Code & Cloud Chronicle

Specright announced general availability of R&D Workbench for Food & Beverage, AI Assistant, and two new specialist agents.

• R&D Workbench lets...

• Harness variance is 7.8x larger than model variance on SWE-bench Verified
• Swapping harnesses shifted GLM-5.1 scores by 13 points vs 3-5 for model...

JetBrains CPO Kris Kang highlights how AI is reshaping software development beyond tools—driving faster prototyping, governance needs, and focus on...

SAP's new Industry AI portfolio combines 50 years of expertise across 26 industries, customer data ontologies, and forward-deployed engineering teams...

• AMD proposes -fenable-readonly-thp Clang flag to enable instruction Transparent Huge Pages at compile and link time
• Aligns segments to 2 MiB and...

Digital.ai and knowmad mood have expanded their global partnership to deliver AI-powered tools including TeamForge, Agility, Release, Deploy, and...

Nutanix introduced controls to run agentic AI workloads alongside VMs and containers on a single platform.

• Enterprise AI 2.8 now generally available...

• Qwen3.8-Flash-Next delivers hybrid QSA attention and n-gram embeddings to slash long-context latency while scaling parameters efficiently.
• AC2...

GitHub's August 17 outage lasted nearly eight hours after a misconfigured Istio sidecar policy blocked autoscaling during peak load, triggering...

82% of leading AI coding tools offer free tiers versus 61% for automation and agent products, signaling coding's more mature competitive landscape while agents emphasize execution capabilities over broad free access.

Vercel AI Gateway now triggers webhooks after long generations, slashing compute costs by dropping idle long-running processes and improving...

Traditional controls fail against autonomous AI agents operating at machine speed without human oversight.

• 68% of organizations cannot distinguish...

AI agents will become essential enterprise tools in 2026, handling complex workflows and team collaboration at scale. This marks their maturation as core infrastructure for automation and cross-team operations.

AutoSaddler turns manual harness design into an offline learning process that diagnoses failures from execution traces and applies targeted code...

Three new open-source and infrastructure projects show Claude Code maturing into a platform for building, searching, and visualizing AI agents.

-...

Hacktron now closes the loop between vulnerability discovery and patching by validating findings, eliminating false positives, and delivering ready-to-review fixes. Set rules once and it performs remediation like a real engineer on every trigger.

SensWear is an open-source modular platform for building smart rings, wristbands, patches, and custom devices with edge AI, synchronized raw sensor...

Agnost AI analyzes live user conversations with production agents to surface silent failures, behavior drift, hallucinations, user frustration, and...

AI coding evaluations are moving beyond simple test-pass rates to expose true migration readiness.

• Open-weights models now handle complex...

Perplexity AI introduced Portable Computer, an on-device AI agent for Nvidia DGX Spark desktops featuring Blackwell GPUs, 20-core CPUs, and 128GB...