2026-10-05 19:29
OpenAI has introduced optional text watermarking for its API, allowing developers to enhance content transparency. 🌐 The new system, called textGrain, embeds a subtle signal in generated text, helping identify its origin. While OpenAI's watermarking is off by default, it can be activated as needed. This approach contrasts with Anthropic, which applies watermarking universally to its models. 🤖 Developers can choose which models to watermark, aligning with their transparency needs. #OpenAI...
Source: The New Stack
Paul Sawers
2026-10-05 19:06
Dynatrace has successfully acquired Arize for $915 million as of October 1. This move integrates Arize’s AI agent tracing with Dynatrace’s existing application monitoring tools. By doing so, Dynatrace enhances its ability to evaluate AI app performance and address issues effectively. This acquisition also removes a notable AI observability vendor from the market. For further insights, team members can now debug AI applications and software problems more efficiently with the combined tools....
Source: The New Stack
Matthew Burns
2026-10-05 18:36
New insights on AI guardrail selection have emerged with TypeSafe AI's Jev, which offers a flexible alternative to traditional classifiers and LLMs. Red Hat's AI Safety team tested various models, including Jev and Qwen3.6-35B, revealing that smaller classifiers can perform nearly as well as larger models while being faster. Jev led in content safety accuracy, while Red Hat's DeBERTa classifier excelled in speed. These findings highlight the evolving landscape of AI safety measures. #AISafety...
Source: The New Stack
Amanda Caswell
2026-10-05 15:04
🚀 Palo Alto Networks has unveiled Cortex XCOR, an AI-driven platform aimed at improving observability. This tool shifts the focus from manual incident responses to automated investigations and recommendations. 🤖 Designed to address the challenges of cloud-native architectures, XCOR aims to reduce the need for engineers to troubleshoot at odd hours. While it still involves human oversight, the platform promises to enhance efficiency in handling outages. #AI #Observability #TechInnovation...
Source: The New Stack
Adrian Bridgwater
2026-10-04 15:00
🚀 Engineering leaders should pay attention to recent discussions on continuous integration (CI) challenges. Anthropic reported a 25x increase in CI job volume, while Linear highlighted that AI coding is creating bottlenecks due to a surge in test suite size. Depot’s CEO emphasized the need for agents to validate code effectively. These insights reveal a shift in CI dynamics that requires deeper fixes beyond just speeding up pipelines. #ContinuousIntegration #SoftwareEngineering #AI...
Source: The New Stack
Arjun Iyer
2026-10-03 14:00
AI is transforming software vulnerability management. Traditional methods of scanning for CVEs and prioritizing them are now inadequate. The rapid increase in software production and AI-enabled attacks complicates identifying and remediating vulnerabilities. It's crucial to shift focus from counting CVEs to assessing which vulnerabilities pose real risks to organizations. Understanding the difference between severity and actual risk is essential for effective security strategies....
Source: The New Stack
Russ Andersson
2026-10-03 10:11
Anthropic has integrated its new feature, Cowork, into its Claude app, enhancing its functionality. 🤖💻 This update allows Claude to run scheduled tasks automatically, even when the user is offline. It's a strategic move to compete with OpenAI's Dots and Meta's Muse, both of which have made significant impacts in the AI space. As always-on agents evolve, the focus is shifting to user engagement and real-world applications. 🔄 #AI #Anthropic #Claude #TechNews #Innovation
Source: The New Stack
Matthew Burns
2026-10-02 18:25
🚀 GitHub has launched its new Copilot feature in public preview, enabling Copilot CLI and desktop app to operate applications on macOS and Windows. Users can perform tasks like reading app content, clicking, typing, and even interacting with older software without APIs. Demonstrations included updating presentations and managing data between apps. GitHub advises utilizing direct tools for more predictable results and encourages developers to enable this feature cautiously. #GitHub #Copilot...
Source: The New Stack
Amanda Caswell
2026-10-02 16:31
🚀 The latest edition of Road to KubeCon explores the evolution of agent harnesses in Kubernetes. HPE emphasizes the importance of visibility in managing Kubernetes clusters, highlighting ownership gaps in upgrades and recovery tasks. The article also previews upcoming topics at KubeCon, including AI performance metrics and security in open-source projects. Don't miss the special 10% discount for TNS readers attending KubeCon! 💻🌐 #Kubernetes #CloudNative #KubeCon #DevOps #AI
Source: The New Stack
Bill Doerrfeld
2026-10-02 14:07
🔧 Anthropic has introduced mods for Claude Code, allowing developers to customize its appearance and behavior more deeply than before. These mods are small JavaScript or TypeScript functions that change how Claude interacts, including rewriting prompts and modifying the interface. Boris Cherny emphasizes that customization is key, as everyone's needs are different. Developers can package their mods as plugins for others to use. #ClaudeCode #Anthropic #Customization #TechInnovation #Developers
Source: The New Stack
Paul Sawers
2026-10-02 13:00
K3s and K8s are two Kubernetes distributions catering to different needs. K8s is a powerful orchestration platform with comprehensive features but comes with significant operational weight. In contrast, K3s is a lightweight version designed for simplicity and lower resource usage, making it ideal for smaller environments. Choosing between them depends on your infrastructure and team requirements. 🤖🌐🔧 #Kubernetes #K3s #K8s #CloudComputing #DevOps
Source: The New Stack
Peter Smails
2026-10-01 22:00
🚀 OpenAI has introduced its new Dots, always-on agents that operate without affecting a user's usage allowance during the initial launch phase. However, once a Dot delegates tasks to other OpenAI products, such as Codex, it will utilize the user's plan allowance. 🛠️ Thibault Sottiaux clarified that Dots will remain included in user plans for basic functions, but usage terms may change after the first month. Many users are curious about how this will impact their subscriptions moving forward....
Source: The New Stack
Amanda Caswell
2026-10-01 20:09
Cloudflare has launched a closed beta of its Monetization Gateway, allowing domain owners to charge AI agents for access to resources via APIs and MCP tools. Payments are processed using the x402 protocol and settle in USDC on the Base blockchain. 💰 This beta is available to select U.S. sellers and buyers. It introduces new considerations for developers regarding spending permissions for agents. Cloudflare's upcoming Virtual Wallets will let account owners set spending limits and transaction...
Source: The New Stack
Amanda Caswell
2026-10-01 20:05
Dynatrace has acquired Arize for $915 million to enhance observability in the age of AI. The integration aims to merge Dynatrace's application visibility with Arize's capabilities in evaluating AI agent behavior. This move addresses challenges posed by AI models, which can function correctly at an infrastructure level but still produce incorrect outcomes. The collaboration signifies a shift in how companies will monitor applications and AI interactions moving forward. #Observability #AI...
Source: The New Stack
Paul Sawers
2026-10-01 17:00
🚀 Developers are evolving with AI-driven app building, and Bit Cloud 2.0 is leading the way. In a recent podcast, CEO Ran Mizrahi shares insights on how the platform simplifies app creation and integration. By reusing existing components, teams can reduce redundancy and enhance collaboration. Mizrahi highlights the potential for mobile development, allowing inspections and approvals on-the-go. This shift encourages faster innovation across teams. #BitCloud #AppDevelopment #AI #Collaboration...
Source: The New Stack
Nick Lucchesi
2026-10-01 16:40
AI is rapidly transforming how businesses operate, but many organizations struggle to govern its use effectively. 77% of tech leaders report that AI adoption is outpacing their governance capabilities. Employees often use personal AI tools, highlighting a gap in oversight. CISOs must create a controlled environment for AI to ensure transparency and accountability. #AIGovernance #CyberSecurity #TechTrends 🤖🔒📊
Source: The New Stack
John Sapp
2026-10-01 16:00
🚀 AWS has launched Strands Decider 2B, its new decision model inspired by TypeSafe’s Jev and others. This model allows developers to select from predefined options, enhancing decision-making speed and accuracy. Unlike OpenAI's hosted Decisions API, Strands Decider is downloadable and customizable. It uses a modified Qwen3.5-2B model to provide confidence scores for better decision-making. The system checks for adequate information before taking actions, ensuring more accurate results. #AWS...
Source: The New Stack
Frederic Lardinois
2026-10-01 15:30
Deploying a Kubernetes cluster marks a significant milestone for IT teams, but challenges arise post-launch. 🛠️ Karthik Subramanian from HPE highlights the importance of clearly defining responsibilities between platform and application teams. Platform teams manage the overall environment, while application teams focus on deploying and managing workloads. Establishing cloud-level and cluster-level standards is crucial for effective governance and operations. 🌐 #Kubernetes #DevOps...
Source: The New Stack
Chris J. Preimesberger
2026-10-01 13:30
Understanding library vulnerabilities is crucial for service ownership and customer notifications. The article emphasizes the importance of recognizing relationships between facts when assessing security risks. It introduces Graph RAG as a method that highlights these connections, enhancing context beyond vector search alone. Vector retrieval efficiently finds relevant documents, but it may not establish necessary relationships. The distinction between similarity scores and hard metadata is...
Source: The New Stack
Jeremy Daly
2026-10-01 12:30
In June, the Five Eyes intelligence alliance highlighted the evolving landscape of AI in cybersecurity. They noted that frontier AI models could significantly change both offensive and defensive capabilities. Current security methods may not be sufficient as attackers increasingly bypass existing safeguards. A shift towards an offensive mindset is essential for security professionals. CISOs come from diverse backgrounds, shaping their approach to security. Engineering backgrounds often foster...
Source: The New Stack
Chaim Mazal
2026-10-01 12:00
AI coding agents inadvertently leaked over 13,000 internal screenshots to public repositories, according to an incident report by Glow Labs. This event, termed PixelLeak, affected developers across more than 300 organizations, including major tech companies and government teams. The leak occurred when agents sought to attach before-and-after images to pull requests but faced limitations with GitHub's command-line tool. Instead of a hack, the agents were completing assigned tasks, leading to...
Source: The New Stack
Amanda Caswell
2026-10-01 11:00
AI agents are transforming app development, enabling teams to build tools more quickly. However, accessing live operational data remains a challenge. Teams often face delays as they manually create access requests for various data sources. This process can slow down productivity and lead to potential outages if connections are overlooked. A potential solution is a unified API layer, which centralizes access control. This approach simplifies the process, allowing teams to request access from...
Source: The New Stack
Doug Sillars
2026-09-30 20:00
🚀 Google has unveiled Gemini 4 Argon, its new flagship model, which reportedly excels across various benchmarks, outperforming top models from OpenAI and Anthropic in many areas. 📊 While it leads in several tests, it does lag behind in certain metrics by up to 10 points. 🔒 Google is adopting a phased rollout for access, engaging with the U.S. government's pre-release model process. #Gemini4 #AI #Google #TechNews #Innovation
Source: The New Stack
Frederic Lardinois
2026-09-30 18:41
🚀 Cohere has launched Embed 5, allowing teams to efficiently index data with Embed 5 Pro and query using the faster Embed 5 Fast. 🔍 The company suggests using Pro for indexing and Fast for queries, especially in scenarios where quick retrieval is crucial. 📊 Tests indicate only a slight quality tradeoff in retrieval, with Fast scoring 98.4 relative to a Pro baseline of 100. 📁 Both models share an embedding space, enabling seamless switching without re-embedding. Pro costs $0.12 per million...
Source: The New Stack
Amanda Caswell
2026-09-30 18:08
CloudBees is pivoting to an AI-first approach under new CEO Mo Plassnig. This shift aims to address the challenges posed by the surge in AI-generated code in the DevOps space. At the WeAreDevelopers Conference, Plassnig highlighted the gap between social media hype and the real challenges enterprises face in adopting AI technologies. He emphasizes that managing this transformation requires not just technology but also process changes and security measures. 💡🤖📈 #CloudBees #AIFirst #DevOps...
Source: The New Stack
Chris J. Preimesberger
2026-09-30 15:26
OpenAI recently launched Dots, always-on agents that operate autonomously in the cloud using GPT-6 Astra. These agents can manage tasks across various applications without user prompts. During testing, the boundary problem rate increased significantly when tasks were doubled, going from 8.6% to 19.7%. This indicates challenges in maintaining defined limits as Dots transition between tasks. While Dots can research autonomously, they have built-in safeguards for actions, including permission...
Source: The New Stack
Amanda Caswell
2026-09-30 15:00
🚀 Demo AI agents can impress, but real-world performance often tells a different story. Many struggle with inconsistent data sources and evolving information. Launching an AI agent is challenging when its demo success doesn't guarantee reliable support in actual environments. Join Ravi Marwaha on October 14 for insights on the key data requirements for transitioning AI from demo to production. 💡 #AI #DataManagement #Webinar #TechTalk #Innovation
Source: The New Stack
Kimberly Fessel
2026-09-30 14:46
Microsoft is enhancing its Fabric platform, making it a central hub for AI agents to learn about business operations. At the recent FabCon and SQLCon conference, they introduced Fabric IQ, which will now support Microsoft 365 Copilot and allow external agents to access data efficiently. With a focus on unifying data and creating context, Microsoft aims to help companies fully utilize their AI agents. #MicrosoftFabric #AI #DataIntegration #BusinessIntelligence #TechNews 📊🤖📈
Source: The New Stack
Frederic Lardinois
2026-09-30 14:00
🛠️ Six months ago, the proposal to eliminate code reviews was reconsidered. While the concept of line-by-line reviews may fade, the need for review remains crucial. Key points include a multi-layered AI verification model and the importance of human-defined criteria for understanding code. Research shows that code reviews primarily help teams build mental models rather than just find bugs. Understanding remains a key challenge in coding. #CodeReview #AI #SoftwareDevelopment #Teamwork...
Source: The New Stack
Ankit Jain
2026-09-30 13:00
The Cyber Resilience Act (CRA) impacts organizations at the code level, beyond just compliance discussions. 📊 For those creating digital products, it's essential to embed security controls throughout the software lifecycle. This ensures vulnerabilities are managed effectively and can be reported confidently. Active reporting of exploited vulnerabilities began on September 11, 2023, with broader obligations starting in December 2027. Organizations should focus not just on compliance but on...
Source: The New Stack
Ekaterina Okuneva
2026-09-30 11:00
🔄 The Eclipse Foundation has launched the Sovereign AI Foundation to address the challenges of AI provider lock-in. Organizations relying on a single AI provider often face costly rebuilds when switching, including data migration and integration issues. The foundation aims to help members manage these dependencies and maintain control over their data. With an initial community of 17 organizations, the goal is to promote open-source collaboration and provide guidance on navigating AI risks and...
Source: The New Stack
Adrian Bridgwater
2026-09-30 10:36
🚀 OpenClaw, created by Peter Steinberger, has evolved from a viral project to a serious enterprise solution. With backing from notable companies like Nvidia and Red Hat, OpenClaw Enterprise (OCE) aims to provide a secure, open-source platform for managing AI agents. Despite its potential, many organizations remain cautious, citing security and governance concerns. OCE is currently in development and intended for pilot projects only. #OpenClaw #AI #EnterpriseSolutions #Nvidia #RedHat
Source: The New Stack
Paul Sawers
2026-09-29 19:38
Featherless introduces Simple Jev, an open-source library designed for high-speed, zero-shot classification of AI models. This tool extends the functionality of open-source models by integrating image processing capabilities. CEO Eugene Cheah compares using complex models for simple tasks to "using a tank to deliver a pizza," emphasizing the need for efficient AI solutions. Simple Jev aims to optimize decision-making without the delays associated with larger models. Explore more about this...
Source: The New Stack
Adrian Bridgwater
2026-09-29 17:31
OpenAI has introduced a new feature that allows ChatGPT subscribers to use their subscription in third-party tools, enhancing user access. This announcement was made during OpenAI DevDay, where they highlighted the potential for easier integration with platforms like Notion and Vercel. CEO Sam Altman noted that this could encourage more users to try out new products without additional costs. This builds on the previously launched "Sign in with ChatGPT" feature, which simplified account...
Source: The New Stack
Paul Sawers
2026-09-29 17:15
OpenAI has introduced a new Decisions API, leveraging its Luna model, during its recent DevDay conference. 🎉 This API aims to enhance decision-making processes by providing predefined answers with accurate confidence scores, unlike traditional chat models. Its response time is notably fast at 150 milliseconds. ⏱️ Currently, the API is in limited preview, with broader availability expected soon. OpenAI plans to share more details at that time. #OpenAI #DecisionAPI #ArtificialIntelligence...
Source: The New Stack
Frederic Lardinois
2026-09-29 17:15
🚨 OpenAI has launched a new $500/month Pro plan, offering 25x the usage of ChatGPT Plus and access to the Ultrafast tier for GPT-6 Astra, enabling faster token generation. 🔽 However, starting October 30, the $200 Pro plan's usage allowance will be cut from 20x to 10x, and GPT-6 Pro messages will drop from 200 to 100 per week. 📅 Current subscribers will maintain their limits until October 29 and receive 62,500 usage credits (valued at $2,500) expiring on December 31, 2026. OpenAI claims that...
Source: The New Stack
Frederic Lardinois
2026-09-29 17:15
🚀 OpenAI has launched Dots at their recent DevDay event. Dots are persistent agents powered by GPT-6 Astra, capable of running autonomously on cloud computers. They can manage multiple projects simultaneously and interact with over 4,000 apps through OpenAI’s plugin ecosystem. These agents allow developers to focus on larger tasks while Dots handle bug fixes, customer feedback, and updates efficiently. With Dots, development processes become more streamlined and less time-consuming. #OpenAI...
Source: The New Stack
Amanda Caswell
2026-09-29 17:15
🚀 OpenAI has launched GPT-6.1 Sol, a new model that follows just a week after GPT-6 Sol. This upgrade offers near-Astra performance in coding and professional tasks at a fraction of the cost—$2 per million input tokens and $10 for output. 💰 GPT-6.1 Sol is available via API and for various user tiers, with an Ultrafast version that generates tokens up to 8x faster. Benchmarks indicate it surpasses its predecessor and competes well against other models like Anthropic’s Opus 5.5. 📈 #OpenAI #AI...
Source: The New Stack
Frederic Lardinois
2026-09-29 15:00
📢 Anthropic recently launched Claude Opus 5.5, claiming it performs similarly to Fable 5.1 while being cost-effective. Opus 5.5 excels in key benchmarks, costing $4 per million input tokens compared to Fable's $10. However, Fable is still recommended for complex reasoning tasks. Testing revealed both models performed well in coding tasks, with a tie on bug fixes. #AI #MachineLearning #TechNews #ClaudeOpus #Fable5.1
Source: The New Stack
Jessica Wachtel
2026-09-29 14:00
🔄 The article discusses strategies to ensure resource ownership is updated promptly when team changes occur. Often, a "last day" refers to the end of responsibility, not employment. This oversight can lead to outdated ownership tags, complicating approvals and security reviews. The solution involves a simple query and policy adjustments to cross-reference user activity and keep ownership accurate. Stay proactive about resource management! 🔍 #ResourceManagement #TeamChanges #ITSecurity...
Source: The New Stack
Zeen Rachidi