Mastodawn

Google announcing a so call revolution for the LLMs the same day @ylecun first and light world model is published: https://le-wm.github.io/. Coincidence ? #llms #ai #worldmodels

LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels

End-to-end joint-embedding predictive architecture from pixels.

sayzard 4d ago

Google AI Studio (@GoogleAIStudio)

프로젝트 Genie를 위해 세계 모델을 시각화하는 새로운 방법이 필요했고, 이를 위해 AI Studio를 사용해 바이브 코딩으로 구현했다고 소개했다. 창의적 코더들의 작업 방식과 AI Studio를 활용한 실험적 개발 사례를 보여주는 내용이다.

https://x.com/GoogleAIStudio/status/2036926844503810115

#aistudio #vibecoding #worldmodels #genie #creativcoding

Google AI Studio (@GoogleAIStudio) on X

for project genie, the team needed a new way to visualize world models, so they fired up AI Studio and started vibe coding hear how they did it, from creative coders @isofluff & @kaloyaan

X (formerly Twitter)

eicker.news ᳇ tech news 6d ago

#WorldModels, a new class of #foundationmodels, learn to predict future states from action-labelled gaming clips. Unlike traditional video models, World Models incorporate actions as a form of compression, enabling them to efficiently simulate complex, dynamic environments. This ability to “compute the uncomputable” makes World Models a promising path. https://www.notboring.co/p/world-models?eicker.news #tech #media #news

World Models: Computing the Uncomputable

A Co-Written Essay with General Intuition's Pim DeWitte

Not Boring by Packy McCormick

Romy Ilano Mar 14

For my homies a slice of life walking into the sense ai world models hackathon at founders inc. #sliceoflife #worldmodels #senseai

Tim Mar 10

- OpenAI and Google workers file brief in support of Anthropic: https://www.wired.com/story/openai-deepmind-employees-file-amicus-brief-anthropic-dod-lawsuit/

- Yann LeCun and Saining Xie launch AMI Labs AI to build world models: https://amilabs.xyz

- André Karpathy packaged his autoresearch and nanochat packages to recursively self-improve nanochat: https://github.com/karpathy/nanochat/commit/6ed7d1d82cee16c2e26f45d559ad3338447a6c1b https://github.com/karpathy/autoresearch

- More Google releases rumoured this week (source: Bindu Reddy)

#AI #AINews #openai #google #anthropic #amilabs #ylecun #karpathy #worldmodels #agenticdev

OpenAI and Google Workers File Amicus Brief in Support of Anthropic Against the US Government

Google DeepMind chief scientist Jeff Dean is among the AI researchers and engineers rushing to Anthropic's defense.

WIRED

Renaud JOLY Mar 10

Yann Lecun est financé 💰https://www.bfmtv.com/economie/entreprises/une-premiere-levee-de-fonds-historique-la-start-up-ami-labs-du-francais-yann-le-cun-recoit-1-milliard-de-dollars-pour-revolutionner-l-ia_AV-202603100065.html #yannlecun #ia #worldmodels

Une première levée de fonds historique: la start-up AMI Labs du français Yann Le Cun reçoit 1 milliard de dollars pour révolutionner l'IA

AMI Labs a annoncé ce mardi avoir levé plus d'un milliard de dollars. La start-up qui veut "révolutionner l'IA", fondée par Yann Le Cun, est déjà valorisé à 3,5 milliards de dollars.

BFM BUSINESS

sayzard Mar 9

fly51fly (@fly51fly)

LG 관련 연구(2026): 'Next Embedding Prediction' 기법을 제안해 임베딩 예측으로 월드 모델(world models) 성능을 강화했다고 보고한 논문입니다. 저자 G. Bredis, N. Balagansky, D. Gavrilov, R. Rakhimov(소속: T-Tech)이며 arXiv 링크가 함께 제공되었습니다. 월드 모델·임베딩 개선을 다루는 최신 연구 결과입니다.

https://x.com/fly51fly/status/2030763111259820192

#embeddings #worldmodels #research #lg

fly51fly (@fly51fly) on X

[LG] Next Embedding Prediction Makes World Models Stronger G Bredis, N Balagansky, D Gavrilov, R Rakhimov [T-Tech] (2026) https://t.co/qBQiY3X9u8

X (formerly Twitter)

sayzard Mar 7

Ilir Aliu (@IlirAliu_)

월드 모델은 로보틱스에서 핵심이 되고 있으나 속도와 장기적 물리 일관성에서 어려움이 있다. 'Interactive World Simulator'라는 새로운 액션-조건 월드 모델이 이 두 문제를 개선하며 10분 이상 길이의 시뮬레이션을 생성할 수 있다고 발표함(새 시스템/연구 소개).

https://x.com/IlirAliu_/status/2030358594210640002

#worldmodels #robotics #simulation #interactiveworldsimulator #research

Ilir Aliu (@IlirAliu_) on X

World models are becoming central in robotics. But most still struggle with two things: speed and long-horizon physical consistency. A new system called Interactive World Simulator pushes both forward. It’s an action-conditioned world model that can generate 10+ minutes of

X (formerly Twitter)