Today, I'm going to show you how to write code with LangChain Deep Agents Code, or dcode, inside a governed NemoClaw OpenShell sandbox. The flow is simple. You write code using dcode, a terminal-based coding agent.Dcode uses Nemotron 3 Ultra, served from Baseten, through an OpenAI-compatible endpoint. And then the agent runs inside a NemoClaw-managed OpenShell sandbox, so teams get a controlled environment around where the agent runs and what it can access. That means developers get their familiar coding ag ...

We're entering a new era in agent engineering, where open models are solving problems that used to be out of reach, running capable agents at a scale that just wasn't economical before. In this video, you'll see how to get rolling with one such model, Nemotron 3 Ultra, using Baseten as our model provider. Nemotron 3 Ultra is a 550 billion parameter model with strong reasoning performance running up to a blazing 300 tokens per second at a fraction of the cost of frontier models. On an Artificial Analysis Int ...

AI Agent Ecosystem & Strategic Development - The industry has reached a critical inflection point where AI agents, grounded in domain-specific knowledge and iterative tool usage, have become highly practical for enterprise applications [3][5][6] - Enterprises are shifting from general-purpose AI models to building proprietary "super agents" that leverage specialized intellectual property to optimize complex workflows like supply chain and chip design [8][25][26][39] - The "harness" (framework) has become as essential as the model itself, serving as the operating system for future business processes by providing necessary control, security, and integration [4][33][49] Performance & Economic Benchmarks - Open-weight models are achieving frontier-level performance, with Nemotron 3 Ultra reaching an 86% score on internal benchmarks, closely trailing Claude Opus at 87% and outperforming DeepSeek and Minimax models (82%-83%) [14][15] - Cost-effective intelligence is a key driver for adoption, as Nemotron 3 Ultra is reported to be 10 times cheaper than frontier alternatives, enabling more frequent iterations and larger search space exploration [15][16][17] Enterprise Implementation & Governance - The deployment of agentic systems requires a robust, secure runtime environment (such as OpenShell) to manage access control and data privacy, mirroring traditional IT onboarding processes for human employees [45][49][50] - Enterprises are encouraged to adopt a hybrid strategy: utilizing frontier models for general tasks while developing specialized, high-ROI agents for core business functions that cannot be outsourced [24][30][43] - The integration of AI agents is creating new job categories, shifting the focus of software engineers from manual coding to systems engineering, evaluation design, and guardrail implementation [59][60]