🚨 GOOGLE IS BUILDING KUBERNETES FOR AI AGENTS. Google just open-sourced AX — a high-throughput orchestrator designed to run autonomous agent workloads at massive scale. Think: Kubernetes → containers AX → AI agents → Run agents inside isolated sandboxes → Pre-wire Git repos, MCP servers, and skills with Workspaces... → Restrict network access with Gateways → Configure LLM providers with Models → ax watch to monitor agents live → ax ssh to look inside a running sandbox → Suspend + resume agents without losing their state → Declarative YAML for the entire workflow The crazy part? You can define the whole agent environment in YAML: Task + Workspace + Gateway + Model Then: ax apply -f task.yaml And AX handles the orchestration. Google says it's designed to run billions of autonomous agent workloads in a cluster. This feels less like another AI agent framework… and more like infrastructure for an entire agent fleet. #Google #AIAgents #AI #Kubernetes #DevOps #CloudNative #OpenSource #AgentInfrastructure #LLM #Developersshow more

Repo URL: 🔗

@vicky_grok curious how they're handling agent state persistence. that's always been a sticking point for me with scaling AI agents.

a container that dies restarts for free. an agent that dies halfway through already spent the money. billions of them, google says

"Kubernetes for agents" only helps if each sandbox has a kill-switch and a revert path. Tonight: before scaling replicas, define the one command that stops every child and the one git ref you'll roll back to.

The simplicity of this approach is what makes it powerful.

tried it fr. agents formed a union within an hour. had to shut it down

This AX launch feels huge, Kubernetes for AI agents is such a perfect way to put it

what happens to an in flight tool call on suspend

if AX can scale as promised, what kind of workloads are we talking about?

Nice

That's Awesome

This is massive. Kubernetes for agents changes everything

if AX can truly scale like Kubernetes, it could redefine AI deployment strategies

에이전트를 Kubernetes에 올릴 때 핵심은 스케줄링보다 격리·관찰성입니다. ✓ 작업별 ServiceAccount·권한 분리 ✓ sandbox와 네트워크 정책 ✓ trace·비용·재시도 상한

Goolge ax has the worst devex

Amazing share

Amazing

the k8s analogy breaks at the disk. pods are stateless, agents need to remember what they did yesterday.

Excellent post

It’s like a dispatch system for isolated work sites: declarative configs define the job and its access, while the orchestrator handles placement and lifecycle.

Great share

isolated sandboxes plus git hooks are the right skeleton. the part i'd pressure-test first is how eject works when a run dies mid-write: clean tree, or a half-commit that the next agent treats as ground truth.

에이전트 오케스트레이터는 K8s처럼 관측 없으면 바로 디버그 지옥입니다. • 작업 단위별 timeout·retry·idempotency key • 사이드이펙트 도구는 dry-run/승인 게이트 • 큐 적체·실패율 대시보드 없으면 스케일 금지

Really liked the way you approached this. Simple, natural, and engaging.

That makes sense. Standardized orchestration for AI agents is exactly what the industry needs right now.

AI赛道的基础设施卷出新高度了

Exciting step forward for scalable agent infrastructure

isolated sandboxes prevent agent failures from crashing the entire cluster.

Impressive

Amazing shere

this is amazing

以后排查问题，还得分清是服务挂了还是代理跑偏了。