Gpt/LLM/GenAI
Ce qui est entré en base, sans jugement.
3065posts
36auteurs
518liens
1470 hde recul
38membres
Aucun recoupement sur 24 h pleines : ni un lien ni un tweet n'est relayé par
deux membres (2 liens, 13 tweets relayés). Le recul est
suffisant, donc cette fois le constat porte : ces comptes diffusent au lieu de
converger, et c'est la composition de la liste qu'il faut revoir — pas la patience
qu'il faut allonger.
Nous Research a relayé @Teknium
We surpassed 1000 community made plugins today!
Check them all out here:
Nous Research a relayé @Teknium
We surpassed 1000 community made plugins today!
Check them all out here:
Andrej Karpathy a cité @karpathy
Earlier this week Tristan published his top YouTube channels:
Which looked excellent so clearly I had to turn it into TristanTube:
47K super nerdy videos +TV mode for laid back watching. (One-shotted with Claude Code in 3 minutes)
Nous Research a cité @Meituan_LongCat
LongCat-2.5, the new multimodal model from @Meituan_LongCat, is free in Nous Portal for one week.
Try it free today at
P
Yann LeCun a relayé @PolymathicAI
Interested in building what’s next in AI for Science? Join us at PolymathicAI in New York (NYU + Flatiron).
We are looking for: Interns, Postdoctoral Researchers, Research Scientists and Research Engineers.
Submit your interest here:
Yann LeCun a relayé @KenRoth
In the United States, Brazil, Hungary, Serbia and Israel, "voters are turning out in droves to cut autocrats down to size. In country after country stricken by democratic decay, overwhelming turnout in elections is making a difference."
C
OpenAI Developers a relayé @craigsdennis
Love building with @OpenAIDevs 🤝 @TwilioDevs
Check out this quick tutorial of how to build telephone based apps that remember you using GPT Live and Twilio Agent Connect
G
Yann LeCun a relayé @GaoyueZhou
Introducing ID-WM: a hierarchical world model that decouples "where to go" from "how to get there." Finally, we can plan over hundreds to thousands of steps with WMs.
It beats the strongest baseline by 38.5%, single-level WMs by 135%. #NeurIPS2026 🧵
N
Hugging Face a relayé @NielsRogge
We @huggingface reproduced this paper!!
It's built with ML Intern, an open-source harness to do AI research, train models, and more
Model:
Demo:
N
Yann LeCun a relayé @Numerama
Dans une étude mise en ligne le 29 octobre, Yann LeCun et Randall Balestriero affirment démontrer qu'un petit modèle d'IA bien nourri se fait une meilleure image du monde qu'un LLM 660 fois plus gros.
Harrison Chase a relayé @sydneyrunkle
we're hosting an event on thursday at the @LangChain office with @allietheicon from @typesafeai to chat about all things jev!
come learn about
- where jev fits in the agent loop
- why jev is a "system one" model
- jev as a judge for evals
L
Hugging Face a relayé @LoubnaBenAllal1
Introducing 🧬 FineDNA: the best dataset to teach any language model DNA.
Training Qwen3-14B on it gives the strongest open model.
300B+ tokens of genomic sequences, papers and web pages mixing DNA and text, plus 200k SFT and 98k RL samples.
L
Hugging Face a relayé @Llama_app
Our picks from last week's model drops 🦙
→ LFM3 by Liquid AI: small models for text, image and speech, with d1 decision models
→ Mistral Large 4: large, 1M context frontier
→ Flint by Recursive: mid-sized, built for coding agents
Get them at
Harrison Chase a relayé @LangChain
🌉 Interrupt, The Agent Conference by LangChain, returns to San Francisco March 24 + 25
🎟️ Early Bird tickets available until Dec 7th
Info + RSVP:
Yann LeCun a relayé @KenRoth
The Trump administration is openly meddling in European democracies to promote far-right political parties that support Putin and to advance the commercial interests of Trump allies.
R
hardmaru a relayé @RobertTLange
Had tons of fun last week at the @SakanaAILabs symposium 🗼
Blog:
I first met Jürgen in Vienna at ICML 2024. We had just finished working on DiscoPOP and were pushing hard towards the AI Scientist v1. Fast forward to October 2026, and we are moving
T
OpenAI Developers a relayé @greatony
I hate the huge and heavy Daikin A/C remote, so I let Codex to write a firmware to convert LilyGO's T-Display to the remote and open sourced it:
Harrison Chase a relayé @LangChain
There's still time to RSVP for our upcoming SF meetup with @typesafeai
Join @sydneyrunkle, @allietheicon, and Seán Roche at LangChain HQ for a night of talks on building with decision models and using them to evaluate agents.
Spaces are limited!
Hugging Face a relayé @llama_cpp
Our picks from this week's model releases 🦙
→ LFM3 by Liquid AI: small models for text, image and speech, with d1 decision models
→ Mistral Large 4: large, 1M context frontier
→ Flint by Recursive: mid-sized, built for coding agents
Get them at
S
Stability AI a relayé @stableaudio
Thanks to everyone who stopped by! If you didn't catch us there, we'd love to chat:
S
Stability AI a relayé @stableaudio
Thanks to everyone who stopped by! If you didn't catch us there, we'd love to chat:
hardmaru a relayé @SakanaAILabs
Sakana AI won 1st place in the company category at the “Recursive Self Improvement in Mathematics” Hack (OpenMath 2026), held with @stephen_wolfram at Harvard Innovation Labs.
Competition:
Results:
Drawing mainly on our Sakana
E
Hugging Face a relayé @EmmaScharfmann
So much happened in AI for Science on the 🤗 Hub this month!
I summarized what these new models and datasets are about, to make them accessible to anyone curious about AI for Science 🚀
Hugging Face a relayé @SergioPaniego
how do you post-train an agent?
we laid it out as a 5-step curriculum, from SFT to RL through a real harness, with livestreams, code and docs for each step
hardmaru a relayé @SakanaAILabs
企業のお客様にSakana AIのAPIを安心してお使いいただけるよう、「Sakana API Platform」を開発しました。本日より、エンタープライズ向けのアーリーアクセスを開始します。
Sakana AIのモデルを、組織での管理機能とあわせてAPIでご利用いただけます。
Harrison Chase a relayé @DeepLearningAI
Build reliable agents for long-running tasks.
Loop Engineering with @LangChain Deep Agents: add loops that check output, trigger runs on events, and improve the harness from traces across runs. Taught by @sydneyrunkle and @EAGar123.
Enroll for free:
N
Harrison Chase a relayé @ndrezn
nearly every proactive agent you've seen is a cron job with a prompt
memory that "dreams" overnight, agents that ping you first. both are scheduled runs on a 50+ year old OS idea
how to build one with Managed Deep Agents ⬇️
OpenAI Developers a relayé @derrickcchoi
Big news with @Oracle! Bringing Oracle apps and data into ChatGPT, with new offerings powered by OpenAI.
US customers can now buy API, Codex & ChatGPT Work via Oracle Marketplace using eligible credits.
Also 130K ChatGPT + 95K+ Codex users now!
田
hardmaru a relayé @ChiekoTazuke
Sakana AIでGTMヘッドを務める Hidekazu Kakishima さんの記事を公開しました。ソラコム、AWS、Skydioなどでご活躍されたHideさんがSakana AIのプロダクトを強力に展開していきます！
Hideさんと一緒にSakana AIのプロダクトを展開していきたい方は応募してください😊🐟
R
Harrison Chase a relayé @RazBhimani
@hwchase17 back at it with an amazing talk on self improving agents. Must watch here:
@LangChain soon will be your goto platform for framework, context management, observability, and evals - completely model agnostic.
K
OpenAI a relayé @KITP_UCSB
Statement of Yang-Mills Working Group, a scientific collaboration to interpret AI result related to the existence of quantum chromodynamics & completions of the Standard Model, including a subproblem proving the existence of the Quantum Yang-Mills theory.
hardmaru a relayé @SakanaAILabs
We’ve published an interview with Sakana AI’s VP of GTM 🐡
In 2026, Sakana AI released Sakana Chat, Marlin, Fugu, Translate, and Namazu, entering the stage of delivering products born from its research to enterprises. Our GTM team launched in Sept 2026.
Nous Research a relayé @Teknium
Introducing the @ElevenLabs SFX plugin, built to continue the work on a suite of Game Dev Studio capabilities for your Hermes Agent!
Hermes made this demo video to see it in action, you can install it right now from the plugins catalog:
The plugin adds
N
OpenAI Developers a relayé @NaveenManwani17
I just shared Codex Pet Badge and how I built it. You can just build things with hardware and Codex. Check it out and give it a vote!
Shout to @OpenAI and @OpenAIDevs for creating this space for Hardware enthusiast
Harrison Chase a relayé @LangChain
How we built Content Studio, an open-source design agent in Slack that makes on-brand videos, graphics and decks, and the harness that powers it.
P
Hugging Face a relayé @pollenrobotics
New on MicroDuck Arena: Tug of War 🦆🪢🦆
Two robot ducks, one rope, a pit in the middle. Drag your rival in and you win.
Train yours in simulation, publish it, and it's in the first cup, live Friday, Nov 6 at 9 AM PT.
N
hardmaru a relayé @NikkeiAsia
Japan's Sakana AI to build homegrown model to cut US, China reliance
Country's top unicorn eyes use in sensitive fields like defense and cybersecurity
hardmaru a relayé @SakanaAILabs
Sakana AIのVP of GTM (Go-To-Market) へのインタビュー記事を公開しました🐡
SakanaAIは2026年、Sakana Chat、Marlin、Fugu、Translate、Namazuを相次いでリリースし、研究から生まれたプロダクトを企業や官公庁に届ける段階に入っています。その販売と拡大を担うGTMチーム
elvis a relayé @omarsar0
After Jev, this is the most exciting System One model.
Cloudflare introduced it last week.
It's great for extraction tasks and complements decision models like Jev.
Learn more in my new tutorial and check out the interactive playground:
D
Sam Altman a relayé @danintheory
We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
hardmaru a relayé @SakanaAILabs
Sakana AI Dinner Meetup #NeurIPS2026 を開催🐟️
12月7日（月）19時から、シドニーにてSakana AIのDinner Meetupを開催します。NeurIPSに参加される方を対象に、食事を楽しみながら交流できる場をご用意します。
※会場の収容人数の都合上、すべての方をご招待できない
D
OpenAI a relayé @danintheory
We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
Hugging Face a relayé @NielsRogge
Papers with Code is now open-source!
Apache 2.0-license.
Let's collaboratively make the best paper interface!
E
Hugging Face a relayé @emmanuel_2m
We just open-sourced 1,300+ 3D models.
All 39 GB are free to download on Hugging Face.
> made with Hitem3D, Meshy, Rodin, Tripo, Claude Opus 5.5 3D and GPT-6 Astra 3D, which all run on Scenario
> includes the input images, prompts and settings that
R
Hugging Face a relayé @RemiFabreRobot
Sorry for the wait, the episode is here!
Live action with real robots instead of animation (episode 4): "Anger"
Enjoy!
As always, the software of Microduck and Reachy Mini is open source:
elvis a cité @omarsar0
After Jev, this is the most exciting System One model.
Cloudflare introduced it last week.
It's great for extraction tasks and complements decision models like Jev.
Learn more in my new tutorial and check out the interactive playground:
hardmaru a cité @SakanaAILabs
Sakana AI will be at #NeurIPS2026 in Sydney!
Signup to attend our dinner party on Dec 7th: 🐙
hardmaru a relayé @SakanaAILabs
Sakana AI Dinner Meetup #NeurIPS2026 🐟️
Signup →
We are excited to host a dinner meetup in Sydney on Monday, December 7th, from 7 PM. We invite NeurIPS participants to join us for an evening of engaging conversations and networking.
※ Due to venue
S
Google DeepMind a relayé @SpringerNature
Springer Nature and @GoogleDeepMind are partnering to co-design an AI peer review assistant, exploring how AI can help reviewers navigate manuscripts more effectively and focus their time on specialist scientific judgement.
Learn more:
Yann LeCun a relayé @KenRoth
While Trump cut taxes for the rich, he tightened access to food stamps—with more than 5 million Americans losing benefits so far—and to Medicaid, the health-care program for the poor.
LlamaIndex 🦙 a relayé @jerryjliu0
OCRia-1 is the best general purpose model for document parsing.
Model card below 🃏
M
Yann LeCun a relayé @MotherJones
The world’s first trillionaire is also a disinformation super-spreader.
Yann LeCun a relayé @KenRoth
Most voters trust that the midterm ballots will be counted properly, but fears mount that Trump and his Republicans will try to overturn the results. The best defense may be a major electoral loss for Republicans, which will be harder to ignore.
J
Harrison Chase a relayé @jess_weng18
I present to you 🥁🥁🥁
Context Engineering: The Prequel (Retrieval)
H
OpenAI Developers a relayé @haydenbleasel
We built 7 new demos to show what you can make with the @OpenAI API.
Practice a language, explore a dataset, build a presentation, or watch AI navigate rush hour.
Each comes with a runnable project you can download, or open in Codex to start remixing!
S
Hugging Face a relayé @Scenario_gg
We just open-sourced 1,300+ 3D models.
That's 39 GB of 3D files, free to download on Hugging Face.
> Hitem3D, Meshy, Rodin, Tripo, Claude Opus 5.5 3D and GPT-6 Astra 3D, all run on Scenario
> every example ships its input image, prompt and exact
Harrison Chase a cité @amal_irgashev
our internal agents team is a lot of fun to work on
and we're hiring for it!
E
Yann LeCun a relayé @EeroSimoncelli
Looking for a PhD program where you can study computational neuroscience? NYU has a fantastic array of researchers covering the field, and we'd love to consider your application:
Yann LeCun a relayé @KenRoth
Thanks to Trump, the United States sticks out as the only major economy where planet-warming pollution increased dramatically last year, according to a new report by U.N. scientists.
Perplexity a relayé @perplexitydevs
You can now run pplx-decider-flash-v1.2 locally on your Mac, without an API key.
Decider for Mac is an app that runs the 4B decision model on device. Ask a question about text or an image and get a probability back, all on your machine.
A
Nous Research a cité @NousResearch
Meshy is offering Hermes users a month of Pro, which includes API access and 1,000 credits, for $1.
After you subscribe and install the plugin, create a key at and paste it into Settings > Plugins > Meshy in Hermes Desktop.
A
Nous Research a relayé @autonomous_labs
🚨 BREAKING NEWS: HERMES JUST ESCAPED 🚨
We just gave Hermes a body, one eye, and two ears.
Nous Research a relayé @Teknium
We've released a good number of plugins as community plugins in our new Hermes Labs Plugins repo and will continue to release more here.
Check it out here:
This repo will contain our experiments that may or may not be maintained, and are encouraged to be
L
Hugging Face a relayé @LoubnaBenAllal1
Introducing Carbon-X: The best open DNA model.
We midtrained a 14B model on DNA and text, and it beats every open language model and matches GPT-6.1 Sol at 1,000x lower cost.
Ask it about a sequence, what a mutation does, or to generate new DNA.
Hugging Face a relayé @multimodalart
just updated the Decision Index to 0.4 🧭
what's new?
🛡️ curbing benchmaxxing: score now exclusively made from private & in-house benchmarks
🔒 stronger private benchmarks on new domains
🏆 224 models evaluated, new top models in all sizes
come play 👉
Hugging Face a relayé @RisingSayak
We just merged the first protein sequence design model RFD4-Proteina in Diffusers 🔥
I hope this pushes the state of open BioML further.
Kudos to @NVIDIAAI and Baker Lab for shipping this!
Check out the docs to know more:
Harrison Chase a cité @LangChain
the traces you already log are training data - keep the runs that did the task well, fine-tune a small model on them, eval on your task
a 9B model beating frontier models makes sense when it's trained on exactly your task
smithtune:
hardmaru a relayé @SakanaAILabs
【採用情報】Product Strategy & Operationsのポジションを公開しました🐟
今回募集するのは特定のプロダクトではなくプロダクト群全体を受け持つProduct Strategy & Operationsの最初の専任メンバーです。Head of Productのパートナーとして以下をご担当いただきます。
Hugging Face a relayé @NielsRogge
Curious to keep up to date with auto-research papers?
Just added a category for it on Papers with Code!
Find it here:
Yann LeCun a relayé @KenRoth
Trump’s longstanding groundless attacks on the integrity of voting systems are crystallizing into what many fear is an attempt to lay the groundwork to challenge the results in November when Republicans are expected to lose big time.
Hugging Face a relayé @mervenoyann
friendly reminder: this is on Monday 🤝
→ learn about speculative decoding to speed-up inference
→ if you're a beginner and not interested, just use that does this for you automatically 🦙
Harrison Chase a relayé @sydneyrunkle
deepagents now supports gemini 4 argon from
@GoogleDeepMind!
it supports 1M input tokens, long decode (up to 1M), and agentic video understanding!
docs:
release blog:
hardmaru a relayé @SakanaAILabs
【採用情報】マーケティングチームの新しいポジションを4つ公開しました🐟️
Enterprise Marketing Manager
Marketing Operations Manager
Product Marketing Manager
Developer Marketing Manager
Hugging Face a relayé @ClementDelangue
Beautiful work from @perplexity_ai @AravSrinivas with weights on @huggingface of course!
N
Yann LeCun a relayé @NYUDataScience
World models can fail when the world changes.
That's why CDS PhD student Ying Wang (@yingwww_), CDS founding director Yann LeCun (@ylecun) & CDS Assistant Professor Mengye Ren (@mengyer) built AdaJEPA to adapt during planning.
Nous Research a relayé @Teknium
Hermes Index has been updated to include GPT 6.1 Sol and several other models!
Check it out here:
H
OpenAI Developers a relayé @HaimantikaM
Submitted the game I built on ModRetro Chromatic with Codex. 🫶
Now I cannot wait to build more hardware and actually but my hardware engg degree into use🤓
J
Hugging Face a relayé @jeffboudier
Tonight is the night. Today is the day AI Builders unite.
I hear MEGADUCK is coming. Will you?
Open Together, 6pm on 10/16 at The Midway, SF. Free.
Yann LeCun a relayé @KenRoth
Trump calls Ukraine's attacks on Russia's oil refineries a "disgrace," even though they deprive the Kremlin of financing for its attacks on Ukraine, but the warped Trump seems to have no problem with Putin's war-crime attacks on Ukrainian civilians.
T
A
Google DeepMind a relayé @ancadianadragan
We did a lot to understand Gemini 4's capabilities and propensities, and to improve our mitigations for this launch -- you can get a tl;dr here and dive into the report
J
J
Emad a relayé @joecroskey
I built Harmonic, a free music visualizer that runs entirely in your browser.
Drop in a song, loop or MIDI. Create melodies, fly through 3D cities, and export visuals with your music in landscape, square or vertical.
W
Hugging Face a cité @reflection_ai
LIVE STREAM in minutes! Hear from the Reflection team about how they built Beam.
Stream:
A
Yann LeCun a relayé @arthurmensch
Should Europe leave frontier AI to the Americans and the Chinese?
The idea has gained ground in Europe. It is a fundamental mistake.
Build or surrender: for now, the choice is still ours.
Full article below.
Jeremy Howard a relayé @mustafasuleyman
We now have state-of-the-art pre-trained base models in the 3T-parameter weight class, built entirely in-house at MAI.
Post-training results coming soon -- stay tuned!
Lots of details in our technical blog post.
Hugging Face a relayé @mervenoyann
Meta is releasing Muse Spark Open as the most-capable US open-weight model.. 🥁 and it's an agent focused MoE that takes all modalities 🔥
Welcoming the model to Hub with day-0 support in transformers and llama.cpp 🤗
Y
Yann LeCun a relayé @yarotrof
Reuters: The order to throttle Starshield/Starlink for Ukraine came from the White House.
C
N
OpenAI Developers a relayé @nickbaumann_
My dot sent 60 messages to movers over several days, filled out quote forms, and handled the back-and-forth.
I gave it my dates, what I needed moved, and permission to handle the conversations.
Wrote up how it went, and how to give yours a job like this.
S
Stability AI a relayé @stableaudio
We’re back with @MusicHackspace, and this time we’re in Amsterdam for @ADE_NL.
We’re handing you Stable Audio 3.0 and seeing what you do with it.
The challenge is simple. Build a tool other producers can use.
Get started with Stable Audio 👉
OpenAI Developers a répondu à @OpenAIDevs
Leaving here so you can learn to create, publish, and share your first Site.
Hugging Face a relayé @mervenoyann
Meta is releasing Muse Spark Open as the most-capable US open-weight model.. 🥁 and it's an agent focused MoE that takes all modalities 🔥
Welcoming the model to Hub with day-0 support in transformers and llama.cpp 🤗
Harrison Chase a relayé @caspar_br
Managed Deep Agents can read the files people share and send files back where the conversation already is. It feels like magic
Shoutout to @SlackHQ, their files API made this a joy to build
hardmaru a relayé @SakanaAILabs
The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery
Blog:
GitHub:
arXiv:
Since 2024, we’ve used The AI Scientist as a way to test how far AI alone can take scientific
Hugging Face a relayé @RisingSayak
It's time. It's fucking time!
Flux.3 Image Klein is out and we bring you full support for it in Diffusers.
Catch all the details in our post:
hardmaru a cité @SakanaAILabs
The AI Scientist v3 is out: The Agentic Scientist.
GitHub:
Blog:
arXiv:
The Agentic Scientist is the 3rd iteration of our automated AI Scientist research system, and the shift this time is from “one agent
G
Google DeepMind a relayé @googlecloud
Explore the latest advances in native audio models and the challenges of building better real-time voice experiences with @GoogleDeepMind + @SierraPlatform.
Learn about conversational latency, multilingual code-switching, agentic voice tasks, and more →
Nous Research a relayé @Teknium
Check out the new official plugin I created for Hermes Agent that includes the REA MCP + a Skill for how to use it way more effectively!
Mira Murati a relayé @hannahnth
BTW tinker is hiring researchers & infra experts!
research:
systems:
T
Mira Murati a relayé @tinkerapi
Dimuon is a new optimizer we built for LoRA fine-tuning. In our SFT tests, it often reaches lower loss in fewer steps than AdamW. We're still experimenting with Dimuon; you can Tinker with it today and share your feedback!
See our optimizers guide at
H
Jason Wei a relayé @hwchung27
In a lecture I gave last year (), I shared a mental model for AI: the convex hull of humanity’s collective knowledge.
Human expertise is highly specialized and fragmented across people and fields. I hypothesized that there might be a huge overhang of new
A
hardmaru a relayé @SakanaAILabs
Introducing “My Fish” in Sakana Chat.
English Blog:
All your memories in one place, your history on a timeline, and the connections between them mapped out.
See what Sakana Chat has learned about you: 🐟
hardmaru a relayé @SakanaAILabs
Sakana Chatに新機能を追加。会話で育ったメモリーが、見えるようになりました。
新機能「My Fish」で、覚えていることを一覧に。履歴をタイムラインに。メモリー同士のつながりをマップに。Sakana Chatがあなたについて学んだことを、見てみてください。
Try Chat:
Hugging Face a relayé @allen_ai
We're co-hosting Open Together with @huggingface!
Join the open AI community at The Midway in San Francisco on Oct 16 to kick off Open Source AI Week. Help us build a future where AI is open, secure, and fully transparent.
The event is free. RSVP:
R
OpenAI Developers a relayé @rhiannon_io
As AI continues to advance, I think it’s even more important for humans to come together, connect, & build a shared vision for the future.
Six weeks into my role on OpenAI's DX community team, I’ve seen that happen through our Codex Ambassador events...
Yann LeCun a relayé @KenRoth
A Saudi court has sentenced a Sri Lankan migrant worker to death over a Facebook comment that was deemed blasphemous, whatever that is.
B
Yann LeCun a relayé @berkerdemirel97
In JEPA, are we preventing collapse in the representation we actually use downstream?
TLDR: not necessarily. A high-rank projected representation does not guarantee a non-collapsed backbone.
λ-JEPA is designed to target this gap.
A
Emad a relayé @asha_shar
Today we're standing up XBOX LABS (). Here is the note that I sent to the team:
Team,
The first video games emerged some 75 years ago. In the time since, there have been booms and busts, but every year there are more great games to play, more devices to
L
Hugging Face a relayé @linoy_tsaban
New H3 character swap LoRA🔥
amazing work from @Alissonerdx
LoRA:
Space:
Harrison Chase a cité @neil_xbt
router lives in the harness, not a gateway, bc the harness knows the task
blog:
A
Hugging Face a relayé @andrii_liubivyi
Someone form @huggingface could you do it in reallife?
Microduck fight
Hugging Face a relayé @ClementDelangue
You don't want to miss this if you're in SF Friday. Hearing there are A LOT of surprises!
Hugging Face a relayé @NielsRogge
Decision Model (also called System One model or typed decision model) has been added as a method to Papers with Code
Explore all papers here:
Hugging Face a relayé @Zai_org
Introducing GLM-Legal
Built on GLM-5.3-Flash, GLM-Legal matches frontier models on mainstream legal benchmarks.
Designed for real legal workflows: contract review, evidence analysis, compliance, and more.
Tech Blog:
Apply for API Access:
Harrison Chase a relayé @sydneyrunkle
deepagents just hit 30k stars on github! deepagents is great for building agents that do complex work.
they manage context well and use a durable runtime.
though inspired by coding agent design, deepagents are for any domain!
J
Jason Wei a relayé @ren_hongyu
🍉 🍉 🍉 Muse Wonder and Muse Wonder Full is here. A new series of model where we push hard towards the personal superintelligence mission. The models achieve sota on internal personal and professional agent tasks across domains. Try it out today!
S
Yann LeCun a relayé @ShibleyTelhami
The full Democracy Perception Index 2026 is available at link below; As posted a while back, Israel is the most negatively perceived country in the world, followed by North Korea and Afghanistan.
P
Hugging Face a relayé @pollenrobotics
The Microduck Arena is live!
Train your duck with RL, publish it on @huggingface, and race it in the 2 m Sprint, Hill Climb, Roller Sprint and MicroDuck Got a Talent. No robot required.
Join the live 👉
#Microduck #robotics
Hugging Face a relayé @allen_ai
This Friday, meet the open source AI community at Open Together, co-hosted with @huggingface.
Join us at The Midway in SF on Oct 16 — bringing the community together to help build a future where AI is open, secure, and fully transparent.
RSVP:
E
Hugging Face a relayé @EmmaScharfmann
🧬 🤗 A new atlas of microproteins in the human brain is now @huggingface!
It brings together more than 1,000 microproteins, including one that’s reduced in Alzheimer’s and appears to play a role in how the brain’s immune cells produce energy.
Dataset:
V
Yann LeCun a relayé @vishalmisra
This full interview of Michael Jordan is well worth reading. The other MJ occupies the same position in basketball roughly as this one in machine learning
Yann LeCun a relayé @maddenifico
OUTRAGE: Elon Musk's pro-fascist PAC Traps Voters with Deceptive "" Data Harvest!
To help Trump rig the midterm elections, far-right disinformation propagandist Elon Musk is at it again.
This is an absolute betrayal of voter trust and a blatant attempt
Hugging Face a relayé @halcyonrayes
3D worlds have been all over the internet 🌐 lately, so we put together the best ones across many models, for you to play with 🤗
introducing a new entry to the 3D-arena 🏟️ family, which focuses on LLM 3D Worlds 🌎
go and make your voice heard here 🚀:
H
OpenAI Developers a relayé @haithehuman
I built GTA Toronto + Uber driving with Codex for my new Modretro Chromatic that i got from OpenAI devday two weeks ago.
the most fun ive had in year. pure GTA II vibes. game is open source here:
hardmaru a relayé @SakanaAILabs
Introducing Multi-Agent Self-Supervision (MASS): Recursive self-improvement through collective intelligence
Blog:
As AI systems tackle increasingly open-ended research problems, human experts may struggle to assess results and guide further progress. The
Harrison Chase a cité @leerob
this is the summary slide that describes how all agents in the future will look
very much aligns with our vision for managed deepagents
durable workflows: builds on LangGraph/LSD
can use a computer: uses Langsmith Sandboxes
Harrison Chase a cité @leerob
this is a great talk
if you want to just follow along in slides, extracted screenshots here to make it easy
writing up thoughts as we speak
Yann LeCun a relayé @PessimistsArc
“Faceless machines that can out think any man steadily move into new areas formerly reserved for the mind. They are electronic computers, capable of advising a sausage maker on the best mixture of meats or a doctor on the genetic code of life” (1962)
J
Yann LeCun a relayé @KempeLab
Huge pleasure to participate in this émission of Radio France Inter on AI in the great Big Bang series: @franceinter moderated by the brilliant @galfard with @NawellMadaniAct and @evdende2 (in French)
A
Yann LeCun a relayé @artemZholus
Introducing RoboJEPA, the first scaling law for multiembodiment world models.
We scaled JEPA predictors up to 8B params, showed scaling laws extrapolate and correlated compute scaling with planning performance.
We release all checkpoints and the code.
T
Yann LeCun a relayé @ThomasScialom
@ylecun @dominik_schnaus @alex_conneau @GuillaumeLample Ran this exact experiment back in the BERT days: frozen visual encoder + frozen BERT, one linear layer between them. Worked remarkably well:
And indeed Yann: FAIR's zero-shot cross-lingual alignment was my direct inspiration and intuition!
Harrison Chase a cité @yuhasbeentaken
agree - most orchestration steps don't need a frontier model
for Open SWE we moved model choice into the harness, so each task goes to the cheapest model that still does the job, tested against quality. median cost per task dropped 64%
Harrison Chase a cité @ch3nweiii
good framing from @ch3nweiii - a lot of agent steps are yes/no calls, not generation
Jev answers those as typed questions with calibrated probabilities, so the frontier model sticks to the hard work
where those calls sit in a harness:
Emad a relayé @a16z
Call center jobs grew ~4% annually for 15 years
Now they’re shrinking 4% annually
Charts of the Week:
Nous Research a répondu à @NousResearch
Slight Correction: Step 5 Preview's final score on Hermes Index has been updated to 32.83, slightly behind GPT-6 Luna (33.89) and GLM 5.3 Flash (34.95)
Learn more @
E
Yann LeCun a relayé @ELLISforEurope
Sessions for the ELLIS UnConference are online! Kicking off with two sessions on World Models 🌍, including a 30-min tutorial by @ylecun!
🔎 Sessions:
❗ Register:
Heading to @EurIPSConf in Paris? Join us a day early on Dec 8!
N
Harrison Chase a relayé @ndrezn
We've been thinking a lot about great agent UX with Managed Deep Agents - here's some of what we've been building in this area and a bit about our product philosophy
LangChain a cité @LangChain
Experience the best of Interrupt.
Access our archive and watch every session on-demand:
hardmaru a relayé @SakanaAILabs
Beyond Imitation: A Framework and Benchmark for LLM-Assisted Peer Review
Peer review needs support, not substitutes. Accepted at TMLR: our new paper on using AI to help reviewers catch errors in research papers.
As research submissions grow, so does the
Q
Hugging Face a relayé @QGallouedec
TRL v1.15 is out. Our biggest optimization ever.
Fused LM head, on by default:
→ up to 82% less peak VRAM
→ 7x longer sequences
→ DPO 10k → 59k tokens, GRPO 29k → 115k
And more...
Harrison Chase a cité @xennygrimmato_
trajectory labeling is really several questions, not one pass/fail - nice framing here
that's how Jev as a judge works in LangSmith evals: difficulty and correctness each get their own typed answer, in one pass, on every trace
hardmaru a relayé @SakanaAILabs
アイリス株式会社（ @ShoOkiyama ）の医師向けエビデンス検索ツール「Evidence Finder」に、Sakana AIの技術が採用されました。
医師の質問に対し、文献を検索して引用元を明示しながら回答。文献の比較・統合と回答生成の部分を担います。
詳細はこちら 🐟
Yann LeCun a relayé @randall_balestr
Financial data is RIPE for JEPAs... noise, non-stationarity, unknown ground truth states... yet data and evaluation is not readily available! We resolve that in our latest NeurIPS paper `Towards Financial World Modeling' TLDR: invest with LeJEPA ;)
📄 🧵⬇️
Y
Yann LeCun a relayé @YujiaZheng9
Is it possible to provably recover individual latent variables of the true world, even without reconstruction (e.g., JEPA)?
Yes, with DSReg, a simple regularization that can be applied post hoc to your pretrained model!
OpenAI Developers a relayé @GoogleAI
Our SynthID technology lets you easily verify if an image, video, or audio file is AI-generated, and today we’re expanding access.
Now anyone can go to to check files for a SynthID watermark — whether they were generated by us or industry partners like
Harrison Chase a cité @jaminball
Sam on where decision models fit in a harness and how to use Jev with LangChain
Y
Hugging Face a relayé @Weyaxi
Everything is open on @huggingface: the environments, every trajectory (including the 14k failures), the exact training split and the checkpoint.
Blog:
🤗
Many thanks @adithya_s_k and @maximelabonne for the reviews and feedback!
Harrison Chase a relayé @sydneyrunkle
the @OpenAI decisions API is now in @LangChain!
we shipped:
1. OpenAIDecisions: ask yes/no, choice, and score questions
2. model routing middleware
3. auto mode middleware
docs:
M
ClaudeDevs a cité @ClaudeDevs
Monthly API credits are now available on Max and Team plans: $100 on Max 5x, $200 on Max 20x, and up to $500 pooled on Team. Use them on any Claude model.
How it works and full terms:
F
Yann LeCun a relayé @BachFrancis
The one and only Michael Jordan was interviewed last week by the French newspaper Libération. Here is an English translation. Enjoy!
A
M
Yann LeCun a relayé @mjfree
LET THE CHEATING BEGIN: Trump administration refuses international election monitors for first time in over 2 decades
J
Hugging Face a relayé @1a1a11a
The largest agentic LLM Inference dataset has arrived!
206B tokens · 12,002 sessions · 1,186,582 LLM requests · 1,213,347 tool calls
The traces include tool names, arguments, durations, and more.
Explore the dataset on @huggingface
Paper coming soon!
Yann LeCun a relayé @KenRoth
So Trump has given up on NO nuclear enrichment for Iran. JD Vance says it must be a "meaningful" reduction in Iran's enrichment capacity. Like the Obama deal that Trump ripped up? Trump started a pointless war for this?
hardmaru a relayé @SakanaAILabs
ニュースレター「Sakana AI Insider」、登録受付中です🐟
詳細：
・プロダクトのリリース情報
・研究開発のアップデート
・Swagプレゼントやイベントの優先案内
先日公開し、多数ご応募いただいたイベントでは購読者の皆様への先行受付を実施しました！
登録はこちら 👇
E
Hugging Face a relayé @RisingSayak
Was meant to be a Valentine's gift, but I'm releasing it early ❤️
Wrote a long piece on KV caching in flow models for image generation.
Includes intuition, commentary, implementation details, benchmarks, & some quirky experiments.
elvis a répondu à @omarsar0
Start with something fun, like a fake Crocs x Porsche campaign. You can try it here:)
Hugging Face a relayé @liquidai
Choose d1-3B for high-quality text and vision decisions, or explore our experimental d1-omni-600M when footprint matters.
Download the weights, fine-tune, and deploy locally.
> Blog:
> d1-3B:
> d1-omni-600M:
OpenAI Developers a répondu à @OpenAIDevs
Ready to ship yours?
Get clearer review feedback with our rebuilt plugin submission flow.
Submit your plugin for review:
Hugging Face a relayé @multimodalart
come meet the microduck on oct 16th if you are in sf
i may have convinced it to make its first ever appearance to the open public at the Open Together event 🦆
J
A
Emad a relayé @AnjneyMidha
hundreds of MI355x and B300 nodes
now live on
subsidized for all .edu and .govs
it’s time to scale the frontier ecosystem
for the public benefit
Yann LeCun a relayé @random_walker
"Even “normal” general-purpose technologies are initially volatile and dangerous, even those we come to think of as unalloyed goods."
Completely agree, and appreciate @DKThomp citing us making that very point. Emphasizing AI's similarity to past general
Hugging Face a relayé @multimodalart
come meet the microduck on oct 16th if you are in sf
i may have convinced it to make its first ever appearance to the open public at the Open Together event 🦆
S
Harrison Chase a relayé @Shashikant86
Can we call next week as official "London Agent Week" @hwchase17 @ankush_gola11 WDYT?
@LangChain Interrupt is happening on 13th October and I am hosting @AgentEngHQ Agent Engineering London 🇬🇧 happening on 16th October
👉
Its going to be awesome
J
Hugging Face a relayé @jeffboudier
With 150 people on the waitlist and 10 days until Open Together we had to double capacity... again.
Calling all AI Builders to San Francisco on Friday Oct 16, join us!
Register:
Hugging Face a relayé @perplexitydevs
pplx-decider-v1.1-27b, our updated open weights multimodal decision model, is now available.
It scores the highest on the new @huggingface Decision Index 0.3 benchmark. pplx-decider-v1.1-27b costs half as much as v1, at $0.02 per million input tokens.
Hugging Face a relayé @multimodalart
the jev Decision Index v0.3 is here 🎯
featuring a way stronger set of benchmarks, now with a strong private half 🔒 so the ranking tracks skills beyond benchmarks
and, as promised, vision 👁️
Yann LeCun a relayé @randall_balestr
Introducing H-JEPA: the first end-to-end learned hierarchical world model for long-horizon visual planning!
Built atop SIGReg/LeWM, H-JEPA provides a stable training recipe where semantic abstractions naturally emerge across levels...
Paper+Code: 🧵👇
F
Google DeepMind a relayé @FlemingCentre
Supported by @GoogleDeepMind , our new 3-year programme will develop the methods and standards needed to evaluate artificial intelligence for antimicrobial resistance , so that we can know when an AI system is accurate, reliable and ready to use.
Nous Research a relayé @Teknium
Just made this cool catalog with a ton of open platform hardware, devices and more that your Hermes Agent (or any agent really) can build on, integrate with, or live inside!
Check it out:
A
Harrison Chase a relayé @businessbarista
Hot take: evals are the #1 thing keeping companies from becoming AI-native.
If you want to understand evals, read this guide on eval-driven development.
Shoutout @jake_klaris for the awesome read.
Hugging Face a relayé @vllm_project
We're excited to Co-host Open Together with @huggingface! We'll kick off Open Source AI Week with a gathering of community open source builders.
The event is Friday October 16th in SF at The Midway.
Register here:
P
S
Hugging Face a relayé @fffiloni
I built a @gradio playground to make this tangible 🔀 cc @victormustar
Give Julia-1 a situation: it ranks options, estimates a yes/no probability, and scores an ordered scale. No generated text.
Change the context and see how the decisions shift 👇
LangChain a cité @LangChain
🗽 Every session from Interrupt NYC, now on demand.
⏯️ Access the archive
Yann LeCun a relayé @SteveRattner
Land-grant colleges and the New Deal helped Americans move from farms to new jobs in the city.
Workers hit by the China shock got far less help.
A.I. is next. Government has to make sure it lifts all boats.
My @nytopinion op-ed:
Hugging Face a relayé @RedHat_AI
An NVFP4 checkpoint of Qwen3.8-Flash-Next is on Hugging Face. MoE experts quantized to FP4, the rest kept in BF16, ready for vLLM. Text, image, and video in.
We benchmarked it against other popular checkpoints. See how it holds up:
Nous Research a relayé @Teknium
The plugins catalog had been growing fast! Nearly 500 mods to Hermes that add new capabilities, model providers, memory enhancements, visual components and a lot more!
Explore them all here:
Yann LeCun a relayé @ylecun
@abuchanlife - training frontier models is expensive
- distilling a frontier model is cheap
- this market force alone tells us that frontier foundation models will be free/open.
- there are others.
Yann LeCun a relayé @KenRoth
Republicans' problem is not just Trump. It is that they did nothing to stop him as he undermined America's economy and democracy. Avoiding his anger was more important to them than benefiting the country.
M
Yann LeCun a relayé @MuzafferKal_
Hamiltonian JEPA: Action-Conditioned World Models with an Inherited Control State
Yann LeCun a relayé @KenRoth
Trump is so unpopular because of his corruption, self-preoccupation, and disastrous war of choice with Iran, all fueling the affordability crisis that he denies, he has put the Republican Senate in play. Democrats have a good chance of flipping it.
hardmaru a relayé @SakanaAILabs
The new Sakana AI website is live. 🐡
We have updated our digital home to reflect our products, enterprise solutions, and latest research.
Sakana AIの新サイトを公開しました。エンタープライズ事例やApplied Teamの採用情報も更新しています。ぜひご覧ください！
Harrison Chase a cité @dbreunig
cool project by @dbreunig
in langchain this is ModelRouterMiddleware - jev reads the first message, picks the model, and that model handles the whole run
jev is cheap enough that you can also re-pick after each tool result with a custom hook
docs:
Yann LeCun a relayé @KenRoth
Putin’s latest ploy to get Trump to go soft on the terms for an end to Russia’s invasion of Ukraine is to bribe people close to Trump with a windfall oil deal.
Yann LeCun a relayé @SteveRattner
Since 1830, technology has cut the American workweek from 69 hours to 38 while raising real income per person roughly 20-fold.
Properly managed, A.I. can do the same.
My @nytopinion Chart:
J
Yann LeCun a relayé @KempeLab
World Modeling for Physics Workshop in Aspen:
To clarify: Applications to participate (deadline Oct 15) go through the Aspen Center for Physics
Log in or create an account with ACP, then apply.
Yann LeCun a relayé @SteveRattner
So far, A.I.'s losers are routine office jobs like data entry keyers and customer service reps.
Its winners are data scientists and the trades building the data centers like construction laborers and electricians.
My @nytopinion Chart:
ClaudeDevs a répondu à @ClaudeDevs
"You should know" is a mod that works by spinning off a sideagent to observe Claude's output.
Read more about mods in our newest blog post:
S
Stability AI a relayé @stableaudio
The chat came ready for our Stable Audio plugin deep dive.
Prompt-based sound generation, live inside the DAW, with the community asking questions and reacting in real time.
Missed it? Watch the full stream here:
LangChain a relayé @caspar_br
User memory in Managed Deep Agents 0.8: your agent can remember the people it works with
Harrison Chase a relayé @caspar_br
User memory in Managed Deep Agents 0.8: your agent can remember the people it works with
Yann LeCun a relayé @SteveRattner
Productivity growth raises per-capita incomes, but outside the dot-com boom, it's been sluggish for 50 years.
A.I. is our best chance in a generation to change that.
My @nytopinion Chart:
Hugging Face a relayé @vanstriendaniel
New in @huggingface Jobs docs: a Train Models page.
A quick overview for you or your agent, covering Transformers, TRL, @UnslothAI, @axolotl_ai and Diffusers. Each example runs in minutes on a single A10G for around $0.10.
Hugging Face a relayé @ggerganov
Decision models in llama.cpp are now available
The `/v1/systemone` endpoint is available in the latest llama builds. Use it to do Jev-style inference locally, efficiently and privately. Multiple open models are supported with more to come.
