AI Briefing#kubernetes
The latest AI and developer news about #kubernetes, with the original source and a short summary.
The latest AI and developer news about #kubernetes, with the original source and a short summary.
Released as v0.4.2, Agent Sandbox reduces operational complexity compared to StatefulSets and supports kernel isolation.
NVIDIA has open-sourced the full Run:ai platform, including KAI Scheduler and Model Streamer, under the NVIDIA DSX AI Factory GitHub organization.
Performance improved on the same resources by reducing the remote DRAM access ratio from 50% to 0.8%.
Reduces operational complexity compared to StatefulSet and resolves cold starts using gVisor isolation and WarmPool.
A multi-tenant all-in-one solution installable on both on-premises and cloud environments, supporting dynamic allocation and automatic reclamation.
Microsoft Research Asia released Agent Lightning v1.0, a 3,500-line framework that raised Qwen3.5-9B's Pass@1 on SWE-bench Verified by 14.6 percentage points using only 6,000 training samples.
AICR v1.0 establishes a stable compatibility contract across CLI, REST API, and Go SDK while supporting 10 GPU accelerators and 11 Kubernetes services.
The Jev 1.13.0 model achieved a 76.2% pass rate on SREGym-Lite benchmarks, performing close to GPT-5.6 Sol's performance while being approximately 7 times faster and 200 times cheaper.
AI21 eliminated 20 weekly manual interventions and reduced hero job starvation from 72 to 12 hours by implementing Kueue on a 10,000-GPU cluster.
A security benchmark found that default Ray and vLLM configurations on Kubernetes expose critical control-plane ports to cross-namespace attacks, which standard scanners fail to detect.
Global model accuracy improved by 26 percentage points over local models in asymmetric data environments
MS unveiled Physical AI Toolchain features that offload robotics inference to the edge or cloud, extending battery life by up to 160% and improving performance.
KT Cloud has proposed a solution to reduce deployment failure root cause analysis time by 90% by applying the MCP ecosystem.
Google has released AX, an open-source orchestrator for declaratively managing agent workloads.
AWS Elastic Beanstalk has launched a fully managed Cluster Mode based on EKS to support infrastructure sharing and simplified operations for multiple applications.
Toss Securities shared its experience of overcoming the limitations of GPU-awareness by building a GPU-native cluster equipped with validation and resource control capabilities.
Google has released 'Agent Substrate', an open-source runtime that enables AI agents to run at 10x higher density on GKE and restart in under 500ms.
Amazon EKS has launched a feature allowing customers to directly configure five control plane parameters, including those for kube-scheduler.
Envoy AI Gateway has been renamed to 'Agent Router' as an independent AAIF project, while maintaining compatibility with existing CRDs and CLI.
By leveraging Run:ai Streamer and cache optimizations for serving Gemma 4 with vLLM on Amazon EKS, cold start times were reduced from 428 seconds to 226 seconds.
InfraLab implemented Envoy AI Gateway, leveraging model abstraction and tiered routing to reduce LLM costs by 78%.
Migrating a Kafka Streams application from AWS EC2 to Kubernetes and applying KEDA and Consumer Lag-based scaling reduced infrastructure costs by over 80%.
A photographer at Woowa Brothers used AI to develop a game supporting thousands of concurrent users, focusing on failure scenario validation and architecture design.
The periodic memory spikes occurring after the EKS migration were confirmed to be caused by log compression and page cache.