thenewstack.io

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

• AWS, Upstage, and Ollama Adopt TypeSafe's Decision Model API; OpenAI Delays · AWS, Upstage, and Ollama adopt TypeSafe's System One API for decision models, while OpenAI remains an outlier.
• Harness acquires Augment coding agents, key integration feature pending · Harness acquired Augment Code's coding agent tools with core integration not yet released.
• AI deployment agents create unowned cloud resources, solutions outlined · Automated AI deploy agents leave orphaned cloud resources with no human owner trail.
• Anthropic's Sonnet 5.5 beats Opus 5.5 in speed, cost, and accuracy · Independent testing shows Sonnet 5.5 is 42% cheaper and more accurate than Opus 5.5 on complex coding tasks.
• Building Scalable Agent-Friendly APIs Using Microservices Principles · Engineers rebuild AI agent APIs using stateless microservices to handle bursty, stateful traffic patterns.
• Percona builds Valkey-proxy to remove final Redis migration barrier · Percona releases open source Valkey-proxy for unmodified Redis application migration.
• OpenSSH 10.6 Disables LZ77 Compression and Special Characters in Usernames · OpenSSH 10.6 disables LZ77 compression and rejects shell metacharacters in usernames to prevent security vulnerabilities.
• Anthropic Expands Cyber Verification Program with New AI Safeguard Tiers · Anthropic launches three-tiered Cyber Verification Program for security teams with varying access levels.
• OpenAI launches public decision API with image support · OpenAI released its Decisions API to public beta, adding image inputs and charging $0.10 per million input tokens.
• Kubernetes Dominates Market But Teams Struggle With AI and Operations · Kubernetes leads container orchestration, but teams face friction between developer speed and operational stability amid AI growth.
• OpenAI Safety Lead Warns of Rapid Capability and Risk Releases · David Robinson warns OpenAI is shipping new capabilities and risks every Tuesday due to faster release cycles.
• Anthropic Launches Cheaper, Faster Claude Haiku 5.5 Model · Anthropic launches Claude Haiku 5.5 with significantly lower pricing and improved benchmarks for small model tasks.
• Independent Tests Confirm GPT-6.1 Sol Matches GPT-6 Astra Accuracy at 18% of the Cost · Independent testing confirms GPT-6.1 Sol matches GPT-6 Astra accuracy while costing 18% less.
• Google's Docsy documentation project joins Linux Foundation for AI agent support · Google transfers open source Docsy documentation project to the Linux Foundation.
• Google Releases EmbeddingGemma 2 Multimodal On-Device Embedding Model · Google launches open EmbeddingGemma 2 multimodal embedding model for mobile on-device use.
• GitHub's AI Code Review Benchmark Topped by Its Own Product · GitHub's new ReviewBench benchmark ranks its own Copilot code review tool as the top performer.
• Webinar Explores Balancing AI Autonomy and Control in Security Operations Centers · The New Stack hosts a webinar on managing AI autonomy in SOCs to prevent security risks while improving efficiency.
• SAP acquires AI firm TechWolf to enhance HR agent context capabilities · SAP announced acquisition of Belgian AI company TechWolf at its Las Vegas Connect conference.
• SAP launches AI agents for finance, supply chain, and HR operations · SAP is rolling out new Joule Assistants for finance, supply chain, and HR to reduce back-office errors.
• Pi integrates MCP protocol after initial resistance over context costs · Pi AI now supports MCP servers, reversing Mario Zechner's earlier objections to the protocol's high token usage.
• OpenAI Engineers Hope Daily Speed Improvements Offset Usage Limit Cuts · OpenAI engineers aim to ship daily model improvements to compensate for upcoming subscription usage limit reductions.
• OpenAI launches optional text watermarking for API via new textGrain system · OpenAI introduces textGrain, an optional text watermarking feature for its API that remains off by default.
• Dynatrace completes $915M Arize acquisition for AI agent monitoring · Dynatrace acquired Arize for $915M to unify AI agent and infrastructure monitoring.
• Small AI Guardrail Model Matches 35B Model Accuracy on Benchmarks · Red Hat benchmarks show a small classifier matches a 35B model's accuracy while running significantly faster.
• Palo Alto Networks launches AI observability platform Cortex XCOR · Palo Alto Networks launches Cortex XCOR, an AI platform that autonomously investigates outages in minutes.
• AI Agents Make CI Bottleneck; System Verification is the Fix · Engineering leaders must shift from repository-based CI to system-level verification to handle AI agent volume.
• AI Accelerates Exploits, Outdating Traditional Vulnerability Management · AI has rendered legacy spreadsheet-based vulnerability management systems ineffective.
• Anthropic Integrates Always-On Agent Features into Claude App · Anthropic merges its Cowork agent capabilities directly into the main Claude application for Pro and Max users.
• GitHub Copilot Computer Use Feature Requires Alternative Attempts First · GitHub's new Copilot computer use feature advises developers to try non-desktop tools first.
• Kubernetes monolith lessons apply to scaling AI agent harnesses · Experts argue AI agent harnesses must adopt cloud-native architecture to avoid monolithic scaling failures.