A social AI that learns, stores, and intelligently utilizes personal context in its interactions with you.
Persona is a social AI platform that learns, stores, and intelligently utilizes personal context in its interactions with you. It offers immediate value through practical social assistance like analyzing a crush's text, deciding which Instagram post to upload, creating Tinder responses, and generating gift ideas - all while building a deeper understanding of your personality, social graph, interests, and communication style for when you want to dig deeper.When you dig deeper, you'll use Persona since it already contains your personal context.
General LLMs are limited in memory and social adaptability, they yap a lot, don't adjust for your communication style, and don't learn much about you. Competitor personalized AIs (Replika, Dot, Pi, etc.)are either AI therapists (pathologizing users and requiring a lot of uer burden before they become useful) or sci-fi virtual friends (creating uncanny valley relationships which alienates general audience). Persona avoids these traps by providing shot-term hooks that are useful to a general audience, while naturally building personal context for when you do dig deeper.
Analyze crush's texts, decide between Instagram posts, craft Tinder lines, plan a date).
Turn-by-turn interactions without the AI yapping, and adaptive communication style.
Builds user context naturally through lightweight interactions rather than lengthy onboarding.
Builds a social graph of people in the user's life with relevant details, these are then put in context when they're mentioned.
Neither an AI therapist (which pathologizes users) nor a sci-fi virtual friend (which creates uncanny valley relationships).
Purpose: Passively save and update data of anecdotes, personality, and communication style.
How It Works:
Anecdote Collection
Records stories and experiences shared during conversations.
Personality Assessment
Analyzes communication patterns to determine traits and preferences.
User Response Analysis
Engagement Tracking
Bayesian Trait Modeling
Updating confidence scores (e.g., directness = 4/5, confidence = 4/5) through interactions.
Purpose: Put relevant data into context from the saved database.
How It Works:
Social Graph Queries
When a name is mentioned, pulls relevant node data (relationship, stories, personality, areas of conflict).
Vector Embeddings
Uses vector embeddings to search through database for relevant anecdotes and context.
Context-Specific Feature Adjustment
Adjusts features based on conversation type (e.g., for attachment issues: directness=1/5, emotional support=4/5).
Purpose: Draft an initial message while keeping personal context in mind.
How It Works:
Conversation Style Integration
Uses selected conversation style button (questioning, crush text analysis, Instagram decisions, etc.) as initial system prompt.
Length Control
Controls maximum response length based on message characterization thresholds.
Turn-by-Turn Design
Instructed to send questions before continuing generation or starting new conversation branches.
Purpose: Adjust the answer to fit the conversation and user better while keeping the main idea.
How It Works:
Style Adaptation
User-Style Matching
Adapts to user's communication style and preferences based on past interactions.
Context-Specific Adjustments
Modifies tone and approach based on conversation purpose and user's emotional state.
Content Preservation
Ensures the main ideas and value of the response remain intact while adjusting presentation.
Companies can access hyper-personalized insights locally through Persona, without ingesting, storing, or cross-linking user data on their end.
Examples:
Contextually relevant product and service suggestions.
Examples:
People are often reluctant to give their data when they think it's needlesly exploited.
but!