The New Stack Podcast

In a recent episode of The New Stack Agents from the Open Source Summit in Amsterdam, Jim Zemlin, executive director of the Linux Foundation, discussed the evolving landscape of open source AI. While the Linux Foundation has helped build ecosystems like the CNCF for cloud-native computing, there's no unified umbrella foundation yet for open source AI. Existing efforts include the PyTorch Foundation and LF AI & Data, but AI development is still fragmented across models, tooling, and standards. 

Zemlin highlighted the industry's shift from foundational models to open-weight models and now toward inference stacks and agentic AI. He suggested a collective effort may eventually form but cautioned against forcing structure too early, stressing the importance of not hindering innovation. Foundations, he said, must balance scale with agility. On the debate over what qualifies as "open source" in AI, Zemlin adopted a pragmatic view, acknowledging the costs of creating frontier models. He supports open-weight models and believes fully open models, from data to deployment, may emerge over time. 

Learn more from The New Stack about the latest in AI and open source, AI in China, Europe's AI and security regulations, and more: 

Open Source Is Not Local Source, and the Case for Global Cooperation 

US Blocks Open Source ‘Help’ From These Countries 

Open Source Is Worth Defending 

Join our community of newsletter subscribers to stay on top of the news and at the top of your game./

Enterprise AI is still in its infancy, with less than 1% of enterprise data currently used to fuel AI, according to Raj Verma, CEO of SingleStore. While consumer AI is slightly more advanced, most organizations are only beginning to understand the scale of infrastructure needed for true AI adoption. Verma predicts AI will evolve in three phases: first, the easy tasks will be automated; next, complex tasks will become easier; and finally, the seemingly impossible will become achievable—likely within three years. 

However, to reach that point, enterprises must align their data strategies with their AI ambitions. Many have rushed into AI fearing obsolescence, but without preparing their data infrastructure, they're at risk of failure. Current legacy systems are not designed for the massive concurrency demands of agentic AI, potentially leading to underperformance. Verma emphasizes the need to move beyond siloed or "swim lane" databases toward unified, high-performance data platforms tailored for the scale and complexity of the AI era.

Learn more from The New Stack about the latest evolution in AI infrastructure: 

How To Use AI To Design Intelligent, Adaptable Infrastructure

How to Support Developers in Building AI Workloads 

Join our community of newsletter subscribers to stay on top of the news and at the top of your game. 

Anthropic's Model Context Protocol (MCP) has become the standard for connecting AI agents to tools and data, but its security has lagged behind. In The New Stack Agents podcast, Tzvika Shneider, CEO of API security startup Pynt, discussed the growing risks MCP introduces. Shneider sees MCP as a natural evolution from traditional APIs to LLMs and now to AI agents. However, MCP adds complexity and vulnerability, especially as agents interact across multiple servers. 

Pynt’s research found that 72% of MCP plugins expose high-risk operations, like code execution or accessing privileged APIs, often without proper approval or validation. The danger compounds when untrusted inputs from one agent influence another with elevated permissions. Unlike traditional APIs, MCP calls are made by non-deterministic agents, making it harder to enforce security guardrails. While MCP exploits remain rare for now, most companies lack mature security strategies for it. Shneider believes MCP merely highlights existing API vulnerabilities, and organizations are only beginning to address these risks.

Learn more from The New Stack about the latest in Model Context Protocol: 

Model Context Protocol: A Primer for the Developers 

Building With MCP? Mind the Security Gaps 

MCP-UI Creators on Why AI Agents Need Rich User Interfaces

Join our community of newsletter subscribers to stay on top of the news and at the top of your game. 

Rahul Auradkar, executive VP and GM at Salesforce, grew up in India with a deep passion for cricket, where his love for the game sparked an early interest in data. This fascination with statistics laid the foundation for his current work leading Salesforce’s Data Cloud and Einstein (Unified Data Services) team. Auradkar reflects on how structured data has evolved—from relational databases in enterprise applications to data warehouses, data lakes, and lakehouses. He explains how initial efforts focused on analyzing structured data, which later fed back into business processes. 

Eventually, businesses realized that the byproducts of data—what he calls "data exhaust"—were themselves valuable. The rise of "old AI," or predictive AI, shifted perceptions, showing that data exhaust could define the application itself. As varied systems emerged with distinct protocols and SQL variants, data silos formed, trapping valuable insights. Auradkar emphasizes that the ongoing challenge is unifying these silos to enable seamless, meaningful business interactions—something Salesforce aims to solve with its Data Cloud and agentic AI platform.

Learn more from The New Stack about the evolution of structured data and agent AI: 

How Enterprises and Startups Can Master AI With Smarter Data Practices 

Enterprise AI Success Demands Real-Time Data Platforms

Join our community of newsletter subscribers to stay on top of the news and at the top of your game.

In this week’s episode ofThe New Stack Agents, Scott Carey, editor-in-chief of LeadDev, discussed their first AI Impact Report, which explores how engineering teams are adopting AI tools. The report shows that two-thirds of developers are actively using AI, with another 20% in pilot stages and only 2% having no plans to use AI — a group Carey finds particularly intriguing. Popular tools include Cursor (43%) and GitHub Copilot (37%), with others like OpenAI, Gemini, and Claude following, while Amazon Q and Replit lag behind.

Most developers use AI for code generation, documentation, and research, but usage for DevOps tasks like testing, deployment, and IT automation remains low. Carey finds this underutilization frustrating, given AI's potential impact in these areas. The report also highlights concern for junior developers, with 54% of respondents expecting fewer future hires at that level. While many believe AI boosts productivity, some remain unsure — a sign that organizations still struggle to measure developer performance effectively.

Learn more from The New Stack about the latest insights about the AI tool adoption: 

AI Adoption: Why Businesses Struggle to Move from Development to Production

3 Strategies for Speeding Up AI Adoption Among Developers

AI Everywhere: Overcoming Barriers to Adoption

Join our community of newsletter subscribers to stay on top of the news and at the top of your game. 

While AI training garners most of the spotlight — and investment — the demands ofAI inferenceare shaping up to be an even bigger challenge. In this episode ofThe New Stack Makers, Sid Sheth, founder and CEO of d-Matrix, argues that inference is anything but one-size-fits-all. Different use cases — from low-cost to high-interactivity or throughput-optimized — require tailored hardware, and existing GPU architectures aren’t built to address all these needs simultaneously.

“The world of inference is going to be truly heterogeneous,” Sheth said, meaning specialized hardware will be required to meet diverse performance profiles. A major bottleneck? The distance between memory and compute. Inference, especially in generative AI and agentic workflows, requires constant memory access, so minimizing the distance data must travel is key to improving performance and reducing cost.

To address this, d-Matrix developed Corsair, a modular platform where memory and compute are vertically stacked — “like pancakes” — enabling faster, more efficient inference. The result is scalable, flexible AI infrastructure purpose-built for inference at scale.

Learn more from The New Stack about inference compute and AI

Scaling AI Inference at the Edge with Distributed PostgreSQL

Deep Infra Is Building an AI Inference Cloud for Developers

Join our community of newsletter subscribers to stay on top of the news and at the top of your game 

In the latest episode ofThe New Stack Agents, Naveen Rao, VP of AI at Databricks and a former neuroscientist, reflects on the evolution of AI, neural networks, and the energy constraints that define both biological and artificial intelligence. Rao, who once built circuit systems as a child and later studied the brain’s 20-watt efficiency at Duke and Brown, argues that current AI development—relying on massive energy-intensive data centers—is unsustainable. He believes true intelligence should emerge from low-power, efficient systems, more aligned with biological computing.

Rao warns that the industry is headed toward “model collapse,” where large language models (LLMs) begin training on AI-generated content instead of real-world data, leading to compounding inaccuracies and hallucinations. He stresses the importance of grounding AI in reality and moving beyond brute-force scaling. Rao sees intelligence not just as a function of computing power, but as a distributed, observational system—“life is a learning machine,” he says—hinting at a need to fundamentally rethink how we build AI.

Learn more from The New Stack about the latest insights about the evolution of AI and neural networks: 

The 50-Year Story of the Rise, Fall, and Rebirth of Neural Networks

The Evolution of the AI Stack: From Foundation to Agents

Join our community of newsletter subscribers to stay on top of the news and at the top of your game. 

Fal.ai, once focused on machine learning infrastructure, has evolved into a major player in generative media. In this episode of The New Stack Agents, hosts speak with Fal.ai CEO Burkay Gur and investor Glenn Solomon of Notable Capital. Originally aiming to optimize Python runtimes, Fal.ai shifted direction as generative AI exploded, driven by tools like DALL·E and ChatGPT. Today, Fal.ai hosts hundreds of models—from image to audio and video—and emphasizes fast, optimized inference to meet growing demand.

Speed became Fal.ai’s competitive edge, especially as newer generative models require GPU power not just for training but also for inference. Solomon noted that while optimization alone isn't a sustainable business model, Fal’s value lies in speed and developer experience. Fal.ai offers both an easy-to-use web interface and developer-focused APIs, appealing to both technical and non-technical users.

Gur also addressed generative AI’s impact on creatives, arguing that while the cost of creation has plummeted, the cost of creativity remains—and may even increase as content becomes easier to produce.

Learn more from The New Stack about AI’s impact on creatives:

AI Will Steal Developer Jobs (But Not How You Think) 

How AI Agents Will Change the Web for Users and Developers 

Join our community of newsletter subscribers to stay on top of the news and at the top of your game. 

Traditional headless browsers weren’t built for AI agents, often breaking when web elements shift even slightly. Paul Klein IV, founder of Browserbase and its open-source tool Stagehand, is tackling this by creating a browser infrastructure designed specifically for AI control. On The New Stack Agents podcast, Klein explained that Stagehand enables AI agents to interpret vague, natural-language instructions and still function reliably—even when web pages change. This flexibility contrasts with brittle legacy tools built for deterministic testing. Instead of writing 100 scripts for 100 websites, one AI-powered script can now handle thousands.

Klein’s broader vision is a world where AI can fully operate the web on behalf of users—automating tasks like filing taxes without human input. He acknowledges the technical challenges, from running browsers on servers to handling edge cases like time zones and emojis. The episode also touches on Klein’s concerns with AWS, which he says held a “partnership” meeting that felt more like corporate espionage. Still, Klein remains confident in Browserbase’s community-driven edge.

Learn more from The New Stack about the latest insights in AI browser based tools: 

Why Headless Browsers Are a Key Technology for AI Agents 

Ladybird: That Rare Breed of Browser Based on Web Standards 

Join our community of newsletter subscribers to stay on top of the news and at the top of your game.

In a recent episode of The New Stack Agents livestream, Antje Barth, AWS Developer Advocate for Generative AI, discussed the growing developer interest in building agentic and multi-agent systems. While foundational model knowledge is now common, Barth noted that developers are increasingly focused on tools, frameworks, and protocols for scaling agent-based applications. She emphasized the complexity of deploying such systems, particularly around navigating human-centric interfaces and minimizing latency in multi-agent communication.

Barth highlighted AWS’s support for developers through tools like Amazon Q CLI and the newly launched open-source Strands SDK, which AWS used internally to accelerate development cycles. Strands enables faster, flexible agentic system development, while services like Bedrock Agents offer a managed, enterprise-ready solution.

Security was another key theme. Barth stressed that safety must be a “day one” priority, with built-in support for authentication, secure communication, and observability. She encouraged developers to leverage AWS’s GenAI Innovation Center and active open-source communities to build robust, scalable, and secure agentic systems.

Learn more from The New Stack about AWS' support for developers through tools that support multiple agents: 

Code in Your Native Tongue: Amazon Q Developer Goes Global 

AWS Launches Its Take on an Open Source AI Agents SDK

Amazon's Bedrock Can Now 'Check' AI for Hallucinations 

Join our community of newsletter subscribers to stay on top of the news and at the top of your game.