The Real Talk:
- “Vibe coding” is giving way to “spec-driven development,” and that shift matters more for your security posture than your dev velocity
- If your company is staffing up with AI agents, who manages them, who do they report to, and who’s accountable when they go rogue? Nobody has a good answer yet
- Network automation adoption has been stuck at 30% for a decade. AI agents might finally remove the “you need to be a unicorn” barrier
A Closer Look:
Vibe Coding Grows Up John Capobianco, head of AI and DevRel at Itential, described his shift from vibe coding (letting AI write whatever, with little structure) to spec-driven development: a structured, Git-tracked process of writing a constitution, a spec, a plan, and tasks before any code gets generated. For security and audit teams, every decision gets documented and traceable, which is the kind of paper trail compliance work actually needs.
The IT Department Becomes an HR Department Capobianco referenced NVIDIA CEO Jensen Huang’s prediction that the future of IT isn’t managing infrastructure. It’s managing agents. Eric Brown, CISSP‘s team is already building specialized agentic developer roles (a project manager agent, a security agent, a researcher agent) for client work. That raises real questions about reporting lines and who’s responsible when an autonomous agent makes a bad call.
Networking’s 30% Problem For ten years, network engineers have been told to learn Python and automate everything, and adoption has stalled at roughly 30%. Capobianco’s argument: spec-driven development removes the requirement to be a “unicorn” who knows both networking and software development. Describe the requirement in plain language, and the configuration follows. That matters for any organization still running undocumented, decade-old network gear.
Bottom Line:
Vibe coding, spec kits, agent armies. The tools are moving faster than the policies meant to govern them. Organizations that get ahead of agent management, accountability, and documentation now will be in better shape than the ones still treating AI as a novelty, or worse, as a thinking colleague instead of a tool that needs the same oversight as everything else on the network.
Tune into the full episode to hear Samuel Santiago Aponte Cala, Nick Mellem and Joshua Schmidt add commentary about “Ralph loops” (named after that Ralph Wiggum), AFK coding overnight, the AI that minted its own cryptocurrency, and why John thinks the solo human billionaire company is closer than you’d expect.
🔗 Ep 91 – Next-Level AI: VibeOps, Agentic Employees and Rouge Bots
Listen wherever you get your podcasts – Subscribe to our YouTube channel to stay up to date on breaking cybersecurity news.
Learn more at www.itauditlabs.com
