Disposable Software, Hyper-Personalization & the Future of AI Development

Disposable software means building a custom AI tool for one job, using it, and rebuilding it in minutes when better AI models come out. Armando J. Perez-Carreno and Mike Carlo share how they build and secure their own AI tools and agents. In this episode: Mike spent about an hour and a half helping a coworker try vibe coding with GitHub Copilot, and a week later she had built four tools on her own. They rebuilt their first app on a new tech stack by having the AI write down every feature of the old version and then rebuild the whole thing overnight. Treat an AI agent like a new employee, with its own GitHub organization, its own email account, a separate machine or virtual machine, and least-privileged permissions. Armando used Claude Code to build a scoping call app with his questions, checkboxes, and pricing, and at the end of the call it created the full statement of work. When your agent learns a new task, have it write itself a skill so it knows what to do the next time, since agents can lose track of what they've already learned. Guest: Mike Carlo, host of the Explicit Measures podcast Chapters 00:00 What disposable tools mean 01:56 Rebuild apps as AI models improve 04:19 Switching from Supabase to Azure 06:12 Treat your AI agent like an employee 10:23 Build a quick tool for one task 11:30 Hyper-personalized workflows for each person 13:26 Turn AI work into reusable skills 17:21 Why AI agents forget and how memory helps 28:41 Focus on problem solving over code 30:33 Managing agents versus managing people Episode page: https://webtalk.show/episodes/054-disposable-tools-build-it-use-it-throw-it-away/ Hosted by Armando J. Perez-Carreno. Learn more at https://perezcarreno.com