Here is your today’s AI Dev Brief from Marktechpost, covering core research, models, infrastructure tools, and applied updates for AI developers and researchers.
NVIDIA AI Releases Nemotron 3: A Hybrid Mamba Transformer MoE Stack for Long Context Agentic AI
NVIDIA Nemotron 3 is an open family of hybrid Mamba Transformer MoE models, designed for agentic AI with long context and high efficiency. The lineup includes Nano, Super and Ultra, all using a Mixture of Experts hybrid Mamba Transformer backbone, multi environment reinforcement learning and a native 1 million token context window for multi agent workflows. Super and Ultra add LatentMoE, multi token prediction and NVFP4 4 bit training for better accuracy and throughput, while Nemotron 3 Nano is already available with open weights, datasets and NeMo Gym based RL tools for developers who want to build and tune specialized agentic systems on NVIDIA GPUs and common inference stacks. Read the full insights/article here.
Anthropic AI Releases Bloom: An Open-Source Agentic Framework for Automated Behavioral Evaluations of Frontier AI Models
Bloom takes a single behavior definition, for example sycophancy or self preferential bias, and automatically generates scenarios, runs rollouts and scores how often that behavior appears, all from a seed config. It uses a 4 stage pipeline, understanding, ideation, rollout and judgment, and plugs into LiteLLM, Weights and Biases and Inspect compatible viewers for analysis. Anthropic is already using Bloom on 4 alignment focused behaviors across 16 models, and finds that Bloom’s automated judgments track closely with human labels while distinguishing intentionally misaligned “model organisms” from production models. For teams working on evals, safety and reliability, Bloom looks like a useful open source starting point for building behavior specific evaluation suites that can evolve with each new model release. Read the full insights/article here.
Project Notebooks/Tutorials
▶ [Open Source] Rogue: An Open-Source AI Agent Evaluator worth trying Codes & Examples
▶ A Coding Guide to Design a Complete Agentic Workflow in Gemini for Automated Medical Evidence Gathering and Prior Authorization Submission Codes Tutorial
▶ How to Build an Agentic Voice AI Assistant that Understands, Reasons, Plans, and Responds through Autonomous Multi-Step Intelligence Codes Tutorial
▶ Build a Multi-Agent System for Integrated Transcriptomic, Proteomic, and Metabolomic Data Interpretation with Pathway Reasoning Codes Tutorial
▶ How to Build a Model-Native Agent That Learns Internal Planning, Memory, and Multi-Tool Reasoning Through End-to-End Reinforcement Learning Codes Tutorial
▶ Build an Autonomous Wet-Lab Protocol Planner and Validator Using Salesforce CodeGen for Agentic Experiment Design and Safety Optimization Codes Tutorial