← Videos
When AI uses Tools - Agents

AI from the Ground Up8 Nov 2026 · 4:30

When AI uses Tools - Agents

What changes when an AI assistant can choose a tool, inspect its result and decide what to do next?

From the episode description

What changes when an AI assistant can choose a tool, inspect its result and decide what to do next? We follow a small search-and-read task through the application around the model. The example shows how a proposed action becomes an actual tool call, how the returned evidence influences the next step, and where permissions and stopping conditions belong. In this episode: • What a tool definition tells the model. • Why proposing an action is different from executing it. • How an agent loop differs from a fixed workflow. • Why one tool call does not by itself establish an autonomous agent. • How limited permissions and clear stopping rules keep the task bounded. The fictional Workshop Notes walkthrough is an authored illustration, not a live agent run. Its tiny structured request explains the idea of tool calling without claiming to show a complete protocol message. Follow an assistant’s visible actions with four questions: what did it request, what actually ran, what came back, and why did it stop? The next lesson uses a working MCP teaching server to make the connection more concrete. ResearchNode: AI from the ground up • Episode 07 of 10. Previous: How AI Uses Your Documents: RAG Explained Next: MCP Explained: How AI Connects to Your Knowledge Presented by Michelle, an AI-generated presenter.

Descriptions and chapter markers are supplied by the channel. They are not a transcript or an independent check of every claim in the video.