Professional description
This episode is a beginner-friendly crash course on AI agents. Remy Gasill explains the difference between chat models and autonomous agents, the agent loop (observe, think, act), and the four core components: LLM, loop, tools (MCP), and context. The video demonstrates building an executive assistant using local markdown files, connecting Gmail/Notion/Stripe via MCPs, creating memory and skills, and scheduling workflows to automate business tasks safely and iteratively.
– Agent fundamentals: Defines chat vs agent (question→answer vs goal→result) and illustrates the agent loop in real demos.
– Architecture & harnesses: Covers LLMs, agent loops, tool connections via MCP, and the role of agent harness platforms (Claude Code, Codeex, OpenClaw, etc.).
– Onboarding & memory: Shows agents.md (system context), memory.md (persistent preferences), security scoping, and local markdown as the canonical workspace.
– Skills & automation: Describes skills as SOPs for AI, how to create/package them, chain and schedule tasks, and examples like daily briefs and ad analysis.
Quotes:
Chat is question to answer; agent is goal to result.
Skills are SOPs for AI: explain once, never explain again.
Everyone’s going to have an AI operating system.
Statistics
| Upload date: | 2026-03-17 |
|---|---|
| Likes: | 6361 |
| Comments: | 262 |
| Statistics updated: | 2026-04-16 |
Specification: Building AI Agents that actually work (Full Course)
|
