AI Intelligence Brief
March 12, 2026 · Last 24 Hours
đź”´ NVIDIA Releases Nemotron 3 Super: 120B Hybrid MoE Model for Agentic AI
Source: NVIDIA Technical Blog · March 11, 2026
NVIDIA released Nemotron 3 Super, a 120B total parameter model with 12B active parameters using a hybrid Mamba-Transformer Mixture-of-Experts architecture. The model delivers up to 5x higher throughput than previous Nemotron models and features a native 1M-token context window. Key innovations include Latent MoE for 4x expert capacity at same cost, Multi-Token Prediction for faster generation, and native NVFP4 pretraining optimized for Blackwell architecture. It scores 85.6% on PinchBench for agentic AI tasks and is fully open with weights, datasets, and training recipes.
🟠TIME Magazine Deep Dive: Anthropic’s War With the Pentagon
Source: TIME · March 11, 2026
TIME published an extensive investigation into the conflict between Anthropic and the U.S. Department of Defense. The report details how Anthropic refused to allow Claude to be used in fully autonomous weapons or mass surveillance of American citizens, leading the Trump administration to designate Anthropic as a “supply-chain risk to national security” on February 27. OpenAI subsequently signed the military contract instead. The article reveals Anthropic’s annualized revenue from Claude Code alone topped $2.5 billion by February 2026, with 70-90% of code used in developing future models now written by Claude itself.
🟡 JetBrains Previews Air: Agentic AI IDE Built on Fleet
Source: The Register · March 10, 2026
JetBrains introduced Air, a new agentic AI development environment that allows multiple AI agents to run tasks concurrently. Built on the codebase of the previously abandoned Fleet IDE, Air supports OpenAI Codex, Anthropic Claude Agent, Google Gemini CLI, and JetBrains Junie. It uses the Agent Client Protocol (ACP) for vendor-neutral agent-editor communication. Currently in public preview for macOS with Windows and Linux coming later. JetBrains also released Junie CLI, making its AI agent fully standalone.
đź’¬ Community Buzz
Anthropic Institute Announced to Advance AI Safety Research
Source: Anthropic Newsroom
Anthropic announced the creation of The Anthropic Institute, a new research initiative dedicated to advancing AI safety and alignment research. The announcement came on March 11, 2026, as part of Anthropic’s ongoing commitment to developing safe AI systems as their technology becomes more powerful.
NVIDIA GTC 2026: Agentic AI Era Dawns
Source: NVIDIA GTC · March 16-19, 2026
Industry anticipation is building for NVIDIA’s GTC 2026 conference starting March 16, with expectations focused on the “dawn of the agentic AI era.” The conference will feature breakthroughs in physical AI, agentic AI, inference, and AI factories, with CEO Jensen Huang’s keynote expected to outline major infrastructure announcements.
All sources verified as published within last 24 hours