The Mission: From "Hello" to "Welcome Back" 🧠

The biggest limitation of most AI apps is that they have "Goldfish Memory"—every time you open the app, it’s like meeting a stranger.

Tonight, we change that. We are building an agent that uses Gemini 3’s 2-million-token context window and a new Memory Skill to create a truly personalized experience. Your app will remember that you hate cilantro, prefer 15-minute meals, and that you scanned an onion three days ago.

What You’ll Build:

• The Context Engine: A system that stores user "Facts" (Allergies, Likes, Dislikes) and injects them into every AI prompt automatically.

• The History Vault: A UI that allows users to "scroll back in time" to see previous scans and recipes.

• The Proactive Chef: An agent that doesn't just wait for a scan; it suggests a meal plan based on the combination of your last three scans.

The Tech Stack:

• Gemini 3 Flash: For lightning-fast reasoning over long-term context.

• Google Antigravity: Our agent-first IDE platform.

• Flutter: The framework for building beautiful, multi-platform apps.

• The Memory--Architect Skill: Our custom "Brain Pack" for handling persistent data.

🎟️ "Wait, I missed Sessions 1 and 2!"

No problem! Like all sessions in this series, this is a standalone event. We provide "Jumpstart Scripts" that will catch you up to speed in under 60 seconds. You’ll get the "Vision" and "Vibe" capabilities from previous sessions pre-installed so you can focus entirely on AI Memory.

Agenda:

• 6:00 PM: The "Memory" Demo — Showing an app that "knows" its user.

• 6:15 PM: The Jumpstart — Installing the Memory--Architect Skill.

• 6:30 PM: Personalization Logic — Teaching the AI to store and retrieve "User Facts."

• 7:00 PM: The "Omniscient" Prompt — Combining Vision, Brand, and Memory.

• 7:20 PM: Show & Tell + Q&A.