Compose a PRView SourceOpen Issue

KubeCon NA 2026 — CFP Drafts

Event: KubeCon + CloudNativeCon North America 2026 CFP window: Typically opens June, closes August Watch: https://events.linuxfoundation.org/kubecon-cloudnativecon-north-america/

Also consider co-located events:

  • Multi-Cluster Management Day
  • AI + Cloud Native Day
  • Platform Engineering Day

Talk A — “The AI-Maintained Open Source Project: 90 PRs in Two Weeks”

Session type: Talk (35 min) Track: Contributor Experience / Open Source Sustainability

Abstract

What happens when you hand the majority of code authorship to AI agents?

KubeStellar is a CNCF-ecosystem multi-cluster Kubernetes project where the bulk of code is written, reviewed, and maintained by a system of AI agents — with humans setting direction and approving merges. In a two-week window in early 2026, over 90 pull requests were merged. Each was generated by a coding agent (Claude, Copilot, Gemini, or Codex), reviewed by an AI code-review agent, validated by automated tests and security scanners, and approved by a human maintainer.

This talk covers the architecture of the Hive multi-agent system: how agents are coordinated, how quality gates prevent regressions, what the bug-to-merge pipeline looks like end-to-end, and — critically — what humans actually do when AI handles volume. We’ll share concrete metrics on velocity, defect rates, and contributor experience, and address the hard questions: Can you trust AI-written code? What breaks? Where does human judgment remain irreplaceable?

Audience takeaways:

  • How to structure AI agent pipelines for open source projects
  • Quality gates that make AI-authored PRs trustworthy
  • What changes in project governance when AI is a first-class contributor
  • The new role of human maintainers in an AI-assisted project

Speakers: Core KubeStellar maintainer team

Talk B — “Multi-Cluster AI Infrastructure: Running LLM Workloads at the Edge with KubeStellar”

Session type: Talk (35 min) Track: AI + ML on Kubernetes / Edge Computing

Abstract

Running AI workloads in production means solving a distribution problem: models need to run near data, across jurisdictions, at scale, without single-cluster bottlenecks. KubeStellar’s multi-cluster orchestration engine — combined with its new A2A (Agent-to-Agent) and MCP (Model Context Protocol) server — provides the infrastructure layer that AI agent frameworks have been missing.

This talk walks through the architecture of running distributed AI workloads with KubeStellar, using the Cornell University Software-Defined Farm as a case study. We’ll cover: how KubeStellar’s placement engine routes AI workloads across clusters based on policy and data locality; how the A2A + MCP server lets AI agents coordinate across cluster boundaries; and what the operational reality looks like for a team managing inference workloads on edge hardware.

Audience takeaways:

  • Multi-cluster deployment patterns for AI/ML workloads
  • Using A2A + MCP for cross-cluster agent coordination
  • KubeStellar placement engine: policy-driven workload distribution
  • Case study: Cornell University SDF in production

Speakers: KubeStellar maintainer + Cornell University SDF team

Talk C — “Building a Production Multi-Cluster UI Without Writing Most of It”

Session type: Talk (25 min) Track: Cloud Native UX / Platform Engineering

Abstract

The KubeStellar Console is a production-grade multi-cluster management UI. It’s also a case study in what’s possible when you combine a clear plugin architecture, an AI-maintained codebase, and a community marketplace for extensions.

This talk shows what building and maintaining a complex Kubernetes UI looks like when the majority of code is AI-authored: how the plugin/card system enables rapid feature addition, how the community marketplace lets contributors extend the UI without touching core code, and how automated quality gates keep the AI-authored code trustworthy at scale.

Practical and demo-heavy. Attendees will leave with patterns applicable to their own platform tooling.

Audience takeaways:

  • Plugin architecture patterns for extensible Kubernetes UIs
  • AI-authored frontend code: what works, what doesn’t
  • Community marketplace model for UI extensions
  • Live demo: the console in action

Speakers: KubeStellar Console maintainers

Submission Checklist

  • CFP opens — monitor https://events.linuxfoundation.org/kubecon-cloudnativecon-north-america/
  • Identify confirmed speaker(s) for each talk
  • Finalize abstracts with speaker input
  • Submit all 3 proposals (increases odds of at least one acceptance)
  • Submit Talk A to Multi-Cluster Management Day CFP separately
  • Submit Talk B to AI + Cloud Native Day CFP separately

Related: kubestellar/kubestellar#3789