531 posts tagged with "AI agents"
AI agents and autonomous systems
Kubernetes 1.36 Ships Gang Scheduling In-Tree: What It Actually Buys an AI-Agent Batch Fan-Out
Kubernetes 1.36 ships PodGroup and Workload APIs in-tree, letting the scheduler gang-schedule a multi-pod fan-out atomically. Here's exactly what the five alpha feature gates do, what they replace Volcano/Kueue for, and how far alpha still is from safe for tenant workloads.
LiteLLM's Agent Platform Fixes Session Amnesia With One Postgres Table, Not a New Kubernetes Primitive
BerriAI's alpha-stage LiteLLM Agent Platform keeps an AI agent's session alive through a pod restart by storing state in Postgres instead of the sandbox itself — here's the actual mechanism, how it stacks up against E2B, Daytona, and Modal, and what it means for any MCP server with real deploy authority.
Volcano's Headlamp Plugin Kills the vcctl Guesswork: A Practical Walkthrough for Gang-Scheduling AI-Agent Batch Jobs on a Cluster API Fleet
A production case study put GPU idle time at 38 percent from partial gang-scheduling failures alone. Here's how Volcano's gang scheduling and queue fairness fix that, what the new Headlamp plugin actually shows instead of vcctl output, and whether it's worth running over hand-rolled priority classes on a Cluster API fleet.
WebMCP Lets a Web Page Register Its Own AI Tools. Does a Deploy Dashboard Need One?
Chrome's new WebMCP API lets a web page register its own AI-callable tools, but a deploy dashboard already has a REST/GraphQL API and an MCP server. Here's the actual capability delta, and why it's not worth building yet.
A2A Turns One and Hits v1.0: What a Standard Agent-to-Agent Handshake Actually Buys a Deploy-From-Chat Pipeline
A2A hit v1.0 and 150+ organizations in its first year under the Linux Foundation. Here's a worked example of what its signed Agent Cards and task-approval states actually change for a deploy pipeline where one agent requests a rollback and another approves it — and whether a self-hosted PaaS's MCP server needs to speak it yet.
Giant Swarm's AI Agent Platform Proves Cluster API Scales — But Not Every Layer It Runs
Giant Swarm's production numbers — 2.8x lower cost per agent run, 500 parallel agents — validate Cluster API and GitOps as an AI-agent substrate. But its Backstage-centric IDP layer is a $300K-$700K/year commitment a simpler git-push control-plane API doesn't need.
GuardFall: Why 10 of 11 Open-Source AI Coding Agents Can't Tell What Bash Will Actually Run
Adversa AI's GuardFall research found 10 of 11 open-source AI coding agents check a shell command's raw text for danger, then hand that text to bash, which rewrites it before running. Here are the five bypass classes, per-agent results, and what it means for building agent-callable deploy tools.
Kelos Turns Autonomous Coding Agents Into Kubernetes CRDs
Kelos turns an AI coding agent's entire working context into four Kubernetes CRDs you can kubectl get and git-revert. Here's what each primitive actually does, and the concrete case for treating it as a reference architecture rather than a dependency.
Kubernetes 1.36's Declarative Validation Went GA — But Not for the CRDs Your Agent Actually Writes To
Kubernetes 1.36 graduated Declarative Validation to GA — for built-in types only. The CEL rules that actually govern your CRDs have been GA since 1.29. Here's the real timeline, a worked example from Cluster API's own CRDs, and what it means for an agent generating manifests.