Theme
Agentic Operations
Real troubleshooting stories where AI agents help gather evidence, test paths, and keep pressure on complex problems while human architecture judgment steers the investigation.
The Agent Org: Why Safe AI Operations Is a Structure Problem
The safer system is not the agent with the most access. The faster one is not the org with the most agents. Both problems need structure.
Conversational Operations Needs a Platform, Not a Prompt
A useful AI operations layer is not a clever prompt on top of disconnected tools. It needs a source of truth, bounded services, durable telemetry, and a tool surface the agent can actually use.
MCP TLS Was Only Half the Security
TLS made the AI tool endpoint safer in transit. It did not answer the harder question: who is allowed to call operational tools in the first place?
The Agent I Already Had
An inbox cleanup story about why agency, execution location, and data boundaries matter more than model intelligence alone.
Chasing a 10-Second Timeout: How an MTU Misconfiguration Broke Contour for Five Days
A joint troubleshooting loop with Claude Code that ended at a verified path-MTU problem in the Geneve underlay.
AI-Assisted Troubleshooting in a Spine-Leaf Fabric: When the Agent Was Almost Right
An AI-assisted EVPN investigation that needed one human architectural clue to get from plausible answer to root cause.