2026-03-17 16:42
🌐 AI in enterprise software is growing, but security risks are often overlooked. Developers must recognize that AI agents can access sensitive data and execute actions in real-time, potentially exposing vulnerabilities. Virtue AI’s new Agent ForgingGround offers tools to simulate adversarial attacks, helping identify these risks before deployment. Dynamic agents require ongoing security testing to prevent misuse and unauthorized actions. #Cybersecurity #AI #EnterpriseSoftware #DataProtection...
Source: The New Stack
Adrian Bridgwater
2026-03-17 16:00
KubeVirt is an open-source project that integrates virtual machines into the Kubernetes ecosystem, allowing teams to manage VMs and containers together. This approach simplifies orchestration and tooling across both workload types. Recent adoption is driven by organizations seeking to modernize their virtualization without vendor lock-in. Live migration is a key feature, enabling VMs to move between hosts with minimal downtime, which is essential for maintenance and resource optimization. For...
Source: The New Stack
Tiago Castro
2026-03-17 14:00
In the rush to adopt AI, many organizations are facing a new challenge: AI blind spot debt. This form of technical debt arises when companies struggle to implement lessons from DevOps and DevSecOps, leading to a “wait and see” approach. As AI usage expands beyond data science teams, the fragmented ecosystem makes it difficult to govern. Employees across departments are utilizing AI tools, often without proper security measures, creating potential risks. Understanding and managing this hidden...
Source: The New Stack
Yuval Fernbach
2026-03-17 13:00
🚀 Featherless has launched Managed OpenClaw, a solution designed for developers to run open-source AI models without server management. This new service offers a secure, sandboxed environment where the cost is a flat monthly fee, eliminating the unpredictability of token-based billing. As AI workflows evolve, the need for efficiency and predictability becomes crucial. Managed OpenClaw aims to simplify this process for developers. #AI #OpenSource #TechInnovation #Featherless #ManagedOpenClaw
Source: The New Stack
Adrian Bridgwater
2026-03-17 13:00
Agentic AI offers significant potential for SREs and AI/ML teams, yet many face challenges when moving from experimentation to production. Autonomous agents may lead to unpredictable behavior due to issues like hallucination and context misinterpretation. A control plane can address these challenges by enhancing coordination and providing essential observability. Join us on March 24 at 11 a.m. Pacific for a free webinar on scaling agentic AI effectively. Learn from experts how to implement...
Source: The New Stack
TNS Staff
2026-03-17 12:00
🔍 Is your database ready for AI integration? For 25 years, databases have been protected by various layers. Now, with the rise of AI agents, there’s a shift in how we approach database access. The Model Context Protocol (MCP) allows AI to connect to databases, but this poses security risks. The article discusses the need for DBAs to understand MCP and its implications for database security. It highlights that SQL can be very powerful in the wrong hands. Key advancements in MCP have improved...
Source: The New Stack
Dan Baskette
2026-03-17 11:00
Uber faced significant challenges in scaling its machine learning infrastructure as it transitioned from a luxury car service to a global logistics leader. In 2015, data scientists spent most of their time managing servers instead of building models. To address this, Uber developed Michelangelo, a centralized system designed to streamline the ML process. However, as the demand grew, it required a shift to a cloud-native Kubernetes architecture for better scalability. They implemented over 100...
Source: The New Stack
Eric Wang
2026-03-16 20:20
Nvidia has launched the Nemotron Coalition at its GTC conference, uniting several AI labs to develop shared base foundation models. 🤝 Members like Black Forest Labs and LangChain will contribute expertise and data, while Nvidia provides training on its DGX Cloud infrastructure. Their first project aims to create the Nemotron 4 family of models. Nvidia's VP, Kari Briski, emphasized the importance of collaboration, stating that building frontier models requires resources many can't afford...
Source: The New Stack
Frederic Lardinois
2026-03-16 20:05
Nvidia recently unveiled the Nvidia Agent Toolkit at its GTC conference, aimed at building secure and efficient autonomous agents. A key highlight is NemoClaw, which integrates the OpenClaw framework with added security features. This enterprise-grade solution is designed for seamless installation and enhanced privacy. NemoClaw can utilize various coding agents and Nvidia's tools, such as the Nemotron models and OpenShell, a new open-source security runtime. #Nvidia #AI #OpenSource #TechNews...
Source: The New Stack
Frederic Lardinois
2026-03-16 18:17
🚀 Cursor has developed a fleet of AI agents to enhance security within its codebase. This initiative addresses the limitations of traditional security tools that struggle to keep pace with rapid code changes. Travis McPeak, Head of Security at Cursor, highlights the need for precise monitoring to avoid irrelevant alerts and missed critical changes. The new agents can analyze code changes more effectively, focusing on meaningful insights. Cursor has released templates for four security agents:...
Source: The New Stack
Frederic Lardinois
2026-03-16 18:02
🚀 Anthropic has announced a temporary increase in usage limits for its Claude assistant during off-peak hours. This promotion lasts for two weeks, allowing users on all plans to access double the usual capacity on weekends and during specific weekday hours. 🗓️ The goal seems to be encouraging developers to engage more with Anthropic's products, rather than using the Claude API. This strategy also helps manage demand on their infrastructure. #AI #Anthropic #Claude #TechNews #DeveloperTools
Source: The New Stack
Paul Sawers
2026-03-16 14:04
AI workloads are challenging traditional Kubernetes observability. As organizations scale AI, the complexity of Kubernetes environments increases, leading to potential failures. Robust observability practices are essential for teams to anticipate issues, not just react. Join the free webinar on March 19 at 11 a.m. Pacific to learn AI-powered strategies for better Kubernetes management. Register now for actionable insights! 📅🔍💻 #Kubernetes #AI #Observability #Webinar #TechInsights
Source: The New Stack
TNS Staff
2026-03-16 13:33
📢 Anthropic has announced a significant pricing change for its Claude Opus 4.6 and Claude Sonnet 4.6 models. The 1-million-token context window is now generally available, with standard pricing replacing the previous premium rates for large prompts. This means developers can submit extensive queries without incurring higher costs. Both models support large datasets, enabling in-depth analysis of complex information like code repositories and legal documents. For more insights, check out the...
Source: The New Stack
Paul Sawers
2026-03-16 12:00
Long-running coding agents are evolving. They now run autonomously for hours, managing tasks and submitting pull requests. However, a key distinction arises: code writing is not the same as software engineering. Agents excel at recognizing patterns, but developers bring crucial context, strategy, and judgment to the table. Developers own tasks requiring deep understanding of product strategy, team dynamics, and risk evaluation. Agents can generate code, but they can't grasp the nuances of...
Source: The New Stack
Arjun Iyer
2026-03-15 16:00
Unlock the potential of vibe coding! 🌟 This approach allows developers to focus on design and outcomes using generative AI, shifting away from traditional coding syntax. By describing intent and functionality, you can create applications more intuitively. For example, instead of coding loops and functions, simply prompt AI to "build a responsive card list" for your tasks. 🖥️✨ Explore the basics and see how vibe coding can streamline your development process! #VibeCoding #GenerativeAI...
Source: The New Stack
Jessica Wachtel
2026-03-15 14:00
Ex-Snowflake engineers have launched Tower, a startup aimed at addressing challenges in data engineering. 🛠️ Tower focuses on the complexities of running Python data pipelines and AI applications in production, a crucial area often overlooked. Co-founder Serhii Sokolenko emphasizes the difficulties in packaging, deploying, and maintaining these applications. 💻 The platform integrates with coding assistants, allowing developers to manage their applications effectively while receiving feedback...
Source: The New Stack
Paul Sawers
2026-03-15 12:00
🌐 Platform teams and AI engineers are navigating a complex landscape in 2026, facing decision paralysis due to an influx of GPU options from NVIDIA and various specialized providers. The choice of cloud for AI workloads has expanded, requiring careful consideration of new categories. The split between training and inference workloads is growing, with inference expected to dominate future AI compute. A practical taxonomy of the AI cloud market has emerged, categorizing it into six distinct...
Source: The New Stack
Janakiram MSV
2026-03-14 20:00
AI-assisted coding tools were introduced with high expectations for solving team issues in software delivery. However, the challenges persist, as failures often mirror those seen in human teams. The article discusses how large software projects fail not just due to human shortcomings but due to inherent complexities. This complexity acts like gravity, affecting both human and AI agents. Interestingly, experiments show that coordinating multiple AI agents often complicates rather than...
Source: The New Stack
Steve Fenton
2026-03-14 18:00
Discover Tromjaro, a free-trade Linux distribution prioritizing user privacy. 🌐 Unlike traditional OS, Tromjaro does not track users or serve ads. Built on Manjaro, it features a custom version of Firefox, enhanced with tools like Privacy Badger and uBlock Origin for better security. Noteworthy additions include the HUD for streamlined app navigation and RiseupVPN for secure browsing. Explore Tromjaro for a trusted and ad-free computing experience! 💻🔒 #Tromjaro #Linux #Privacy #OpenSource...
Source: The New Stack
Jack Wallen
2026-03-14 16:00
🚀 TypeScript 6.0 Release Candidate is now available, marking a significant step forward for developers! This release serves as a bridge to TypeScript 7.0, which will feature a complete compiler rewrite in Go. Key updates include support for the Temporal API, types for Map.getOrInsert, and RegExp.escape, enhancing date management and string handling. However, several outdated options have been deprecated, including ES5 output and certain module formats. Notably, strict mode is now enabled by...
Source: The New Stack
Darryl K. Taft
2026-03-14 14:00
🚀 The Model Context Protocol (MCP) is becoming essential in the agentic AI stack, facilitating connections between AI models and external systems. As adoption increases, developers face challenges in production use. The project's 2026 roadmap aims to address these issues, focusing on improving scalability and connection handling. Key players like Anthropic, OpenAI, Microsoft, and others are already integrating MCP into their platforms. Stay tuned as MCP evolves! 🔧📈 #AI #ModelContextProtocol...
Source: The New Stack
Paul Sawers
2026-03-14 13:32
AI layoffs are increasing, with major companies like Atlassian and Meta attributing job cuts directly to AI automation. Atlassian announced a 10% workforce reduction, while Meta plans to cut 20%. Other tech giants, including Amazon and Oracle, are also facing significant layoffs, with over 45,000 tech jobs lost globally in March alone. 📉 The debate between the MCP toolchain and APIs is heating up, as some advocate for a shift in approach. Additionally, Mac Minis are becoming preferred for...
Source: The New Stack
Matthew Burns
2026-03-14 12:00
Andrej Karpathy recently shared a 630-line Python script on GitHub that autonomously conducted 50 experiments overnight. 🌙🤖 This initiative, called AutoResearch, automates the process of tuning machine learning models, allowing for efficient exploration of configurations. Key elements include an editable asset, a scalar metric for performance, and a time-boxed cycle for experiments. ⏳📈 These principles show promise beyond just ML training, highlighting a new approach to research methodology....
Source: The New Stack
Janakiram MSV
2026-03-13 19:26
🚀 Exciting news in AI security! NanoClaw has partnered with Docker to enhance the safety of AI agents by isolating them within MicroVM sandboxes. This collaboration aims to address the security flaws associated with OpenClaw by providing a more secure, minimalistic runtime environment. Docker's new Sandboxes allow each agent to operate in its own lightweight MicroVM, ensuring actions are confined and do not affect the host system. This approach enhances protection against potential...
Source: The New Stack
Steven J. Vaughan-Nichols
2026-03-13 18:33
🚨 Attention Android developers! Google is implementing a new verification plan that could impact alternative app stores. F-Droid, a leading open-source app repository, warns this change may threaten their existence and others like them. Starting September, developers in certain countries must register with Google to distribute apps, even outside the Play Store. This move includes a $25 fee and requires a government ID, raising concerns among developers about the implications for app...
Source: The New Stack
David Cassel
2026-03-13 12:00
The article discusses the architecture of agent memory systems in engineering teams. It highlights a dual approach: using a filesystem interface for agent interaction and a database for persistent storage. Key insights reveal that this combination, rather than choosing between the two, is essential for effective system design. Recent evaluations and implementations by teams like LangSmith demonstrate the practicality of this method, especially for coding agents. Understanding the distinction...
Source: The New Stack
Mikiko Bazeley
2026-03-12 20:22
🌐 Agentic AI systems can shift enterprise staff roles from execution to oversight and strategy. However, this shift carries significant risks. 🔍 Key concerns include loss of human control, security vulnerabilities, and unpredictable actions that may be hard to reverse. An early mistake can escalate quickly, complicating accountability. 🤖 As the field is new, understanding and managing these risks is essential. Testing and sharing knowledge among IT professionals are crucial for safe...
Source: The New Stack
Charles Humble
2026-03-12 20:00
🚀 Google has announced it will launch Chrome for ARM64 Linux devices in Q2 2026, addressing a long-standing gap for users of Arm-powered hardware. This release aims to integrate Google’s services into a seamless browsing experience, enhancing access to features like account sync and the Chrome Web Store. While Chromium has been available, the full Chrome browser has been missing for ARM64. Firefox has been a popular alternative during this wait. Google also plans to make Chrome available on...
Source: The New Stack
Frederic Lardinois
2026-03-12 19:54
AI governance is crucial as the technology continues to evolve. SurePath AI has introduced its MCP Policy Controls to enhance security in AI interactions. This new service aims to manage which Model Context Protocol (MCP) servers can be used, addressing potential risks like data leakage and supply chain attacks. As adoption accelerates, oversight remains a challenge. SurePath’s co-founder emphasizes the need for secure management of MCP tools to prevent misuse within organizations. 🔒💡🛡️...
Source: The New Stack
Adrian Bridgwater
2026-03-12 19:22
🚀 Exciting news for developers! Perplexity has expanded its API platform, introducing three new tools: Embeddings API, Agent API, and Sandbox API. These tools enhance access to the orchestration layer, streamlining workflows by unifying search, models, and execution. This means developers can now replace fragmented systems with a single API key. The new suite supports semantic similarity for better context understanding, making it easier to handle proprietary data at scale. #Perplexity #APIs...
Source: The New Stack
Meredith Shubel
2026-03-12 18:00
🌟 Anthropic's Claude has received a major update, enabling it to create interactive charts, diagrams, and visualizations on demand. This feature aims to enhance user understanding during conversations. 📊 The visuals are integrated into chats but are temporary and won’t be saved in the app. Users can request graphics, and Claude may also generate them based on context. While the feature shows promise, there are occasional inaccuracies and longer wait times for visuals. 📈 #AI #Anthropic #Claude...
Source: The New Stack
Frederic Lardinois
2026-03-12 16:21
📊 AI adoption is accelerating, with 95% of US companies now using generative AI. Organizations are experimenting with AI agents, which enhance operational efficiency. However, as deployment speeds up, governance becomes crucial to manage risks effectively. A strong framework should focus on: 1️⃣ People-first governance 2️⃣ Clear guardrails for AI actions Ensuring human oversight and defined responsibilities will help maintain control and accountability in AI operations. #AIGovernance...
Source: The New Stack
João Freitas
2026-03-12 13:00
Runpod's latest report reveals that Qwen has surpassed Meta’s Llama as the most-deployed self-hosted LLM. This shift highlights the growing influence of agentic AI services in the enterprise technology market. 🌐 Runpod's insights, derived from anonymized deployment logs, show real-world usage patterns, contrasting with public perceptions of major AI brands. Interestingly, Llama 4 has seen minimal adoption, while Qwen, developed by Alibaba Cloud, is gaining traction for its multi-modal...
Source: The New Stack
Adrian Bridgwater
2026-03-12 12:00
🚀 Gloo has launched Gloo AI Studio, an AI platform designed for faith-based organizations. This platform ensures values alignment and governance directly within its infrastructure. Faith groups seek AI models that reflect their beliefs, and Gloo provides a flexible solution compatible with various AI models, including those from OpenAI and Google. The platform also supports data integration, allowing organizations to upload their own materials, enhancing AI outputs while keeping latency low....
Source: The New Stack
Frederic Lardinois
2026-03-11 19:48
🚀 Galileo has launched Agent Control, an open-source platform for managing AI agents at scale. This tool enables organizations to set behavioral policies once and enforce them across all deployments. Key partners like AWS and CrewAI will offer this centralized governance framework. With AI agent usage expected to grow significantly, Agent Control aims to standardize guardrails, allowing real-time policy updates without downtime. For more information, check out the article! 📊🔍 #AI #Galileo...
Source: The New Stack
Steven J. Vaughan-Nichols
2026-03-11 18:30
🚀 Nvidia has launched the Nemotron 3 Super, a 120-billion-parameter AI model. This model features a 1-million-token context window, optimized for speed and efficiency. It is designed to support large-scale AI systems, contrasting with the earlier Nemotron 3 Nano, which focused on smaller tasks. 🔍 The Super model can be accessed on various platforms, including Google Cloud and Hugging Face. It utilizes advanced architecture for improved context tracking and has been trained on extensive...
Source: The New Stack
Frederic Lardinois
2026-03-11 17:52
🚀 Tetrate has launched "Built on Envoy," an open source marketplace for Envoy extensions. This platform aims to simplify Envoy adoption for cloud-native applications. Envoy serves as a service proxy that manages network traffic and enhances security. The marketplace provides ready-to-use extensions to help teams overcome common deployment challenges. However, Envoy adoption can face hurdles, such as security issues with Web Application Firewalls (WAF) and SAML integration. Tetrate's...
Source: The New Stack
Adrian Bridgwater
2026-03-11 17:38
Microsoft's VS Code team has shifted to weekly releases after a decade of monthly updates, thanks to AI innovations. Product Lead Pierce Boggan shared that AI tools now empower Product Managers to define and prototype features independently. This change has sparked mixed reactions within the community. Some view it as a game-changer, while others advocate for engineers to take the lead in utilizing these tools. The use of VS Code and GitHub Copilot is central to their workflow, enhancing...
Source: The New Stack
Darryl K. Taft
2026-03-11 16:57
🚀 Developers face a new challenge: Shadow Tech Debt, a term introduced by JetBrains to describe low-quality code generated by AI agents. These agents lack structural understanding, leading to fragmented and potentially harmful code in projects. JetBrains emphasizes that while these tools improve productivity, they may compromise code quality over time. 🛠️ To address this, JetBrains launched Junie CLI, a coding agent designed for better codebase intelligence and integration across various...
Source: The New Stack
Darryl K. Taft
2026-03-11 16:37
Hewlett Packard Enterprise (HPE) explores the concept of "self-healing" IT in their recent research. With increasing IT complexity, operations teams face challenges in prioritizing issues and enhancing reliability. AI-trained models can analyze infrastructure data, helping teams spot unusual behavior and identify changes early. This proactive approach can reduce alert fatigue and improve troubleshooting. HPE's whitepaper discusses the potential for autonomous, self-healing IT systems, aiming...
Source: The New Stack
Jennifer Riggins