← BACK TO SKILLS
FREE

Using Gemini 3.0 Pro as a 10x AI Designer for Websites, SaaS & Mobile Apps

by @gregeisenberg

Business Business★★★★☆ principles

ABOUT THIS SKILL

Greg Eisenberg stress-tests Google AI Studio’s Gemini 3.0 Pro model by designing a personal website, a restaurant-analytics SaaS dashboard, and a fitness mobile app in real time, scoring each on aesthetics, usability, and feedback responsiveness.

TECHNIQUES

vibe designingreference image guided promptingvisual annotation feedbackproduct requirements doc promptingmulti turn design iteration

KEY PRINCIPLES (12)

Reference-Driven Design

High-quality reference images are the fastest way to escape generic Tailwind-purple vibe-coded UIs.

Greg pastes a Dribbble SaaS dashboard and a Teenage Engineering hardware photo to steer color, layout, and tactile button styling.

Why: Vision-language models map pixels to design tokens more faithfully than text alone, reducing hallucinated styling.

"the better reference images you get, you generally get more out of it"

Feedback Loop

Visual annotation inside the live preview compresses iteration cycles from hours to seconds.

Using the built-in annotation tool, Greg circles the white background and writes, "this white background is boring… please adjust," and the model regenerates the CSS within two minutes.

Why: Spatially-grounded feedback eliminates ambiguity that text descriptions introduce.

"draw and write on your app preview to instantly communicate your vision to Gemini"

Taste as Moat

In the era of the idea guy, taste—not coding skill—determines design quality.

Greg argues that anyone can now produce beautiful apps if they supply the aesthetic judgment and reference material.

Why: Generative models commoditize execution, leaving curation and taste as the scarce resource.

"you have to have the taste, you have to have the reference images"

Context Injection

Feeding live URLs or screenshots of your existing site lets the model inherit real content and structure.

Greg pastes a screenshot of gregisenberg.com so Gemini can rewrite copy and reorganize sections while preserving original assets.

Why: Grounding generation in actual content prevents hallucinated lorem-ipsum and keeps brand voice intact.

"the website domain is gregisenberg.com if you want to grab content from there"

Nostalgia as UX Hook

Retro visual metaphors (e.g., Windows XP desktop) create scroll-stopping differentiation for personal sites.

The redesigned site uses XP-style windows, a start bar, and skeuomorphic icons to stand out from generic portfolios.

Why: Familiar but unexpected interfaces trigger emotional engagement and social sharing.

"a lot of people have personal websites… but they don’t have something like this"

Physicality in Digital UI

Borrowing tactile cues from industrial design (knobs, bevels, shadows) makes SaaS buttons feel ‘real’.

Greg references Teenage Engineering’s hardware to push the model toward 3-D, pressable controls instead of flat rectangles.

Why: Tactile affordances increase perceived reliability and reduce cognitive load for non-technical users.

"I want the buttons to feel almost like real buttons in the real world"

Gamification Blueprint

Copying proven gamification patterns (mascot, streaks, leaderboard) accelerates mobile-app engagement.

Using Brain Rot’s App Store screenshots as a template, Gemini produces a fitness app with streak tracking and a reactive mascot named Gains.

Why: Proven behavioral loops reduce risk when entering crowded app categories.

"everything about it feels to me like a blueprint for getting someone to do something"

Prompt Minimalism

Sparse, high-level prompts often yield more creative interpretations than exhaustive specs.

For the personal site, Greg only asks for "more color, more freshness… Microsoft XP experience" and lets the model fill gaps.

Why: Over-constraining can stifle emergent design solutions the model might discover.

"not giving it a lot of instructions… I kind of want to see what it does"

WHAT'S INSIDE

PRINCIPLES
5
TECHNIQUES
12
EXPERT QUOTES

This is a structured knowledge base — not a prompt file. Your AI retrieves principles semantically, understands the reasoning behind each technique, and connects to related skills via a knowledge graph.

Compatible with OpenClaw · Claude · ChatGPT

principles · semantic retrieval · knowledge graph

Free during beta · Sign in to save to dashboard