Prompt. Build. Repeat. – The Vibe Coding Mindset
summary
Live demo shows a working Angular social‑post generator wired to Perplexity API for research, ChatGPT for generation, Lavable frontend and Supabase/Postgres backend; the demo Pokedex and CI fixes run end‑to‑end. Practical setup covers GitHub Copilot, custom instruction files and agents.mmd file, MCP servers (Context7, Angular MCP) and model choices. Claims and numbers: Copilot 55% faster, Cursor users report 5–10x productivity, Lovable hit $100M ARR in 8 months, but studies show 95% of AI pilots fail, LLM‑assisted developers were 19% slower in one benchmark and ~50% of generated code fails top‑10 OWASP tests.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:50Social media generator demo
- 3:36Talk overview and AI coding
- 5:44Wipe coding defined
- 7:53AI tool productivity claims
- 10:21100M paradox and adoption
- 11:26AI code quality impacts
- 13:54Why use AI tools
- 14:36LLM coding limitations
- 16:43Custom AI instructions setup
- 22:23MCP servers for context
- 27:46Right LLM for coding
- 32:46Co-star prompt engineering
- 34:31Spec-driven AI dev
- 35:13Quality gates for AI code
- 37:19AI coding best practices
- 38:23Autonomous coding future
- 39:04Call to action
related talks
12mNavigating the World of AI Development Without Going Insane
23mCode smarter, not harder: How AI Tools Boost Your Productivity
51mLive Panel Discussion: Less Code? More Vibes? Web Dev in the AI Era
31mPrompt Like a Pro & Vibe Code with AI Tools
48mMastering the AI Shift: Adaptive Strategies for Software Evolution
29mDo More using GitHub, AI, and VS Code
46mBreaking AI: Live coding and hacking applications with Generative AI
49mBuilding AI Enabled Web Apps