← Back to all playbooks

Logan

LOGAN KILPATRICK & AMMAAR RESHI: Google AI Studios
Start: 04:31:41
Sketch note for Logan's session
TL;DR

Google AI Studio is evolving from a model playground into a full-stack 'vibe coding' environment that prioritizes high-fidelity UI prototyping from screenshots and one-click deployment over traditional IDE setups.

Key Insights

Figma-to-Functional-UI via JSON Style Injection

04:39:17

To achieve design precision, use a Figma plugin to export JSON code for styles and typography, then pass this directly to the model alongside a screenshot to ensure the generated app matches brand specs exactly.

The 'Loom' as the New Code Diff

04:44:28

For vibe-coded projects where the builder may not understand the underlying code, a video walkthrough (Loom) of the feature is a more effective unit of proof and quality validation than a traditional code review.

Automated UX Discovery via Walkthrough Prompting

04:44:07

When building complex UIs with AI, prompt the model to 'create a walkthrough of the feature' to force it to click through and demonstrate its own interactive logic, revealing hidden UX flaws.

Tool Calling as the Engine of Ambition

04:52:08

The transition from 'slop' to ambitious apps is enabled by tool calling, which allows the model to perform 'on-the-fly context engineering'—making micro-decisions like searching docs or editing specific files autonomously.

Systems

High-Fidelity Vibe Prototyping

  1. Capture a screenshot of a Figma UI component.
  2. Use a Figma-to-JSON plugin to extract exact CSS/style parameters.
  3. Paste both into AI Studio's build mode.
  4. Prompt for an interactive simulation (e.g., 'Make this sidebar functional').
  5. Iterate in the preview panel until the 'vibe' matches the design.
Tools: Figma, Google AI Studio, Figma JSON Plugin

Actionable Data Visualization

  1. Upload a raw CSV file directly to the LLM without pre-cleaning.
  2. Prompt for an 'interactive visualization' to explore the data.
  3. Layer on generative tools (e.g., 'Add a button to generate a YouTube script based on the top-performing row').
  4. Deploy as a functional internal tool.
Tools: Google AI Studio, Kaggle (for datasets)

Tools

Google AI StudioGemini APIAntigravity (Coding Agent)FigmaLoomKaggle