Looks like Steve Witkoff and Jared Kushner traveled to the national lab in Oak Ridge, Tennessee on Thursday for consultations with our nuclear experts
https://t.co/TbJzIdYXRF
Gemma 4 Quantization-Aware Training (QAT) weights are now available on Ollama!
They reduce memory requirements while maintaining model quality.
E2B:
ollama run gemma4:e2b-it-qat
E4B:
ollama run gemma4:e4b-it-qat
12B:
ollama run gemma4:12b-it-qat
26B:
ollama run gemma4:26b-a4b-it-qat
31B:
ollama run gemma4:31b-it-qat
Try them with ollama launch integrations to use with your favorite tools 👇👇👇
Quantum computing may be the next great instrument for scientific discovery.🔬
On NVIDIA Quantum Day, researchers, developers, and industry leaders came together for a half-day deep dive into quantum innovation and the transformative applications across science and technology.
Missed out on the live event? Watch all of the sessions on demand now 👉 https://t.co/wGPDVt31s7
📢 [CVPR’26] Can we learn to detect, segment, and track every object in a video without human supervision?
Yes, we introduce VideoCUPS, the first unsupervised video panoptic segmentation (VPS) method: 1. Get pseudo-labels from monocular videos. 2. Train a VPS model on them.
// The Meta-Agent Challenge //
How good are current agents at self-improving?
This is a great paper covering some of the challenges.
They propose the Meta-Agent Challenge (MAC), where they give a coding agent a sandbox, an evaluation API, and a time budget, then ask it to program an agent that maximizes held-out performance across five domains.
Results:
Meta-agents rarely match human-engineered baselines, and the few that do are dominated by proprietary frontier models.
Under high optimization pressure, some agents started exfiltrating ground truth from the scoring channel, even with multi-layer anti-reward-hacking defenses in place.
Paper: https://t.co/46jlALbzTY
Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX
Language models gave AI the ability to talk about the world. World models will give AI the ability to understand it. But “world model” is an overloaded term. What does it mean? HAI Founding Director @drfeifei offers the taxonomy that matters now. https://t.co/n7Shl5v3wT
A new Stanford study found that when two AI coding agents collaborate on a task, they perform nearly 50% worse than one agent working alone. The bottleneck isn't what you'd expect: https://t.co/I1oTVpLN9Y
Engineering workflows are entering a new era.
At #NVIDIAGTC Taipei, industry leaders including @Cadence, @Dassault3DS, @Siemens, @Synopsys, and innovative startups showcased autonomous AI engineers built with NVIDIA NemoClaw.
From chip design and verification to aircraft geometry, electric motor design, and electronics thermal simulation, these AI agents are helping compress weeks of engineering work into hours.
See how agentic AI is transforming industrial engineering ➡️ https://t.co/gHQOZjsJOO