← Back to Dashboard
1. No-Code Agent Prototypes (The Prototype)
Before writing Python code, we start with a no-code sandbox.
- Default: Google AI Studio (Gemini) for prompt iteration, saved versions, and exporting examples.
- Optional: No-code "agent builder" experiences (e.g., ChatGPT GPT Builder / Gemini "Gems") to package instructions + files into a shareable assistant.
- Pedagogical Goal: Learn structured instructions and "system prompting" without syntax errors.
2. Google Opal (The Visual Logic Builder)
Moving beyond text prompts into Visual Workflows.
- What it is: A node-based tool from Google Labs to build "Mini-Apps."
- Why it matters: It teaches the "Logic Flow"—how to connect a prompt to a search, then to a filter, then to an output—using a visual map instead of code.
3. Actions (The "Hands")
A smart brain is useless if it can't do anything.
- What it is: Using OpenAPI Specifications (JSON/YAML) to teach a GPT how to use an external software tool.
- Example: Teaching a GPT to "Check the Weather" or "Search the Library Catalog."
- The Skill: Reading API Documentation and writing a "Manifest."
Access Policy: Workshop workflows avoiding student billing. We teach API fundamentals via controlled demos or local mock endpoints.
4. MCP (Model Context Protocol)
This is the future of Local AI.
- The Problem: Cloud AIs can't see the local file on your desktop named
secret_thesis_draft.docx.
- The Solution: MCP is a standard way to build a "Server" on your computer that safely exposes your files to an AI.
- The Metaphor: It's a "USB Port" for AI. You plug your data into the model using a standard connector.