I Tested Gemini 3 as a Designer. It’s Terrifyingly Good.

Quick Overview

Gemini 3.0 in Google AI Studio successfully redesigned a personal website to be more modern and fresh, scoring an 8.5/10, but it struggled with designing a mobile app (Brainrot), indicating that while LLMs excel at web design iteration, complex mobile UI generation still requires significant human guidance.

Key Points: Gemini 3.0 successfully redesigned the speaker's personal website based on a screenshot and vague prompts, achieving a score of 8.5/10 for the desktop view. The AI struggled with designing a mobile app called "Brainrot" (a fitness app), creating a basic layout but failing to implement features like the mascot's animation or proper navigation. The AI correctly generated a detailed Product Requirements Document (PRD) for the SaaS dashboard idea, outlining problem alignment, goals, and key features. The speaker notes that while AI is great for iteration and handling simple design changes, it still struggles with complex, context-aware mobile UI generation. The speaker plans to explore Gemini's capabilities further, particularly with mobile app design, in future episodes. The speaker contrasted Gemini's performance with ChatGPT, noting a strong belief that Google AI products are currently 'crushing it' in design tasks.

Context: The video features the speaker testing the capabilities of Google's Gemini 3.0 model within the Google AI Studio environment, focusing specifically on its ability to act as a designer for web and mobile applications. The speaker sets up three main tasks: redesigning a personal website, designing a SaaS dashboard, and designing a mobile fitness app, using previous work and reference images as prompts to gauge the AI's performance against human expectations.

Detailed Analysis

The speaker tests Gemini 3.0 in Google AI Studio across three design tasks: redesigning a personal website, creating a SaaS dashboard design, and designing a mobile app. For the personal website redesign (using a screenshot and vague prompts), Gemini 3.0 produced a result the speaker rated an 8.5/10, noting its good design, modern feel, and correct placement of elements like buttons, though it struggled with the background color. The AI also successfully generated a detailed Product Requirements Document (PRD) for a separate 'Smart Inventory Tracker' idea, outlining goals, high-level approach, and key features. Finally, when prompted to design a mobile app called 'Brainrot' (a fitness app), Gemini 3.0 generated a visually clean main screen but failed to implement interactive elements like the mascot animation or custom navigation, scoring it an 8.5/10 overall for the mobile task. The speaker concludes that while Gemini 3.0 excels at iterative web design improvements and detailed documentation based on vague inputs, complex mobile UI generation still requires significant human input, noting that Google's AI seems to be outperforming competitors like ChatGPT in design tasks.