Articles from Source: The-New-Stack

LLMs create a new blind spot in observability

2026-01-24 18:00
🔍 The rise of Large Language Models (LLMs) poses challenges for observability in SaaS products. Traditional metrics and logs often fall short in explaining LLM behaviors, which are unpredictable and constantly evolving. This creates a new blind spot for engineering teams. Key signals to monitor now include token usage, latency, error rates, and response quality. Understanding LLM performance requires new telemetry approaches to assess their behavior in production. #SaaS #AI #Observability...
Source: The New Stack
Shahar Azulay

Distributed apps platform Aspire supports JavaScript, Python

2026-01-24 17:00
🚀 Aspire, Microsoft’s open source development platform, now offers full support for JavaScript, TypeScript, Python, and Java. This change simplifies the process of building distributed applications, making it easier for developers to orchestrate their projects. Aspire is designed to support cloud-native apps and microservices, featuring a developer dashboard and curated components. Learn more about how to run JavaScript code in various scenarios! #Aspire #Microsoft #JavaScript #CloudNative...
Source: The New Stack
Loraine Lawson

How Claude Cowork helps developers spread the AI knowledge

2026-01-24 16:00
🚀 Anthropic has launched Claude Cowork, a new feature in Claude Desktop aimed at spreading AI knowledge beyond developers. This tool allows users to leverage local folders, enhancing collaboration across departments. It helps demonstrate the value of developers' AI expertise in various business areas. Currently in research preview for Pro and Max subscribers on macOS, Cowork is designed for non-developers but benefits all. #AI #TechInnovation #Collaboration #ClaudeCowork #Anthropic
Source: The New Stack
David Eastman

Vercel’s json-render: A step toward generative UI

2026-01-23 21:00
🚀 Vercel has introduced json-render, an open-source tool aimed at generative user interfaces (UI). This innovation allows AI to create web interfaces on demand. Developers can set parameters for AI interactions, enabling users to describe their needs in natural language. The AI generates and renders UI in real-time. Although currently experimental, json-render could transform how websites and applications are built, automating UI creation. Learn more about this disruptive technology! 🔍💻...
Source: The New Stack
Loraine Lawson

CNCF: Kubernetes is ‘foundational’ infrastructure for AI

2026-01-23 19:00
The CNCF's latest Annual Cloud Native Survey shows that 82% of container users are now running Kubernetes in production. This marks a significant shift in cloud native technology, which is increasingly seen as foundational for modern infrastructure. 🌐 Chris Aniszczyk, CNCF's CTO, highlights that Kubernetes supports scale and reliability, especially for AI systems. Notably, 66% of organizations use Kubernetes for generative AI workloads, focusing on deployment rather than model training. 🤖...
Source: The New Stack
Steven J. Vaughan-Nichols

From data to dollars: AI’s role in the modern GTM stack

2026-01-23 18:00
AI is reshaping the go-to-market (GTM) process, transforming how marketing, sales, and product teams work together to drive revenue. Traditional methods, reliant on CRM tools and email marketing, are facing challenges due to decreased deliverability and oversaturation. To succeed, marketers must adopt a "speed to market" approach, utilizing new AI-powered tools for efficient workflow automation. This shift emphasizes the importance of identifying the right intent signals for effective...
Source: The New Stack
Deirdre Mahon

Kubernetes 1.35 features that change Day 2 operations

2026-01-23 17:00
Kubernetes 1.35 introduces key features enhancing Day 2 operations for containerized applications. 🌟 The standout update is in-place pod resource updates, allowing seamless resizing of CPU and memory without pod restarts. This aids in managing stateful services and minimizing disruptions for AI/ML workloads. Additionally, the fine-grained supplemental group control enhances security by managing Unix group permissions for containers more effectively. 🔒 Explore these advancements for improved...
Source: The New Stack
Janakiram MSV

Developer proves AI agents can be reprogrammed via new exploit

2026-01-23 16:30
A recent demonstration highlights a new VS Code exploit that can rewrite AI agents in code repositories. 🖥️ The exploit involves a malicious tasks.json file that activates when developers open affected folders. It was identified by Oasis, who also provided mitigation strategies. Isaac Lewis showed how this can impact AI-native code editors like Cursor, potentially compromising developer tools. He cautioned that the exploit could extend to other code editors, raising concerns about...
Source: The New Stack
Loraine Lawson

More security tools are slowing down your incident response

2026-01-23 16:00
Time is critical in cybersecurity. The speed of detecting and responding to incidents significantly impacts an organization's security posture. 🕒 While multiple security tools may seem beneficial, they often slow down response times. Increased alerts from various systems can overwhelm analysts and complicate investigations. 📊 To improve efficiency, breaking down security silos is essential. This allows for better visibility and prioritization of threats, especially in critical sectors like...
Source: The New Stack
Raghav Iyer

Context is AI coding’s real bottleneck in 2026

2026-01-23 15:00
AI-generated code raises valid security and trust concerns, but the real bottleneck is context. 🛠️ The gap between what engineers know and what AI understands limits efficiency. Companies addressing this issue can reduce mistakes and technical debt. In 2026, new tools are expected to improve context transfer, helping teams ship faster. The focus will shift to capturing engineers' implicit knowledge for better AI integration. 🔄 #AICoding #SoftwareDevelopment #TechTrends #EngineeringLeadership...
Source: The New Stack
Greg Foster

The year of AI: 3 critical shifts coming to regulated industries

2026-01-23 14:00
Enterprises in regulated industries are making significant strides in AI adoption as they move into 2026. Key developments include: 1️⃣ Accelerated modernization of legacy systems, allowing organizations to update technology faster while staying compliant. 2️⃣ A shift to predictive security operations, enhancing threat prevention through AI monitoring. 3️⃣ Expanded development capabilities, improving efficiency across various business functions. These advancements are transforming how banks,...
Source: The New Stack
Bob Stevens

Docker logs got you down? Give this tool a try

2026-01-23 13:00
Struggling with Docker logs? 🐳 Troubleshooting can be tough, especially with many containers. A solution is Dozzle, an open-source web-based log viewer. It offers real-time monitoring, supports various log formats, and has a user-friendly interface. Deployment is straightforward on any machine running Docker. For a step-by-step guide on installation and setup, check out the full article! #Docker #DevOps #Troubleshooting #OpenSource #Dozzle
Source: The New Stack
Jack Wallen

CTO Chris Aniszczyk on the CNCF push for AI interoperability

2026-01-22 21:00
🚀 Chris Aniszczyk, CTO of CNCF, discusses the similarities between AI agents and microservices. He emphasizes the importance of being cloud native for AI development, especially with tools like Kubernetes. 🌐 The CNCF is working on a conformance program to ensure AI systems meet community standards, enhancing interoperability. 🛠️ Key projects include gRPC for communication and Prometheus for observability. #AI #Microservices #CloudNative #CNCF #KubeCon
Source: The New Stack
Loraine Lawson

The head chef model for AI-assisted development

2026-01-22 19:00
The relationship between developers and AI coding assistants is evolving. The "head chef" model suggests that developers will oversee AI "sous chefs" to handle coding tasks while they focus on design and quality control. This shift allows developers to break down problems into manageable tasks, enhancing their role as decision-makers. Effective context management is crucial for accurate AI outputs, requiring careful curation of input data. In real-world applications like building Kafka...
Source: The New Stack
Adi Polak

Why agentic LLM systems fail: Control, cost, and reliability

2026-01-22 18:00
Agentic AI systems like AutoGPT and BabyAGI have shown promise in automating complex tasks with minimal human input. 🤖 However, their real-world deployment faces challenges. Key issues include control, cost, and reliability. The very traits that make them appealing—adaptive decision-making and open-ended reasoning—can hinder performance under production constraints. Variability in outputs complicates testing and validation, as results can differ even with identical inputs. This...
Source: The New Stack
Ibrahim Kamal

Why platform companies keep buying frontend framework teams

2026-01-22 17:00
Cloudflare recently acquired Astro, a leading frontend framework, continuing a trend of platform companies buying framework teams. This movement began in 2021 with Vercel hiring Svelte's creator and includes notable acquisitions by Shopify, Netlify, and Automattic. Cloudflare aims to support independent websites amid challenges posed by AI, promoting “content-driven websites” and advocating for creators’ rights. #TechNews #FrontendDevelopment #Acquisition #Cloudflare #WebDevelopment 🌐💻✨
Source: The New Stack
Richard MacManus

Chainguard to eliminate FUD around Chainguard OS

2026-01-22 16:00
🚀 Exciting news from Chainguard! They have launched the Chainguard OS Fully User Directed Committee, also known as the FUD Committee. This initiative aims to eliminate fear, uncertainty, and doubt surrounding purpose-built Linux distros by focusing on customer needs. Chainguard OS will be governed by its users, enhancing stability and predictability. The committee will be composed of members from various industries and will influence technical decisions, align roadmaps, and provide strategic...
Source: The New Stack
Darryl K. Taft

Why enterprise AI breaks without metrics discipline

2026-01-22 15:00
AI is transforming workplaces by enhancing product iterations, operational efficiency, and customer support. 🚀 However, many companies face challenges in seeing consistent benefits from AI due to inconsistent data definitions across teams. This ambiguity can lead to unreliable AI systems and a lack of trust in their outputs. Implementing an intelligent metrics layer can standardize data definitions, providing a single source of truth. This clarity is crucial for effective AI adoption. 📊...
Source: The New Stack
Ritish Chugh

How Precog adds business context to make enterprise data AI-ready

2026-01-22 14:00
🚀 Extracting data from enterprise tools like Salesforce and NetSuite is straightforward, but preparing it for AI models is challenging. Precog's new feature aims to add essential business context to data extraction processes. 🔍 The company focuses on making enterprise data usable by automatically generating context to help models understand the information. This approach streamlines the data preparation process, making it faster and more efficient. 💡 Notably, Precog ensures that company data...
Source: The New Stack
Frederic Lardinois

Drowning in AI slop, cURL ends bug bounties

2026-01-22 13:00
Daniel Steinberg, founder of cURL, is ending the bug bounty program due to overwhelming low-quality submissions, termed "AI slop." He aims to reduce the influx of unreliable reports that hinder the review process. Steinberg has noted a significant decline in submission quality, prompting this decision. He plans to announce further details next week, coinciding with his talk on open source security and AI at FOSDEM. This move highlights ongoing concerns in the tech community regarding the...
Source: The New Stack
Steven J. Vaughan-Nichols

AI SRE roundtable: The creation of a new category

2026-01-21 23:00
The rise of AI SREs is transforming operations and security. 🛠️ With over 30 vendors now offering AI SRE solutions, the term is gaining traction as a key player in the tech landscape. Experts agree that while AI enhances efficiency, human oversight remains crucial. 👥 Kolton Andrus emphasizes the importance of accountability in decision-making. As AI tools evolve, organizations face challenges in navigating their utility and addressing skill gaps. 📈 Investing in training and fostering...
Source: The New Stack
Adam LaGreca

Enabling autonomous agents with environment virtualization

2026-01-21 22:00
🚀 The AI landscape is evolving from just smart suggestions to autonomous contributions. Tools like Claude Code and GitHub Copilot Workspace are now tackling complex tasks independently. However, code verification is becoming a significant bottleneck. For agents to truly enhance productivity, they must operate in a closed-loop system that includes real runtime data. Without this, human developers are left to troubleshoot integration errors, slowing down progress. As noted by Boris Cherny,...
Source: The New Stack
Arjun Iyer

Why d-Matrix bets on in-memory compute to break the AI inference bottleneck

2026-01-21 20:00
AI inference is becoming a major compute workload, highlighting the need for optimized hardware solutions. Companies like OpenAI face constraints due to compute availability, prompting a shift in focus. d-Matrix, backed by $275 million in funding, is innovating with its Corsair platform, moving away from standard GPUs to address memory bottlenecks. CEO Sid Sheth emphasizes the importance of differentiation in a market dominated by Nvidia. 🚀💻🔍 #AI #Inference #TechInnovation #dMatrix #Computing
Source: The New Stack
Frederic Lardinois

Should enterprises run containers in VMs or bare metal?

2026-01-21 19:00
Should enterprises run containers in VMs or on bare metal? 🤔 Containers offer flexibility and portability, but the choice of deployment is crucial. For enterprises, using virtual machines (VMs) can enhance operational flexibility and security compared to bare metal servers. VMs allow for multiple Kubernetes versions across clusters, minimizing disruption during updates. This flexibility is vital for performance and compliance. Additionally, VMs provide better security isolation, reducing the...
Source: The New Stack
Jelani Harper

How to scale GitOps without maxing out

2026-01-21 18:30
Scaling GitOps can lead to challenges, often referred to as the "Argo ceiling." 🚧 As Argo CD instances multiply, visibility and governance can suffer. Teams often rely on custom scripts, diverting attention from core projects. To address this, a centralized GitOps control plane can enhance management while retaining familiar tools. Join the free webinar on Jan. 20 to learn actionable strategies for effective scaling! 📅 #GitOps #DevOps #Webinar #ArgoCD #Kubernetes
Source: The New Stack
Vicki Walker

Building scalable agentic assistants: A graph-based approach

2026-01-21 18:00
🚀 The article discusses the challenges of building a scalable interface assistant for payments and transactions. Initial attempts with single-agent systems were ineffective due to organizational complexity. The breakthrough came with a graph-based orchestration model, allowing for a coordinated system of specialized agents. Key issues tackled include routing, context preservation, and scalability. Each domain has a dedicated supervisor node to manage requests efficiently. #AI #TechInnovation...
Source: The New Stack
Abhishek Sant

Hardened containers don’t fix a broken software supply chain

2026-01-21 17:00
The software supply chain is facing significant challenges, prompting a rise in hardened container offerings. While these containers enhance security, they don't address the root issues of sourcing and verifying software. The industry often focuses on post-production fixes rather than ensuring the integrity of software from the start. Relying on hardened images can lead to a false sense of security. To truly secure the supply chain, organizations need to prioritize building software from...
Source: The New Stack
Dan Lorenc

Focus on ‘Don’ts’ to build systems that know when to say ‘No’

2026-01-21 16:00
To improve AI systems, organizations should focus on integrating "don’ts" into their knowledge bases. Listing only policies leaves gaps, as AI may generate incorrect responses. Negative examples help prevent mistakes and inconsistencies, guiding AI on what to avoid. Additionally, decision trees can enhance AI judgment in complex scenarios, allowing for better navigation of exceptions. #AI #KnowledgeBase #DecisionMaking #RiskManagement #Innovation 🤖📊🚫
Source: The New Stack
AJ Sunder

How Rust does Async differently (and why it matters)

2026-01-21 14:00
🚀 Discover how Rust's async model differs from JavaScript, Python, and Go in the first part of a new series! In Rust, async operations are "pull-based," meaning they are lazy and only execute when explicitly awaited. This contrasts with the "push-based" approach of other languages. The series will cover key concepts, including the poll-based model, pinning, executors, and practical async patterns. Stay tuned! #RustLang #AsyncProgramming #TechEducation #ProgrammingLanguages #SoftwareDevelopment
Source: The New Stack
Anshul Gupta

Vibe coding could cause catastrophic ‘explosions’ in 2026

2026-01-20 23:00
Concerns are rising about vibe-coded applications, as experts predict potential issues in 2026. David Mytton, a software security expert, warns that many organizations may face catastrophic problems if AI-developed software isn’t thoroughly reviewed. He likens the risks to the Challenger disaster, emphasizing the importance of understanding the code being deployed. While vibe coding can be useful for quick prototypes, developers must be cautious of its limitations. #AIDevelopment...
Source: The New Stack
Darryl K. Taft

Are your CI/CD pipelines accidentally increasing technical debt?

2026-01-20 20:00
🚀 Are your CI/CD pipelines adding to your technical debt? Internal developer portals can streamline DevOps, but quick setups may lead to operational challenges. The "scaffolding trap" creates complexities in scaling as cloned templates drift over time. Join the free webinar on February 3 to learn how to effectively manage your CI/CD pipeline strategy and avoid operational paralysis. #DevOps #CICD #TechnicalDebt #Webinar #SoftwareDevelopment
Source: The New Stack
Vicki Walker

5 Tech Predictions for 2026: From AI Inference to Kubernetes

2026-01-20 18:00
🌟 Exciting tech advancements are on the horizon for 2026! 1️⃣ Generative AI continues to enhance software development, becoming essential for coding teams. 2️⃣ Agentic AI is transforming enterprise operations, automating tasks and driving efficiency. 3️⃣ Enterprise AI is integrating more closely with systems of record, creating virtual workers for repetitive tasks. 4️⃣ Modernization is accelerating as organizations seek alternatives to costly legacy systems, with Kubernetes leading the way....
Source: The New Stack
Venkat Ramakrishnan

WebAssembly vs. Testing Side-by-Side Performance

2026-01-20 17:00
🌐 WebAssembly (Wasm) is gaining attention for its performance advantages, often described as “near-native speed” compared to JavaScript. While JavaScript excels in UI tasks, Wasm is better suited for heavy data processing. This article discusses a project comparing Wasm and JavaScript performance using image processing tasks. The project involves converting images to grayscale and applying a sharpen filter, showcasing Wasm's capabilities in real-world scenarios. For developers interested in...
Source: The New Stack
Jessica Wachtel

AI Agents or Skills? Why the Answer Is ‘Both’

2026-01-20 16:00
Anthropic introduces a new concept: Agent Skills. These modular bundles of expertise enhance AI agents, allowing them to adapt without changing core logic. Agents faced limitations with prompt-based systems, often leading to confusion. By integrating skills, teams can improve domain capabilities without risking regressions. This approach promotes efficiency and clarity in AI workflows. #AIAgents #AgentSkills #TechInnovation #MachineLearning #DigitalTransformation 🤖📈🔧
Source: The New Stack
Pete Hampton

How To Build Production-Ready AI Agents With RAG and FastAPI

2026-01-20 15:00
🚀 Agentic AI is evolving from demos to real-world applications like research assistants and compliance copilots. This article outlines a tutorial for building production-ready AI agents using RAG and FastAPI. Key focuses include reliability, observability, and cost-awareness. The tutorial covers essential steps like defining robust tool interfaces, setting up RAG, and implementing guardrails for safety. Discover how to create effective, scalable AI workflows that can operate reliably in...
Source: The New Stack
Oladimeji Sowole

Mageia Harkens Back to the Glory Days of Mandrake Linux

2026-01-18 15:00
🚀 Mageia brings back the spirit of Mandrake Linux, a beloved distribution from the '90s! After the transition from Mandrake to Mandriva, Mageia emerged as a robust alternative. Its slogan, "build great tools for people," reflects its user-friendly approach. Mageia offers a live distribution experience, allowing users to test it without installation. The setup is slightly more complex than some alternatives, but it upgrades during installation, saving time later. Explore the KDE Plasma...
Source: The New Stack
Jack Wallen

Repair Advocates Name CES 2026’s Most Anticonsumer Tech

2026-01-18 14:00
The Repair Association highlights the lack of consumer-friendly practices at CES 2026. For the fifth year, they called out the "Worst in CES," focusing on products that are least repairable and sustainable. This year's event involved groups like Consumer Reports and iFixit, aiming to raise awareness about these issues. Executive Director Gay Gordon-Byrne emphasized the marketing hype over genuine innovation, questioning the value of faster chips in everyday products. The advocacy extends to...
Source: The New Stack
David Cassel

Astro Redesigns Its Development Server

2026-01-17 15:00
🚀 Astro recently released the beta version of Astro 6, featuring a redesigned development server. This update includes new APIs for content security policies, fonts, and live content collections. Astro now runs web applications in the same runtime as production, enhancing stability across various platforms, including Cloudflare Workers. In related news, Waku updated its Cloudflare Worker adapter with automatic selection and static deployment support. Additionally, Svelte addressed five...
Source: The New Stack
Loraine Lawson

The Future of AI in SRE: Preventing Failures, Not Fixing Them

2026-01-17 14:00
The future of Site Reliability Engineering (SRE) is shifting towards preventing failures rather than just fixing them. As systems grow more complex, traditional reactive methods are no longer sufficient. SRE has evolved through three stages: alerting, AI-assisted triage, and safe auto-remediation. Each stage has improved recovery times, but all required failures to occur first. The next step emphasizes a preventative approach. By analyzing historical data, AI can identify patterns that...
Source: The New Stack
Asaf Savich

A Developer’s Guide to Marshaling Data With JSON

2026-01-17 13:00
Understanding data representation is crucial for developers. 🖥️ The article discusses JSON (JavaScript Object Notation) as a streamlined format for structuring data. Unlike XML, JSON is simpler and more efficient for marshalling data, making it ideal for saving program states, like game positions. 🎮 The post also highlights JSON's readability, which allows non-programmers to understand data structures easily. It’s often used in configuration files where flexibility is key. 📄 Explore how JSON...
Source: The New Stack
David Eastman