I Broke Kubernetes 3 Ways. Can an AI SRE Agent Find the Real Cause?
Source: YouTube · Kubesimplify · published Oct 1, 2026 · 24:06
BLUF: Nudgeb is an AI-driven platform designed to automate DevOps, SRE, and FinOps tasks by connecting agents through a unified environment that tracks cluster topology and changes, aiming to resolve complex incidents like cascading failures and invisible configuration drifts autonomously 2:26.
Key Takeaways:
• Nudgeb addresses the pain of 2 AM incidents by acting as a central platform where AI agents can investigate root causes without switching between multiple CLI tools like Grafana or kubectl 0:48.
• The architecture consists of a lightweight agent within the cluster that collects telemetry (metrics, logs, traces) via WebSocket and sends it to a cloud platform containing a services server, database, and AI agents 3:04.
• The platform features specialized agents including an AI SRE agent for alert triage and root cause analysis, an AI FinOps agent for cost optimization, and a K8s Ops agent for cluster management 1:41.
• Nudgeb’s UI provides a "Briefing" feature that automatically ranks incoming events by severity (P0-P3) and offers a chat interface for users to interact with the AI teammate during investigations 6:49.
• The video demonstrates testing on a self-hosted bare-metal Kubernetes cluster using GPT-5.6 to validate how the AI SRE agent handles specific scenarios like cascade failures and pinned HPAs 4:35.
Closing statement: By centralizing data and leveraging interconnected AI agents, Nudgeb seeks to reduce manual debugging time and improve operational reliability in Kubernetes environments.
Sources:
- 0:48 Explanation of the manual debugging process Nudgeb aims to replace.
- 1:41 Overview of the different AI agents (SRE, FinOps, K8s Ops) available in the platform.
- 2:26 Description of the platform's core value proposition regarding interconnected agents and environment awareness.
- 3:04 Technical breakdown of Nudgeb's three-part architecture: agent, platform, and AI/workflow servers.
- 6:49 Demonstration of the UI's triage inbox and briefing capabilities.
- 4:35 Details on the test environment setup and the specific LLM used for the demonstration.
Generate CPE Credits
Generate a professional CPE document from this video's transcript.
Estimated credit: 0.5 CPE hours
Estimate uses the video runtime (1 hour ≈ 1 CPE, rounded to the nearest 0.5, minimum 0.5, maximum 2.0). The final amount can be lower after review, never higher.
Topic: Cybersecurity. Commonly maps to: Security and Risk Management, Security Operations. Exact CISSP domains are assigned during generation.
Free account. One generation at a time, with a daily limit.
CPEBuddy is independent and not affiliated with or endorsed by ISC2, ISACA, or any certification body. Exports are formatted for common CPE submissions; acceptance is at your certification body's discretion.
Transcript Preview
First 800 characters of the transcript
It's 2:00 a.m. in the morning and as usual the phone rings, payments is crashing, orders are not ready and the end users are getting errors. And when you open the logs, everything looks perfect. Pods are running, the card cache is running, zero restarts, zero warnings. That's the kind of incident we'll kind of go through today. And uh there are three incidents that we'll try to solve where the incident points to something else but the real cause is really something else and then we'll see if a tool like nudgeb can actually solve it. So what is nudgeb? Think about what you would do today when an alert comes. You open graphana, you check the logs, you do cubectl describe and you try to remember any of the change that you made last week. Uh that's like four to five different steps, four to fi…