Shipping Vibe-Coded Apps with a Multi-Agent AI Workflow.

How different AI personas catch bugs, performance issues, and UX gaps before real users.

I presented this talk at theCursor Sydney meetup on August 12th, 2026.

I described how the LLM industry has progressed since prompt engineering, through spec-driven development, context engineering, and harness engineering, and has now arrived at loop engineering. I also covered how using multiple AI-agent personas to review specs and code can improve the quality of your AI-built applications.