Last week in AI | 23 March

2026-03-23 · Source: Thoughtworks Insights · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

The "Last week in AI" brief from March 23, 2026, highlights key developments, starting with NVIDIA's GTC conference. NVIDIA showcased its Vera Rubin platform, described as an "AI factory in a box," signaling a significant expansion beyond traditional chip offerings to integrated hardware and software solutions. The conference also saw the launch of NemoClaw, an extension of OpenClaw designed to provide essential guardrails, sandboxes, and security features, giving users enhanced control over agentic AI capabilities. Additionally, the brief discussed new research exploring a novel approach to improve Large Language Models' (LLMs) ability to handle complex multi-step tasks. This research involved "running a computer inside a transformer model" to overcome the token and computation intensity limitations of existing methods like tool invocation and agent scheduling.

Key takeaway

For AI Engineers evaluating infrastructure or deploying agentic systems, NVIDIA's shift towards "AI factory" platforms like Vera Rubin suggests a future of highly integrated solutions. You should consider how this impacts your hardware procurement and deployment strategies. Additionally, if you are developing agentic AI, explore NemoClaw for its built-in guardrails and sandboxes to ensure secure and controlled operations, mitigating potential risks associated with autonomous agents.

Key insights

NVIDIA is integrating hardware and software into "AI factory" platforms, while new research enhances LLM multi-step task capabilities.

Principles

• AI hardware is evolving into integrated "AI factory" platforms.
• Agentic AI requires robust safety and control mechanisms.
• LLM multi-step task efficiency can be improved via novel architectures.

Method

A research team experimented by running a computer inside a transformer model to expand LLMs' ability to perform complex multi-step tasks, addressing token and computation intensity.

In practice

• Explore NVIDIA's Vera Rubin platform for integrated AI infrastructure.
• Implement NemoClaw for secure agentic AI deployments.
• Investigate "computer-in-transformer" architectures for LLM efficiency.

Topics

• NVIDIA GTC
• Vera Rubin Platform
• AI Factory
• NemoClaw
• Agentic AI
• Large Language Models
• Transformer Models

Best for: AI Architect, Research Scientist, CTO, AI Scientist, AI Engineer, Director of AI/ML

Related on AIssential

• Nvidia’s version of OpenClaw could solve its biggest problem: security — Nvidia's NemoClaw offers an enterprise-grade, secure platform for deploying and managing AI agents based on the OpenClaw framework.
• NVIDIA Jetson Brings Agentic AI to the Physical World — NVIDIA Jetson now supports agentic AI via JetPack 7.2 and NemoClaw, enabling physical AI agents at the edge.
• LWiAI Podcast #237 - Nemotron 3 Super, xAI reborn, Anthropic Lawsuit, Research! — The AI landscape is rapidly evolving with new agentic tools, model releases, market shifts, and critical policy and research challenges.
• Build enterprise-ready Agentic AI with DataRobot using NVIDIA Nemotron 3 Super — Transforming powerful LLMs into trusted, production-grade enterprise AI systems requires robust infrastructure beyond model capabilities.
• Nvidia intros six new AI chips and new open models — Nvidia's new Rubin platform and specialized open models emphasize a full-stack AI factory approach beyond just GPUs.
• NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI — NVIDIA's latest platforms advance AI, robotics, and autonomous vehicles with significant performance and efficiency gains.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Thoughtworks Insights.