust now serve as testing grounds for hundreds of autonomous vehicles.

New from @kinlinglo.bsky.social: A look at how the ghost town that once symbolized China’s property bust became an ideal place to train self-driving vehicles.

We’re scaling quantum computing even faster with Atlantic Quantum. Google Quantum AI was founded in 2012 and our mission today remains the same — build quantum computers that can solve otherwise unsolvable problems. We’re making steady …

blog.google/technology/r...

E

The Turing Test and our shifting conceptions of intelligence “Can machines think?” So asked Alan Turing in his 1950 paper, “Computing Machinery and Intelligence.” Turing quickly noted that, given the difficulty of defining thinking, the question is “too meaning...

In honor of the 75th anniversary of the Turing test, I'm re-upping my short essay, "The Turing Test and our shifting conceptions of intelligence".

www.science.org/doi/10.1126/...

Percepta | A General Catalyst Transformation Company Transforming critical institutions using applied AI. Let's harness the frontier.

We're finally out of stealth: percepta.ai
We're a research / engineering team working together in industries like health and logistics to ship ML tools that drastically improve productivity. If you're interested in ML and RL work that matters, come join us 😀

🚀 New Model Launch: Granite 4.0

IBM's Granite 4.0 is the latest open-source SLM, built for fast inference, long-context understanding (tested on up to 128K tokens), and cost-efficient deployments. Multiple model sizes provide flexibility for different hardware and use cases.

ModernBERT goes MULTILINGUAL!

One of the most requested models I've seen, @jhuclsp.bsky.social has trained state-of-the-art massively multilingual encoders using the ModernBERT architecture: mmBERT.

Stronger than an existing models at their sizes, while also much faster!

Details in 🧵

Smart move

⏩ FastForward ⏩ A new model for efficient visual relocalization. Accurate camera poses without building structured 3D maps.

nianticspatial.github.io/fastforward/

Work by @axelbarroso.bsky.social, Tommaso Cavallari and Victor Adrian Prisacariu.

Nobody even works here. It's just me and 29 advanced AIs.

Evaluating seasonal forecast improvements over the past two decades We have analysed the performance of operational seasonal forecasting models since their inception. Clear improvements are measured through the different model eras, particularly in the Tropics. For t....

Are you interested in seasonal forecasts and how they have advanced over the past two decades?
Chris O’Reilly (Univ. of Reading) led a new paper on exactly this: rmets.onlinelibrary.wiley.com/doi/10.1002/...

That's an epic multilingual corpus with a trove of knowledge on data preparation! Props on releasing the cleaning tools as well like the custom bpe.

An ELEGOO UNO R3 Project Starter Kit

My 9y old likes playing with this Arduino starter kit (UNO R3). 3D printing your own robot arm and getting it to work also sounds like a fun learning experience, maybe we'll try that next github.com/huggingface/...

5-Day AI Agents Intensive Course with Google Join our 5-day AI Agents Intensive Course with Google, November 10–14, to learn how to build, evaluate, and deploy agents.

We're hosting the 5-Day AI Agents Intensive course with Google on November 10 - 14. We’d love for you to join!

This no-cost course is designed to help you explore the foundations and practical applications of AI agents. More info and registration: rsvp.withgoogle.com/events/googl...

"Don’t underestimate the effect of a smile" that's a good one, whatever energy you send is what you get back

Cloudflare accidentally DDoS'ed itself in dashboard update gone wrong Three-hour service disruption affected dashboard and APIs but spared network infrastructure

It's a beautiful kind of irony when the world's most famous DDoS protector ends up getting taken down by a single useEffect hook. They protected everyone else from the storm, but forgot to board up their own windows.

Cloudflare accidentally DDoS'ed itself
www.sdxcentral.com/news/cloudfl...

When did Git go from "too different" to the industry standard? 🤔 Linus Torvalds explains the generational shift that changed everything. ✨

Learn more about the history of Git in our full interview with Linus. ⬇️
https://github.blog/open-source/git/git-turns-20-a-qa-with-linus-torvalds/

Jiasen Lu, Liangchen Song, Mingze Xu, Byeongjoo Ahn, Yanjun Wang, Chen Chen, Afshin Dehghan, Yinfei Yang
AToken: A Unified Tokenizer for Vision
https://arxiv.org/abs/2509.14476

Can we use video diffusion to generate 3D scenes?

𝐖𝐨𝐫𝐥𝐝𝐄𝐱𝐩𝐥𝐨𝐫𝐞𝐫 (#SIGGRAPHAsia25) creates fully-navigable scenes via autoregressive video generation.

Text input -> 3DGS scene output & interactive rendering!

🌍http://mschneider456.github.io/world-explorer/
📽️https://youtu.be/N6NJsNyiv6I

Even if disturbed, the textures made by building blocks called neural cellular automata have the ability to heal themselves. By understanding how simple components form complex systems, scientists could theoretically coax bodies to regenerate missing limbs. www.quantamagazine.org/self-assembl...

record 17,058 Marie Curie proposals in 2025 vs. 10,360 in 2024!!! the reason is pretty obvious, but wow that is a huge jump marie-sklodowska-curie-actions.ec.europa.eu/news/msca-po...

GSplats for high resolution image compression
arxiv.org/abs/2407.01866