Episode #28: The AI Code Audit: From Prototype to Production
Welcome to Episode 28 of CTRL-ALT-DEV! Today, Sean and Paul strip back the format for a fireside "Deep Dive" into the reality of AI-assisted software development. With tools like Replit, Lovable, and Bolt, founders are building working software faster than ever. But there is a massive hidden gap between a shiny prototype and a production-ready application. Because AI sounds incredibly confident and presents everything as correct, builders often don't realize the risks hiding beneath the surface. I n this episode, we unpack why every piece of "vibe-coded" software needs a comprehensive Code Audit before it goes anywhere near real users. Key Topics & Highlights: The Vibe Coder Dilemma: Vibe coders are often shrewd business owners trying to replace expensive SaaS subscriptions with their own AI-built tools. While they understand their market and price points perfectly, they often lack knowledge of the Software Development Life Cycle (SDLC). The Iceberg Analogy: Users see the visible 20% (beautiful dashboards and UI), but the invisible 80% below the surface holds all the risk. Paul explains the doomsday scenario: an unsupervised AI fixing a bug at 3 AM that inadvertently locks out 100 site workers the next morning. The 11 Pillars of Great Software: Inspired by a pre-AI developer conference, Paul details the foundational pillars of software that AI often ignores. These include security (zero-day vulnerabilities), performance bottlenecks, code maintainability, accessibility, automated testing, and proper database schemas. End-of-Life Compliance: When you roll an app out, it relies on hundreds of thousands of underlying packages and node modules. Paul explains how these frameworks often reach "end of life" for security updates within two years, requiring constant active maintenance. The £3,000/Month AI Trap: The hidden costs of auto-scaling AI apps are huge. Paul shares a story of a client spending £3,000 a month on Replit due to spaghetti code and auto-scaling. They dropped their costs to under £200 a month by exporting the code, moving to Railway for hosting, and using Claude Code. The True Value of AI Builders: Vibe coding is the ultimate prototyping tool. It allows founders to build an MVP, prove market demand, and even secure investment (like one user who successfully deployed a native app to TestFlight with 40-60 users). Once the prototype proves the concept, you hand the blueprint over to a professional dev team to scale safely. Links & Resources Mentioned in this Episode Replit, Lovable, & Bolt: https://replit.com/, https://lovable.dev/, https://bolt.new/ (The AI-assisted development platforms mentioned for rapid prototyping and vibe coding). Claude Code (Anthropic): https://claude.ai/ (The AI tool Paul recommends for maintaining the codebase more affordably once exported). Railway: https://railway.app/ (The hosting platform mentioned where a client moved their exported app to pay just £5 a month). Endoflife.date: https://endoflife.date/ (Paul specifically recommends this website to check when software frameworks and dependencies stop getting security updates). Submit Your Questions: mailto:[email protected] (The email address provided for listeners to submit their topics or code audit requests).

Zig 2026: No-AI Policy, $670K Foundation, Left GitHub & Why Zig Isn’t 1.0 - Andrew Kelley Explains

Andrej Karpathy: From Vibe Coding to Agentic Engineering w/ Stephanie Zhan

Building pi in a World of Slop — Mario Zechner

Episode #27: The AI Operating Layer, Lean Methodology, and Eliminating Business Waste

Episode #29: From Barrister to Vibe Coder: Disrupting a 200-Year-Old Industry with AI

Keynote: After the AI Hype – What’s Real, and What’s Next - Richard Campbell - 2026

Full Walkthrough: Workflow for AI Coding — Matt Pocock

Keynote: The dangers of probably-working software - Damian Brady - NDC London 2026

The Future of AI Agents with Andrew Ng | Interrupt 26

Billionaire's WARNING: I'm SELLING. The Crash Is Already Here!

Ryan Lopopolo - Harness Engineering: How to Build Software When Humans Steer and Agents Execute

Inside the Mind of Anthropic CEO Dario Amodei | The Circuit | Extended Interview

Slow down to speed up: AI and software engineering

Head of Claude Code: What happens after coding is solved | Boris Cherny

How AI agents & Claude skills work (Clearly Explained)

Anthropic Workshop: Build Agents That Run for Hours — Ash Prabaker & Andrew Wilson

Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI

Ed Zitron on CNBC: Generative AI Doesn't Work, And Big Tech Is Out Of Hypergrowth Ideas

NVIDIA CEO Jensen Huang's Vision for the Future

