AI Agents in Practice
AI Agents in Practice
A Founder's Field Guide - 90 Days of Experiments & System Building
An interactive presentation covering my journey building with AI agents over the past 18 months. From early experiments with Cursor and Supabase to sophisticated multi-agent orchestration systems.
What's Covered
- The Journey - Timeline from Cursor + Supabase → Gadget → GTM automation → OpenClaw + Mac mini → Systems orchestration
- The Stack - Three-layer architecture: Agent Harness, Brain/Memory, Intelligence/Model
- Single Agent - One agent (Jarvis), multiple frontends (iMessage, Discord)
- Multi-Profile - Orchestrator pattern with specialist personas (Dev, Research, Writer, Analyst, GTM)
- Paperclip - Advanced multi-agent orchestration for parallel task execution
- Mac mini Setup - Headless, always-on AI infrastructure with stable/experimental user isolation
Key Concepts Explained
- Context Windows & System Prompts
- Agent Harnesses & Agentic Loops
- Skills/Tools & MCP (Model Context Protocol)
- Memory management (short vs long-term)
- Multi-agent orchestration patterns
Built With
- React + TypeScript - Type-safe component architecture
- Vite - Lightning-fast build tooling
- Framer Motion - Smooth slide transitions and animations
- Tailwind CSS - Custom "liquid glass" design system
- Canvas API - Animated neural network backgrounds
Hosted on Vercel, source on GitHub
