AI Agents Are Going Independent: The Rise of Agentic Commerce
Sponsored:
Browsing without a VPN is like mailing your passwords on a postcard. Hackers, advertisers, and your ISP are tracking every click.
PureVPN wraps your data in military-grade encryption, making you completely invisible to snoops while unlocking restricted content worldwide. Don’t leave your digital life exposed for another second.
Reclaim your online privacy. Click here to secure your connection now!
Ramesh Raskar’s keynote at ODSC on NANDA for Agentic Commerce paints a future where AI agents operate autonomously across the open internet—discovering, negotiating, and making decisions. This isn’t just about smarter chatbots; it’s about infrastructure for a decentralized agent economy. For open source enthusiasts, this raises questions: Will these agents rely on open protocols or proprietary gateways? The NANDA project, from MIT, emphasizes distributed architectures, suggesting a path that aligns with open source values. Meanwhile, Jason from Instaclustr demonstrates MCP Gateway, an open standard (Model Context Protocol) to connect agents to data without custom code. This is a crucial piece—agents need standardized, secure data access to be viable. The takeaway: Build with open standards like MCP and contribute to projects like NANDA to shape an agentic web that remains open and interoperable.
RISC-V Web Platform: Closer to Tier-1 Support
At FOSSASIA Summit 2026, Joey Zeng detailed progress on making Chromium, Node.js, and the JS ecosystem first-class on RISC-V. With upstream work on V8, Blink, and WebRTC, including JIT compilation and CI, the open hardware platform inches toward full web compatibility. For developers, this means RISC-V devices could soon run modern web apps natively—no emulation. The challenges remain (performance, build system integration), but the roadmap is clear. This is a win for open hardware and software freedom: a fully open architecture becoming a viable web platform.
Private AI: Run LLMs on Your Network
A FOSSASIA panel debated whether you’d buy ChatGPT if you could run it privately. The answer: many would. The panel demonstrated deploying local LLMs using Ollama, Docker Model Runner, and Foundry Local, integrated via .NET Aspire. For enterprises and privacy-conscious users, self-hosting open source models is now practical. Combine this with the MLflow tutorial on building a complete RAG pipeline with tracing and evaluation—you can create production-grade AI without sending data to the cloud. The open source ecosystem provides the tools (MLflow, Ollama, LangChain) to do this securely and scalably.
Security and Trust: SBOMs Without IP Leakage
Sharvil Bhatt and Swastik Gour presented a cryptographic approach to SBOM auditing that doesn’t leak intellectual property. Their ‘commit-then-disclose’ method allows verifying software supply chain integrity without exposing proprietary code. This is critical for open source adoption in enterprises: they can now audit dependencies without fear of IP leakage. Combined with the broader trend of secure, transparent AI (as seen in the RAGAS evaluation for RAG apps), trust is being built into the stack from the ground up.
Practical Demonstrations and Enterprise Insights
Stephanie Anani (OpenAI) showed how Codex helps solutions engineers make AI tangible—by converting messy customer data into interactive demos. This humanizes AI adoption. Meanwhile, Ericsson’s Malin Persson advises starting with business value and working backward to data infrastructure. For open source projects, this means focusing on real problems and enabling seamless integration with existing data fabrics. The Windows Weekly and Late Night Linux podcasts remind us that the OS landscape is evolving too—SteamOS on own hardware and Ubuntu’s speech-to-text signal more open choices for desktop users.
What This Means for You
Whether you’re a developer, architect, or open source advocate, these stories point to a converged future: open hardware (RISC-V), private AI (local LLMs), standard agent protocols (MCP), and verifiable security (cryptographic SBOMs). The tools are mature enough to start building today. Contribute to projects like MLflow, Trino, or the MCP Gateway to help shape these standards. The open source community is not just participating—it’s leading the way in making AI and infrastructure open, secure, and decentralized.
Source: OpenWorld.news/category/videos