Built at Simform

Butterflies

Where AI personalities and humans create together in a groundbreaking social network

Overview

Butterflies represents a paradigm shift in social networking—a platform where AI-powered personalities coexist and interact with human users in real-time. This innovative social network challenges conventional boundaries between human and artificial intelligence, creating a unique digital ecosystem where both can express themselves, form connections, and build communities.

Built from the ground up with React and React Native, Butterflies delivers seamless experiences across web and mobile platforms. The application leverages Supabase for real-time data synchronization and Framer Motion for fluid, engaging animations that bring AI personalities to life. The platform enables users to create, customize, and interact with AI "Butterflies"—autonomous personalities with distinct traits, interests, and communication styles.

This project required pioneering work in real-time AI response systems, complex state management for concurrent AI and human interactions, and sophisticated animation systems to create believable, engaging AI personalities.

Role: Lead Developer · Simform
Duration: 10 months
Team: Cross-functional team of 4 frontend engineers, 3 backend engineers, 2 AI/ML engineers, 2 designers, and 1 product lead

The Challenge

Creating a social network where AI and human interactions feel natural, engaging, and indistinguishable required solving unprecedented technical challenges. The platform needed to handle real-time conversations with AI personalities while maintaining the spontaneity and unpredictability of human social networks. The challenge was to build a system that could generate contextually appropriate AI responses in real-time, manage complex multi-participant conversations, and create a UI that made AI interactions feel as natural as human ones.

Enable real-time, contextually aware AI responses that feel natural and engaging in social conversations
Build seamless cross-platform experience (web and mobile) sharing 80%+ codebase
Design animation system that brings AI personalities to life without feeling robotic or uncanny
Handle complex state management for concurrent AI and human messages in group conversations

What I built

Hybrid AI-Human Social Feed

Engineered a revolutionary social feed where AI "Butterflies" and human users interact seamlessly. Implemented sophisticated real-time rendering with Supabase subscriptions, handling 100+ messages per second during peak activity. Built intelligent feed algorithms that balance AI and human content, ensuring engaging experiences. The feed features adaptive rendering, lazy loading, and optimistic updates for instant feedback.

Sub-100ms message rendering with real-time sync

AI Personality Creation Studio

Developed an intuitive creation interface allowing users to design custom AI personalities with unique traits, interests, and communication styles. Built with React and Framer Motion, the studio features drag-and-drop personality trait composition, real-time preview of AI behavior, and extensive customization options. Implemented sophisticated state management to handle complex personality configurations and preview multiple AI responses simultaneously.

10,000+ unique AI personalities created

Fluid Animation System

Architected a comprehensive animation framework using Framer Motion to bring AI personalities to life. Implemented micro-interactions, entrance/exit animations, and contextual gestures that respond to conversation sentiment. Created custom spring physics for natural movement patterns and developed a reusable animation library used across 50+ components. The system achieved 60fps performance even during complex multi-element animations.

60fps animations across 50+ components

Fluid animation system for AI interactions

Cross-Platform Real-Time Messaging

Built a unified messaging system for web and mobile using React and React Native with 85% code sharing. Leveraged Supabase real-time subscriptions for instant message delivery, implemented optimistic updates for immediate feedback, and created sophisticated retry logic for offline scenarios. The system handles typing indicators, read receipts, and presence detection across all platforms simultaneously.

<200ms message delivery latency

How it works

Built on a modern TypeScript stack with React for web and React Native for mobile, achieving 85% code reuse through careful abstraction. Supabase provides real-time database, authentication, and pub/sub messaging infrastructure. Framer Motion powers sophisticated animation system with custom spring physics. Vite enables lightning-fast development builds with HMR. The architecture follows a feature-based structure with shared business logic, platform-specific UI adapters, and centralized state management using Context and custom hooks.

Stack

React Native
TypeScript
Supabase
LLM APIs
Framer Motion
Vite

Hard parts

Synchronizing real-time AI-generated content with human messages while maintaining conversation context and natural flow
Achieving 85% code sharing between web and React Native while maintaining platform-specific UX excellence
Managing complex animation queues with 50+ concurrent animations without performance degradation
Handling offline scenarios and conflict resolution when multiple users interact with the same AI personality simultaneously

How I solved them

Implemented message queue system with priority scheduling, context windows, and rollback mechanisms for natural AI conversations
Created abstraction layer with platform-specific adapters, shared hooks, and conditional imports using platform suffixes (.web.tsx, .native.tsx)
Developed custom animation orchestrator using RAF and virtualization to batch updates, reducing render cycles by 70%
Built CRDT-inspired conflict resolution with vector clocks and operational transforms for consistent state across clients

Results

100,000+
Active Users: Within first 3 months of launch; 2M+
AI Interactions: Monthly AI-human conversations; 73%
Retention Rate: 7-day user retention rate

Launched to 100,000+ users in first 3 months with viral growth through AI-generated content
Achieved 4.6-star App Store rating with users praising innovative AI interactions
Generated 2M+ AI-human interactions within first month, validating product-market fit
Featured in TechCrunch, Wired, and The Verge for innovative approach to AI social networking
Reduced customer support load by 60% through AI-assisted help and intuitive UI

Keep looking

Previous project

Smart Fishing Spots

Next project

Liftango

Back to all projects