Beyond Vibe Coding: The AI Software Factory

In this episode of ThriveCast, we speak with John Kennedy, co-founder of Actual AI and a veteran Dev Tools leader (AWS, Acquia, Upsun), who shares how engineering teams are moving beyond manual coding into building “software factories”; by leveraging AI-managed Architectural Decision Records (ADRs), John explains how teams can eliminate “vibe-coded slop,” enforce architectural consistency, and unlock a new model of linear acceleration with enterprise-grade governance.Key Insights• Vibe coding often leads to inconsistent and unscalable code (“slop”).• AI-generated code can degrade architecture, security, and maintainability over time.• Slop is not just bad code it’s anything that slows down development velocity.• Individual AI workflows are powerful, but lack consistency across teams.• Architectural Decision Records (ADRs) act as guardrails for AI-generated code.• Most teams still rely on passive documentation, which is outdated and reactive.• AI can now analyze entire codebases and auto-generate architectural rules.• Software factories enable 24/7 autonomous development with AI agents.• Organizations using software factories can achieve 10x–100x development velocity.• The future of engineering is shifting from writing code to improving the factory itself.Actionable Takeaways• Avoid relying solely on vibe coding for production systems.• Establish clear architectural guardrails (ADRs) before scaling AI development.• Use AI to analyze and codify existing codebase decisions.• Shift engineering focus from writing code to improving systems and processes.• Invest in building a custom software factory within your own cloud (VPC).• Optimize for time-to-production (speed of idea → deployment).• Reduce token usage and iteration cycles through better upfront decisions.• Encourage engineers to become “software factory engineers” instead of coders.Resources Mentioned• Actual AI — Platform that helps teams build AI-powered software factories with guardrails, ADRs, and autonomous agents.• G-Stack (Gary Tan) — Open-source framework for rapidly prototyping AI-powered applications and workflows.• Gastown / GAST — Open-source software factory framework for building agent-driven development pipelines.• AI Tinkerers — Community of builders and engineers exploring practical AI applications and workflows.If you’re a CTO, engineering leader, or developer building with AI, this episode highlights that modern software velocity doesn’t come from writing more code it comes from building systems, enforcing architectural guardrails, and evolving your team into software factory engineers who can continuously accelerate delivery at scale. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.hybridgtm.com