20 mins to Production: A Live AI-Build Challenge
summary
Live builds a habit-tracker app using Claude and Gemini agents, a precreated NextJS on Vercel repo, and browser SQLite for local storage; agents generate a design folder, numbered markdown spec, architecture, data model, checklist and a gap file. Prompts include a “grill me” design prompt and a concise “ponytail” coding prompt; separate agent sessions act as designer, builder and PR reviewer. Result: concrete design artifacts and an initial build started (not fully finished in-session) with repo, prompts and CI warnings preserved in an agent.md pattern.
AI-written summary · about
chapters (9)
- 0:00Talk overview
- 1:55Habit tracker project
- 3:39Design prompt for agents
- 7:32Grill me and ponytail prompts
- 10:45Building with phased instructions
- 14:03Four-phase development approach
- 18:20Starting build and resolving gaps
- 23:21Updating agent files
- 25:05Q&A on AI behavior
related talks
25mThe Ai-Assisted Developer Workflow: Build Faster and Smarter Today
31mPrompt Like a Pro & Vibe Code with AI Tools
57mDeveloping Production-Ready Apps in Collaboration with AI Agents
46mPractical End-to-End AI Development using Prompty and AI Studio
36mBuilding a Rails app live with AI
49mLet's build an AI agent
29mTools and Roadmap for AI-Native Developers
32mStop Chat-and-Hope and Start Building Better Apps with Agents