Itential

The fastest way to find the limits of vibe coding? Put the code to work. In the latest VibeOps Lounge with John Capobianco, Andy Lapteff 🛠️💬 of The Art of Network Engineering gets candid about what happened after his AI-generated code shipped: Git collisions, unreliable automations, mismatched metrics, and a growing system nobody fully understood. The episode digs into what he learned fixing it including worktrees, multi-Claude workflows, and the bigger lesson for anyone building with AI: Getting code to run is only the beginning. Reliability, architecture, and understanding how the system behaves over time still matter. Watch the full episode below. 👇

A lot of buzz at the Georgia Tech Career Fair yesterday!⚡🐝 We had a great time meeting students, talking all things engineering, automation, and AI, and sharing a little about what it’s like to build at Itential. Our co-op program gives Georgia Tech students the chance to take an active role shipping products from day 1 working alongside our teams to contribute to real projects during the semester. If you missed us at the event, reach out to Melissa Chenggis to learn more about our Co-Op program. Thanks to everyone who stopped by the booth. See you next Spring! Jennifer Lu Sarah Toton Melodie Hsieh Sagar Narayana Glenn Butera Andrew VanNus Ankit Bhansali Christopher Calloway Georgia Institute of Technology

AI is changing the kinds of workloads Kubernetes is being asked to run, and that puts new pressure on the infrastructure, orchestration, and control planes underneath them. On Monday, John Capobianco joins BMC Software’s Anthony Anter and Techstrong Group’s Michael Vizard for a live discussion on what teams need to think about as AI infrastructure continues to scale. They’ll cover: • GPU orchestration and what gets harder at scale • Model deployment + inference patterns for production environments • AI-powered operations and self-optimizing clusters • The infrastructure and control requirements behind AI-native platforms For platform engineers, SREs, architects, and infrastructure teams, this is a practical look at how AI is changing the operating model around Kubernetes, not just the workloads running on top of it. 📅 Monday, September 21 | 1 PM ET Register at the link in the comments.

AI agents are getting smarter fast. But intelligence alone doesn’t earn them access to live network infrastructure. John Capobianco joined Swapnil Bhartiya on TFiR to unpack what it actually takes to bring agents into network operations without giving up control. They get into: • Why read-only tasks are a practical starting point for agentic automation • How FlowAI separates AI reasoning from deterministic execution • Why build-time tool selection is the key to controlled access + limiting blast radius • Real use cases spanning ticket triage, source-of-truth reconciliation, and port turn-up The conversation gets at the bigger challenge with agentic operations: giving AI enough access to be useful while keeping the controls, visibility, and auditability network teams need in production. Watch the full conversation. Link in the comments. 👇

AI workloads are putting new demands on the infrastructure platforms beneath them. On September 21, John Capobianco joins BMC Software’s Anthony Anter and Techstrong Group’s Michael Vizard to dig into what that means for Kubernetes, from GPU orchestration and model deployment to AI-powered operations and the control plane required to manage it all at scale. If you’re building the infrastructure AI will run on, check out John’s post below and save your spot today. 📅 Monday, September 21 | 1 PM ET 📍 Register at the link in the comments.

A major North American airline is bringing thousands of network devices back in-house while building the operating model to manage them with a lean team. With Itential, the airline is replacing manual, overnight change processes with governed automation and laying the foundation for AI agents to safely operate across production infrastructure. The expected impact: ✅ 100x more devices managed by the internal team ✅ 95% faster routine changes ✅ 80% fewer after-hours change windows ✅ 60% less engineering time spent on repetitive change work ✅ $3.2M+ in projected annual cost avoidance And every change, whether initiated by an engineer, workflow, or AI agent, runs through the same RBAC, approvals, and audit model. See how the airline is scaling operations in-house today while building toward an AI-ready future. 👇 Full case study in the comments.

A CVE alert tells you there’s a problem. It doesn’t tell you where you’re actually exposed. That’s the work that usually takes time: interpreting the advisory, checking it against inventory, and figuring out which devices are actually impacted. In this demo, John Capobianco and Joksan Flores show how FlowAgents can take a CVE, assess it against inventory in NetBox, Nautobot, SolarWinds, ServiceNow, or any tool, and return the device-level impact without manually selecting devices or working from spreadsheets. And the analysis stays traceable. The audit trail captures the agent’s reasoning, tool calls, inputs, and outputs so operators can see exactly how it reached its answer. That means faster vulnerability impact analysis without giving up the visibility teams need to trust the result. Watch the full demo. Link in the comments. 👇

Building an AI agent that works is just the tip of the iceberg. Below the surface is everything that determines whether it can actually operate safely in production: scale, failure handling, governance, approvals, auditability, and the amount of human-in-the-loop oversight. Holly Holcomb breaks down what that foundation looks like, including: ✅ Building governance + approvals into execution ✅ Adjusting human oversight based on risk + confidence ✅ Keeping every agent-driven change auditable ✅ Creating an operating model teams can trust to adopt See why operating agents safely depends on the tools, controls, and governance built around them, and what that actually looks like in practice. Check out the full blog below.

What does agentic change management look like when the network is highly regulated and every action has to be defensible? For a global managed service provider, it meant putting AI to work across BGP change operations while keeping validation, approvals, audit trails, and human oversight intact. The team built four specialized FlowAgents to handle: 1️⃣ BGP configuration validation 2️⃣ ServiceNow change request creation 3️⃣ Duplicate change detection 4️⃣Scheduled change execution In testing, a FlowAgent caught a configuration error and blocked execution even after the change had already been approved upstream. This is where agents start to change operations: not just assisting with a change, but carrying it through validation, coordination, and execution under governance. See how they built it. Link in the comments.👇

John Capobianco is getting in the flow with FlowAI. 🌊 He gave OpenAI’s new Image 2.5 tool a prompt and let it run with the idea. The result: a little FlowAI stop-motion fun for your Friday.