Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Our first robotics pod with Dyna’s Chief Scientist on their big bet on World Action Models & the Next Frontier of Physical AI
Daily digest / 13 October 2026
AI news digest — 13 October 2026
How to Rewrite Apache Spark UDFs for GPU Acceleration with AI Agents
Enterprise applications have become increasingly demanding. Agentic AI has compounded the need for vast amounts of enterprise data with frequent, complex... Enterprise applications have become increasingly demanding. Agentic AI has compounded the need for vast amounts of enter…
How artificial neural networks build their own learning machinery
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
MIT Lincoln Laboratory hosts National Security Commission on Emerging Biotechnology
Roadshow event highlights Massachusetts' defense-biotechnology ecosystem.
Developing Gemini 4 Argon responsibly and safely
GitHub’s model for secret detection: Finding more secrets with less noise
GitHub’s context-aware secret classifier improves detection quality, raises weighted acceptance by 11.5%, and reduces median scan execution time by 86%. The post GitHub’s model for secret detection: Finding more secrets with less noise appeared first on The GitHub Blog .
Run fast, private agents on-device with LFM2.5-24B-A2B
Applying Gemini 4 for large-scale codebase maintenance in Chromium
The Chromium codebase is millions of lines long and decades old. In the past, working with it has meant navigating decades of tangled features, where touching one component can unexpectedly affect another. Erik Chen, an engineer at Google, is using Gemini 4 Argon to ‘untangle’…
Using Gemini 4 Argon to protect against Cyber attackers
In shadow A/B testing on real-world systems, Wiz security researchers put Gemini 4 Argon to the test against complex cyber threats. Using its advanced reasoning and long-horizon capabilities, Gemini 4 Argon found over two times more vulnerabilities than baseline models and suc…
Kwai Summary Attention V2 Technical Report
Summary-token architectures such as Kwai Summary Attention (KSA) compress distant context into learnable Summary token per chunk, and hybrid variants interleave these layers with full attention to retain exact access to the original text. In such hybrid KSA-Full architecture, …
Gemini 4 Argon for Robotics
Gemini 4 Argon: Built to solve your hardest problems
Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI
Last year, we introduced GEM, the industry’s largest recommendation system (RecSys) foundation model. GEM established Meta’s first LLM-inspired paradigm for ads recommendation, driving major performance gains. Building on this foundation, our next goal is to further scale both…
Preparing chip designers for the nanostack era
IBM is releasing a process design kit for our groundbreaking 0.7-nanometer transistor architecture.
Gemini 4 Argon: Built to solve your hardest problems
We’re giving developers, enterprises, scientists, and consumers access to Gemini 4 Argon, our latest frontier model.
Teaching a humanoid robot to play cornhole with Gemini 4 Argon
In an internal test, Google DeepMind robotics researchers instructed Gemini 4 Argon to watch a video of a human playing cornhole. Using its video understanding and advanced reasoning capabilities, Gemini 4 built a virtual simulation in MuJoCo and accurately calculated the phys…
Untangling 500-page chemistry patents with Gemini 4
Auditing chemistry literature usually means weeks of manual, error-prone reading, where a molecule’s structure can sit 50 pages away from its experimental readout. At Isomorphic Labs, researchers Brian Bender and Santosh Adhikari are using Gemini 4 Argon’s multimodal capabilit…
NVIDIA Run:ai Is Now Open Source as Part of NVIDIA DSX AI Factory Platform
NVIDIA is making the full NVIDIA Run:ai platform available as open source software, giving customers, developers, and ecosystem partners access to... NVIDIA is making the full NVIDIA Run:ai platform available as open source software, giving customers, developers, and ecosystem…
Scaling AI-driven chip design with open agent factories
Open models and llm-d enabled IBM Research to explore quantum error correction hardware at 11.5× lower inference cost than an all-proprietary configuration.
How to transcribe an interview: Accurate, speaker-labeled transcripts
Learn how to transcribe an interview with AI, from verbatim vs. clean styles to speaker labels. ElevenCreative makes it easy to transcribe
Baroness Campbell gives the UK Parliament’s first ever AI-voiced speech with ElevenLabs
Baroness Campbell addressed the House of Lords in her own AI-recreated voice, a first for UK Parliament, built with ElevenLabs.
[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
MSL’s emergence as a frontier lab is finally complete.
Fish Audio Drama 3 Technical Report
We introduce Fish Audio Drama 3, an open-source speech generation model that turns text into directed performances, built on a 35B-parameter mixture-of-experts backbone that activates 3B parameters per token. It clones voices from multiple references, follows natural-language …
How Molmo 2’s open release helped an MIT researcher rethink fish tracking
Molmo 2’s open release helped MIT researcher Kai Van Brunt develop Molmo2Fish, an experimental model that tracks salmon in sonar footage and corrects some tracking errors using everyday language.
Q&A: Building accountability behind our words — and AI’s — in civic conversations
Professor Deb Roy contends there should be someone who is “answerable” for every piece of text in public life, no matter its origins.
Could this be the Frontier Labs’ WeWork moment?
Some ominous signs
Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings
For most languages of the world, language model pre-training operates in a data-constrained regime where models must repeat their training data many times, degrading generalization. Two remedies exist: aggressive hyperparameter tuning such as high weight decay, and mixing in d…