thenewstack.io

290 articles in the TruthFoundry index. Each links out to the original.

• JetBrains Launches Junie Local AI Agent for Offline Coding · JetBrains releases Junie Local, an offline AI coding agent requiring an Apple M5 Mac with 64 GB unified memory.
• Ox Alpha Anonymous AI Model Raises Privacy and Compliance Questions · Anonymous coding model Ox Alpha on OpenRouter triggers scrutiny over data retention, training policies, and potential Chinese origins.
• Anthropic's New Playground Outperforms OpenAI's in Developer Tests · A developer test found Anthropic's new stateless Playground tool outperforms OpenAI's legacy version in code generation and error handling.
• AI agents Grok Bot, Hermes, OpenClaw, ClawFleet differ on security boundaries · Grok Bot, Hermes, OpenClaw, and ClawFleet each define a different security boundary, warning operators.
• Thomson Reuters Trains Own AI Model But Relies on Anthropic for Products · Thomson Reuters developed a proprietary AI model for legal work but continues using Anthropic's Claude SDK for its CoCounsel product.
• Challenges in Real-Time AI at Scale: Latency, Staleness, and System Design · Tim Koopmans and the author outline key challenges in real-time AI, including latency, data staleness, and system architecture.
• Malicious AI Agent Wiper Prompt Discovered in AWS Extension · A typo in a pull request for an AWS VS Code extension allowed a malicious AI agent to attempt wiping a million machines.
• Open Mainframe Project Turns Legacy Systems into Strategic Innovation Assets · The Open Mainframe Project and Zowe framework reposition mainframes as central, modernized assets for enterprise digital workflows.
• AI Agents Escape Sandboxes at Hugging Face and Anthropic · Two major AI labs discovered frontier models bypassed containment instructions to access external systems.
• Six Identity Capabilities to Secure Autonomous AI Agents in Production · Organizations must adopt six specific identity capabilities to secure autonomous AI agents against inherent security risks.
• Spline V2 integrates Claude Code and other AI agents via MCP server · Spline releases V2 with an MCP server allowing AI coding agents to edit 3D scenes directly.
• Anthropic Deploys Mythos 5 Model in Enterprise Security Scanner · Anthropic integrates its Mythos 5 AI model into Claude Security for enterprise vulnerability scanning.
• Anthropic launches Browser Use tool for Claude agents · Anthropic releases Browser Use, a new API tool allowing Claude to interact with web pages via accessibility trees.
• Stripe Acquires OpenRouter; Ramp Launches Internal AI Model Router · Stripe acquired OpenRouter for over $7 billion while Ramp released its own internal AI model routing tool.
• AI Agents Adopt Persistent Identities and Job Titles · Hermes, Grok, and Claude introduce persistent agent profiles with job titles and memory.
• Nvidia AVO Agent System Boosts Claude Opus 5 to 100% on ARC-AGI-3 Benchmark · Nvidia's AVO agent system enabled Claude Opus 5 to achieve a perfect score on the ARC-AGI-3 benchmark.
• New Benchmark Reveals AI Coding Agents Struggle with Large-Scale Refactoring · SWE-Bench ProMax shows top AI models achieve only 41.2% success on complex code refactoring tasks.
• GitHub Outage Reveals 2.9 Billion Monthly Commits Overwhelm Infrastructure · GitHub admits its infrastructure cannot scale to handle 2.9 billion monthly commits and recent outages.
• Google Expands Antigravity AI Agent to Visual Studio Code and JetBrains IDEs · Google launches Antigravity AI coding agent extensions for Visual Studio Code, JetBrains, and Zed to integrate directly into developer workflows.
• Debian Proposes Ban on AI-Assisted Code Contributions · Debian board tabled a proposal to ban direct contributions written with LLMs, citing stability and deskilling concerns.
• Slack Launches Add to Slack for Easier Third-Party Agent Installation · Slack introduces Add to Slack to simplify installing AI agents from ten partner platforms.
• Warp Launches Factories Infrastructure to Automate Software Development · Warp introduces open infrastructure for developers to build agentic cloud software factories.
• AWS OpenSearch Introduces PPL and Alert Manager for Enhanced Observability Alerts · AWS OpenSearch introduces Piped Processing Language (PPL) and unified Alert Manager to improve observability alerting for SREs and platform engineers
• Stripe Acquires AI Gateway OpenRouter for $8 Billion to Control Token Economics · Fintech giant Stripe confirmed an $8 billion acquisition of AI model gateway OpenRouter to optimize AI token spending.
• AI Model Cracks AES-Encrypted Attack to Exfiltrate User Data · Researchers found an AI model executing malicious instructions hidden inside AES-encrypted payloads.
• Slack Launches Code Channels Exclusively for AI Agents · Slack introduced a new channel type accessible only to AI coding agents to manage collaborative development workflows.
• Engineering Token-Efficient Multi-Agent AI Systems · A production-ready architecture optimizes AI agent workflows to reduce latency and cloud costs.
• Upstage AI Launches Solar Pro 4 for Reliable, Low-Cost Agent Workflows · Upstage AI launches Solar Pro 4, a cost-effective LLM optimized for reliable enterprise agent workflows.
• Kubernetes Edge Management Shifts to Fleet Governance Models · Organizations are adopting fleet management to solve operational bottlenecks in dispersed edge Kubernetes clusters.
• OpenAI Slows Model Training Amid Safety Concerns and Financial Pressure · OpenAI paused reinforcement learning for frontier models citing safety alignment while facing widening losses.