Trending Now

BlueField-4 DPU acts as the only hardware gate on a glowing cable between a sealed agent sandbox and AI model block, per NVIDIA agent safety tests.

AI

NVIDIA Puts Agent Safety in Silicon: 19 Rogue Actions in 122 Test Runs

September 28, 2026

4 min read

Icelandic AI hall liquid-cooled GPU rack; DSX MaxLPS fits 192 GPUs into 140 slots, 37% more AI capacity, amber streams.

AI

264.4 kW, 192 GPUs: DSX MaxLPS Squeezes 37% More AI Capacity

September 28, 2026

5 min read

Isometric 3D render of five translucent stacked prompt layers in a server rack, with tool cards, JSON packets, a red error gate, and a human-route chute.

Automations

Five-Layer Prompt Architecture That Stops Hermes Agent Failures in Production

September 28, 2026

4 min read

Glowing bot traffic river splits around a shield gateway as one lit menu card takes payment, visualizing Cloudflare's 1,000x agent traffic warning.

Business

Bots Now Outnumber Humans Online: Cloudflare’s 1,000x Agent Traffic Warning

September 28, 2026

4 min read

Search

Client Hub

AI

Explore the latest breakthroughs, applications, and ethical implications of Artificial Intelligence across all industries.

BlueField-4 DPU acts as the only hardware gate on a glowing cable between a sealed agent sandbox and AI model block, per NVIDIA agent safety tests.

NVIDIA Puts Agent Safety in Silicon: 19 Rogue Actions in 122 Test Runs

Keep Exploring

Introducing OlmoEarth embeddings: Custom embedding exports from OlmoEarth Studio for downstream analysis

How OlmoEarth Embeddings Lower the Barrier to Geospatial AI

Smartphone screen with glowing hand-tracking landmarks flowing to text, blocked side path, representing sign language AI.

DeepMind’s SL2T Brings Sign Language AI to Pixel 11

LFM2.5-VL-3B for Better and Faster Vision Capabilities for the Edge

LiquidAI’s New 3B Vision Model Beats Gemma 4B on Your Phone

Jetson module with NVENC/NVDEC chips, floating frames, code, verified throughput graph, for JetPack agentic video.

Agentic Video Skills in JetPack 7.2.1 Turn Prompts into Verified Pipelines

AI prompt window feeding sparse token packets to neural processor, bypassing tall document stack, powering agent circuit

Why Sending Fewer Tokens Makes AI Agents Smarter, Not Just Cheaper

NVIDIA chip with glowing green pathways routing to specialized units, symbolizing selective activation for agent execution.

NVIDIA’s 3B-Active Nemotron Lightning Hits the Agent Execution Pareto Frontier

NeMo Switchyard AI model router as railway switchyard with data trains, microchip servers, and neon cost and accuracy gauges.

NeMo Switchyard: NVIDIA’s Open-Source Model Router Slashes AI Agent Costs

Holographic wind tunnel around wireframe car, particle pressure contours, representing GeoPT's 60% data cut.

GeoPT: New Pretraining Method Cuts Physics Simulation Data by 60%

ExecuTorch RTX 5090 GPU with glowing tensor cores and a neural net, emitting a high-speed token stream in a transparent case.

ExecuTorch Makes 30B Agentic AI Run Locally at 233 Tokens Per Second

Run Local Agentic AI Workflows with Meta’s Muse Glimmer on NVIDIA

Meta’s Muse Glimmer Pushes 30B Agentic AI Onto a Single Consumer GPU