Build the new generation of AI Agents
summary
A working AI agent is assembled with Google AI Studio driving Gemini models, Vertex AI for deployment, and an Angular codebase used in demos; it ties tools like Google Search, Drive, Maps and a python sandbox into a task queue that creates and executes subtasks. Gemini Pro 1.5 claims a 1,000,000‑token context (up to 30k lines of code, one hour video, 11‑hour podcast, 800 pages) and the live agent demo finds conference facts via search but exhausts free credits before finishing.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:35Google IO 2024 announcements
- 8:24Google Cloud AI tools
- 12:40AI agent architecture
- 15:50Gemini chatbot features
- 21:56Gemini in AI/Vertex AI
- 23:00Gemini 1.5 context window
- 27:23Large context window benefits
- 31:24AI Studio playground
- 32:29Building autonomous agents
- 35:23Live demo: research agent
- 38:13Summary
related talks
01h 45mAI Engineering with the Google Gemini 2.5 Model Family
11mapproach to generative AI
18mIntroduction to Gemini APIs and AI Studio
37mBuild agents with Gemini API
15mBuilding with the Gemini API and AI Studio
01h 48mBuilding Conversational Agents
49mLet's build an AI agent
29mTools and Roadmap for AI-Native Developers