Building AI Agents that actually work (Full Course)
Remy Gaskell teaches beginners how to build AI agents that autonomously complete tasks—moving beyond simple chatbots to systems that observe, think, and act in loops. The episode breaks down essential concepts like agent harnesses, MCPs (Model Context Protocol), skills, and memory systems, demonstrating how to create an AI executive assistant using local markdown files and tool integrations that can save 1-2 hours daily and scale to automate entire business departments.
Key takeaways
- • Chat models are question-to-answer interactions, while agents are goal-to-result systems that loop through observe-think-act cycles until tasks complete, making them fundamentally more productive for real work.
- • Building effective agents requires context engineering (loading business information via agents.md files) rather than prompt engineering, so agents understand your business before tackling tasks.
- • MCPs (Model Context Protocol) standardize how agents connect to external tools like Gmail, Google Calendar, and Stripe, eliminating the need for custom integrations.
- • Memory.md files allow agents to learn and retain user preferences across sessions, automatically improving performance when configured to update themselves based on feedback.
- • Skills are SOPs (standard operating procedures) packaged as markdown files—once you explain a process once, agents execute it identically every time, compounding productivity gains across repetitive tasks.
- • Scheduled tasks enable fully autonomous workflows; for example, running a daily briefing skill every morning at 9 AM that summarizes your calendar, inbox, and projects.
- • Beginners should start with user-friendly harnesses like Co-work or Claude Code before attempting more complex platforms like OpenClaw, building skills locally first before migrating to autonomous systems.
Recommendations (14)
"We're going to learn to drive and once you know how to drive, you can jump in any car, whether it's Claude Code, Codex, Open Claw, Manus, all of them."
Remy Gaskell · ▶ 1:29
"I'm going to open up Codex and anti-gravity and show you this loop actually happening in action."
Remy Gaskell · ▶ 6:46
"I've connected it up to Perplexity. So, it's now using Perplexity to research Greg. So, it's going through its first step of the loop."
Remy Gaskell · ▶ 10:58
"I've connected all of the tools that I use most like Gmail, Google Calendar, Granola, Notion."
Remy Gaskell · ▶ 34:11
Mentioned (7)
More from these creators
23 AI Trends keeping me up at night
Stop Vibe Coding. Start Getting Customers.
Paperclip: Hire AI Agents Like Employees (Live Demo)
Firecrawl AI clearly explained (and how to make $$)
I fixed OpenClaw so it actually works (full setup)
Karpathy's "autoresearch" broke the internet