eXtreme Programming with Artificial Intelligence
summary
Claim: eXtreme Programming reduces AI risk by pairing XP practices with LLMs. Concrete examples: a Norwegian kiosk modernization ships a one‑button web pilot in under ten days, scales from 1→2→5→systemwide, and proves start minimal and evolve. Tools actually used: Claude Gemini ChatGPT IntelliJ RAG Kira Specmatic and CI/version control; a 24‑line human solution versus 1,800 lines from an LLM shows bloat. Problems: hallucinations and intermittent LLM outages; solution: never trust a single LLM and do the simplest thing that could possibly work.
AI-written summary · about
chapters (27)
- 0:00Introduction to AI and Extreme Programming
- 1:11What is XP
- 3:19XP as safety framework
- 5:06Schedule and quality risks
- 7:15Mantras for AI work
- 8:18Norwegian train tickets
- 13:53Start minimal evolve with AI
- 15:39Simplest thing possible
- 17:46Evolutionary design in practice
- 19:31Agile analysis learning
- 20:13Balance discovery delivery
- 21:57Human-AI balance refactoring
- 22:18Fear coding standards
- 23:43Direction delegation AI
- 25:09Human-AI collaboration
- 26:13Simple design expertise
- 27:58XP and AI faster results
- 28:41AI overdelivery bloat
- 29:49Balance human AI knowledge
- 30:12AI skepticism learning curves
- 31:15Fear in AI adoption
- 32:39AI limits hallucinations
- 33:22Trust multiple LLMs
- 34:25Evolutionary design production
- 35:08Beyond sprints vertical slices
- 36:11Avoid AI slot machine
- 36:56Q&A
related talks
53mImproving your life with Extreme Programming
46mIt’s XP Stupid!
13mDesigning AI-Intensive Applications
46mPanel: AI in the Loop - Left-Shifting Platform DX and Governance
28mXtremeJS Xperts Panel: The AI Revolution in Coding – Tools, Trends, and Tomorrow
30mUsing AI to Create Better Requirements and Designs
46mExtreme Programming with AI?
43mAI Development Lifecycle: Learnings of What Changed with LLMs