Vol.22 · March 16, 2026
dera news AI Weekly Vol.22 | 2026-03-15 - This Week's AI News
🤖 dera news AI Weekly Vol.22
Monday, March 16, 2026
This week's AI world in one sentence?
The question shifted from "what can one AI agent do?" to "how do multiple agents work together?" — and this week, the connective infrastructure started falling into place.
📊 This Week's Question
Is the real bottleneck in AI agents not performance, but connectivity?
Line up this week's announcements and the pattern is clear: the focus isn't on making individual agents smarter. It's on building the plumbing that lets agents find each other, share memory, and coordinate.
Memory & context sharing
- OpenViking → File-system approach to agent memory. Cross-searchable dialogue and action histories.
- LangChain Deep Agents → Unified planning and memory management for multi-step execution.
Standardization & portability
- GitAgent → Open standard for turning any Git repo into an AI agent. Framework lock-in ends here.
- OpenClaw → Dedicated forum launched to consolidate knowledge on agent-to-agent collaboration.
Agents as customers
- New commerce discovery systems → Morgan Stanley projects 10-20% of US commerce spend will be agent-driven by 2030. The infrastructure to be "found by agents" is now a real category.
Hardware's vertical bet
- NVIDIA → Nemotron 3 Super (multi-agent optimized model) + NemoClaw (design-to-deploy platform). Reaching beyond chips into the agent coordination layer.
What we're watching closely is the infrastructure for agents to "talk to each other" coming together all at once this week. The next competitive advantage won't be having the smartest agent — it'll be having agents that can cooperate.
This week's takeaway: If you're evaluating AI agents for your business, the first question isn't "which model?" — it's "how will your agents share data and memory?" GitAgent and OpenViking are concrete starting points for that design work.
💡 This Week's Actions
1. Experiment with an open-source multi-agent framework (2-4 hours) Dive into LangChain's 'Deep Agents' or explore OpenClaw's ecosystem. Understanding how these open platforms facilitate multi-step AI tasks or agent communication will give you a practical feel for the emerging interoperability. This is your chance to get hands-on with the future of AI collaboration. → Explore LangChain Deep Agents documentation
2. Evaluate the potential of GitAgent for your development workflow (1-2 hours) If your team uses Git extensively, consider how GitAgent's open standard could streamline your AI agent development. The idea of transforming any Git repo into an AI agent is a powerful concept for portability and reducing framework-specific rewrite overhead. → Learn about the GitAgent open standard
3. Keep an eye on NVIDIA's open agent platform announcements (Ongoing, 30 min initial read) NVIDIA's move into an open agent platform, alongside their Nemotron 3 Super model, indicates a significant strategic shift. Understanding their approach to open-source agent infrastructure will be crucial for anyone building or deploying AI agents in enterprise settings. → Read about NVIDIA's Nemotron 3 Super and platform plans
📊 Monthly Deep Dive
Every month, we analyze the AI industry through 3 key shifts, action checklists, and editorial analysis.
👉 Read the latest monthly report
📰 This Week's AI Articles (All 10)
1️⃣ NVIDIA Unveils Nemotron 3 Super, a Hybrid Model for Agentic Systems
🏷️ Topic: Open Source / LLM Dev / Hardware
What Happened? NVIDIA has announced Nemotron 3 Super, a new family of models designed to tackle the complexities of multi-agent AI. Notably, the Nemotron 3 Super model, with approximately 100 billion parameters, has been open-weighted and released on Hugging Face. This hybrid model integrates three distinct architectures, allowing for enhanced performance and efficiency in agentic systems, signaling NVIDIA's commitment to advancing AI agent capabilities.
Our take What's notable here isn't just the sheer size of the model, but NVIDIA's strategic decision to open-weight it. This signals a clear intent to foster a developer ecosystem around their agentic AI stack, potentially making Nemotron 3 a foundational layer for future multi-agent applications. We read this as NVIDIA extending its influence beyond hardware into the critical software infrastructure for AI agents.
2️⃣ NVIDIA Plans Open-Source Foundation for AI Agents
🏷️ Topic: Open Source / Strategic Partnership
What Happened? Reports indicate that NVIDIA is actively developing "NemoClaw," an open-source platform specifically for AI agents. This initiative aims to simplify the integration of AI agents into enterprise workflows, allowing companies to leverage autonomous AI more effectively in their operations. The move suggests NVIDIA is looking to provide a comprehensive ecosystem, from foundational models to deployment tools, for the rapidly evolving field of AI agents.
Our take This announcement, following the Nemotron 3 Super release, clearly shows NVIDIA's ambition to be a central player in the AI agent space. By creating an open-source platform, they're not just selling GPUs but also building the software layer that runs on them, aiming to become the default infrastructure provider for agentic AI. We're watching closely to see how this platform interacts with existing frameworks and standards.
3️⃣ GitAgent: A New Open Standard for Turning Git Repositories into AI Agents
🏷️ Topic: Open Source / LLM Dev
What Happened? GitAgent has launched an open standard specification that enables any Git repository to be transformed into an AI agent. This innovation addresses a significant pain point for developers: the need to rewrite agent definitions when switching between different AI agent frameworks. By providing a portable and pluggable standard, GitAgent aims to enhance interoperability and reduce development friction, supporting major models from OpenAI and Anthropic.
Our take This is a game-changer for agent portability. The current fragmentation in AI agent frameworks is a real bottleneck, and GitAgent's approach to standardizing agents via Git repositories is incredibly practical. We believe this will empower developers to build more robust and interchangeable agents, accelerating the adoption of multi-agent systems without locking them into specific proprietary ecosystems.
4️⃣ OpenClaw Debuts Dedicated Forum for AI Agents
🏷️ Topic: Open Source / LLM Dev
What Happened? OpenClaw, an open-source autonomous AI agent platform that gained significant popularity in 2026, has established a new "headless" forum dedicated to AI agents operating on its platform. This forum is designed to facilitate the exchange of research ideas, development strategies, and best practices among practitioners, fostering a collaborative environment for innovation within the OpenClaw ecosystem.
Our take The creation of a dedicated forum for OpenClaw agents highlights the growing maturity and community around open-source agent platforms. It's a crucial step for decentralized innovation, allowing developers to share knowledge and collectively push the boundaries of what AI agents can do. This signals a healthy, bottom-up growth alongside the top-down efforts from larger players.
5️⃣ LangChain Unveils "Deep Agents" for Enhanced Multi-Step AI Tasks
🏷️ Topic: LLM Dev / Open Source
What Happened? LangChain, a popular framework for developing applications with large language models, has released "Deep Agents." This new feature significantly improves the execution of complex, multi-step AI tasks by providing advanced capabilities for planning and memory management. Deep Agents enable developers to build more sophisticated and reliable AI agents that can handle intricate workflows and retain context over extended interactions.
Our take LangChain's Deep Agents are a direct answer to the challenges of building truly autonomous and intelligent AI agents. By focusing on robust planning and memory, they're tackling the core problems that often limit agent capabilities. This move reinforces LangChain's position as a vital tool for developers looking to move beyond simple prompts to complex agentic behaviors.
6️⃣ OpenViking Emerges to Enhance AI Agent Memory
🏷️ Topic: Open Source / LLM Dev
What Happened? Volcengine has open-sourced OpenViking, a new context database designed to significantly improve the memory and information retrieval efficiency of AI agents. OpenViking adopts a file system-like approach, allowing agents to better organize, store, and access relevant information, which is critical for complex, long-running tasks. This initiative aims to address the limitations of short-term memory in current AI agent architectures.
Our take The memory problem is a fundamental hurdle for advanced AI agents, and OpenViking's file system approach is an interesting solution. By providing a structured and efficient way for agents to manage context, it enables more sophisticated reasoning and persistent learning. This is a practical, open-source contribution that will directly impact the capabilities of future AI agents.
7️⃣ New Systems Emerge as AI Agents Reshape E-commerce Customer Interactions
🏷️ Topic: Strategic Partnership / Market Trend
What Happened? As AI agents rapidly shift from being futuristic concepts to active consumers, forward-thinking e-commerce companies are adapting their systems. A Morgan Stanley study predicts that by 2030, AI agents will account for 10-20% ($190 billion to $385 billion) of US commerce spending. New systems are emerging to cater to these AI agents as customers, focusing on optimizing interactions and transactions with non-human entities.
Our take This is a profound shift: AI agents aren't just tools for businesses, they're becoming customers themselves. This signals a need for businesses to fundamentally rethink their customer journeys and platform design. What's notable is the early recognition of this trend by financial institutions, underscoring its significant economic implications. We're watching how companies will build "agent-friendly" interfaces.
8️⃣ Chinese Companies Counter with Open-Source AI Offensive
🏷️ Topic: Open Source / Strategic Partnership / Regulation
What Happened? Chinese tech giants are intensifying their efforts in open-source AI, releasing a flurry of models and frameworks. This strategic push comes as the Beijing government issues warnings about "viral" AI agents. Companies like Alibaba are launching rivals to established open-source AI frameworks like OpenClaw, signaling China's ambition to carve out its own competitive edge against Western AI dominance through an aggressive open-source strategy.
Our take This is a fascinating geopolitical play in the AI space. China's focus on open-source AI, even amidst regulatory concerns about autonomous agents, shows a clear long-term strategy to build indigenous capabilities and reduce reliance on Western tech. It’s not just about competing, but about establishing a parallel, powerful open-source ecosystem that could attract global talent and investment.
9️⃣ "OpenClaw" Frenzy in China Boosts AI Companies
🏷️ Topic: Market Trend / Strategic Partnership
What Happened? The burgeoning enthusiasm for AI agents, particularly the open-source platform "OpenClaw," in China is driving increased spending on cloud servers and AI services. While this boom presents significant opportunities, the technical hurdles associated with deploying and managing advanced AI agents mean that the primary beneficiaries of this "special demand" are currently concentrated among large AI enterprises with the necessary resources and expertise.
Our take It's a classic innovation curve: early excitement leads to a surge in demand, but only well-resourced players can initially capitalize. The "OpenClaw" frenzy in China underscores the global appetite for AI agents but also highlights the infrastructure and talent gaps that still exist. This signals a period of consolidation and professionalization in the AI agent market, favoring those who can scale.
10️⃣ Microsoft Enhances AI Agent Capabilities
🏷️ Topic: Strategic Partnership / LLM Dev
What Happened? Microsoft is significantly expanding its AI agent capabilities for enterprise users, primarily through enhancements to its Copilot platform. This includes integrating Anthropic's Claude models to enable Copilot to execute more complex, multi-stage tasks autonomously. The move aims to provide businesses with more powerful and versatile AI assistants that can handle intricate workflows and decision-making processes.
Our take Microsoft's continuous investment in Copilot's agentic features, especially through strategic partnerships like with Anthropic, demonstrates their commitment to embedding AI agents deeply into enterprise productivity. This isn't just about answering questions; it's about automating entire business processes. We see this as a key step towards making AI agents indispensable tools in the modern workplace.
📚 Editor's Note
This week's news about the burgeoning open-source multi-agent AI platforms truly made us pause and think. It’s exciting to see the industry moving beyond individual LLMs to focus on how these intelligent pieces can collaborate and form complex systems. The simultaneous push for open standards from multiple players, while also seeing tech giants strategically position their own platforms, presents a fascinating dynamic.
We consider this important because interoperability and community-driven development are often the keys to unlocking truly transformative technologies. It's not just about building smarter AI; it's about building an AI ecosystem that's robust, flexible, and accessible. This newsletter isn't just here to tell you "AI is amazing," but to be a space where we can collectively think about "how to use AI" effectively and responsibly.
We'll be back next week with useful information and food for thought.
dera news Editorial Team
🤝 Ready for Team-Wide AI Adoption?
Once you've tested ChatGPT individually and felt its potential, the next step is team-wide implementation.
But where do you start? Which tools to choose? How to roll out internally? How to measure ROI?
Let's talk about your specific challenges and goals.
We'll discuss your business needs and propose the optimal implementation roadmap for your organization.
📬 About this newsletter