Patterns of Application Development Using AI
summary
Describes building Olympia.chat, a Ruby on Rails SaaS that embeds AI as discrete components (account manager, engagement bots) using system prompts and function tools; a production bug is fixed by adding a single prompt rule to prevent adding bots to cancelled accounts. Shows concrete patterns: structured IO (JSON/XML inputs and outputs), the ventriloquist prompt for guided retries, and an open source Ray framework for simple chat completions. Claims the app runs in production serving hundreds of custom customers daily and extracts patterns into a 500‑page book.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:08Why AI patterns
- 10:47Narrowing the path
- 14:40Multitude of workers
- 20:17AI components in code
- 23:09Account manager AI example
- 28:51Where AI adds value
- 30:14Structured input/output
- 32:19Ventriloquist approach
- 35:12Further learning
related talks
34mPatterns of Application Development Using AI
46mIntelligence as Infrastructure: What Developers Must Learn to Build with AI
46mDesign Patterns For AI Interfaces
19mThe New Application Layer
26mMocking and Deploying an AI Chatbot in Minutes
59mFrom Pair to Peer Programmer
01h 02mArchitecture Patterns for AI-Powered Applications
35mThe State of the Ruby AI Toolbox