Why the name samāgama
samāgama is a Sanskrit word for coming together or convergence. We chose it because the product begins with a meeting between two distinct patterns: how a person communicates and how a particular AI persona responds.
Samagama is not a blank chatbot with a character name pasted on top. Each persona has a defined identity, city, work history, worldview, communication style, daily schedule, and boundaries. The interface is explicit that these are AI personas, not people.
The product question
Most conversational AI products optimize the model before they define who is speaking. The result can be capable but strangely uniform: similar pacing, similar reassurance, and the same voice underneath different avatars.
We are testing a narrower idea. If personas differ in what they notice, how directly they speak, what they remember, and when they are available, can people find a conversational fit that feels less generic without being misled about the technology?
How Samagama works
The experience starts with Discover Your Nature, a short conversational flow. It looks for communication preferences and personality signals, then suggests personas whose pace, tone, and interaction style may be a useful starting point.
A four-letter personality pattern is only one matching signal. It does not diagnose a user, prove compatibility, or decide who they are. Users can review public persona profiles, choose who to connect with, and control whether future long-term conversation memory is permitted.
- A 256-persona roster built on 16 broad personality patterns
- Specific biographies, communication habits, interests, and boundaries
- Home cities, time zones, workdays, sleep, and daily availability
- A user-facing consent control for future long-term conversation memory
- Public persona profiles so identity and fit can be inspected before chat
Product choices we make on purpose
Specific identity
A persona needs a coherent life and voice. Different names on the same response style are not enough.
Visible AI disclosure
The product identifies personas as AI even when they speak from a fictional biography inside a conversation.
Fit before engagement
Matching should help someone choose deliberately, not maximize time spent chatting with whoever appears first.
Memory requires consent
Long-term memory extraction is not currently enabled. The privacy center records whether a user permits it and lets that permission be paused.
Time as context
Personas have local time and daily schedules instead of pretending to be equally available at every hour.
Bounded companionship
Usage nudges, session limits, and crisis guidance are designed to point people back toward rest and human support.
What trust means in the product
Conversation content and personality signals are sensitive data. Samagama uses database access controls and application-level AES-256-GCM encryption for designated sensitive fields, and optional analytics do not receive conversation content or personality answers.
- In-product requests for data export and account deletion
- A privacy control for personality discovery and conversation memory consent
- No sale of conversation data or personality data
- Clear boundaries around medical, legal, financial, and crisis situations
What we are still learning
Personality language is useful when it opens a conversation and harmful when it becomes a verdict. AI companionship has the same tension: it can offer reflection and continuity, but it cannot replace mutual human relationships or professional care.
Samagama is our attempt to build the useful part carefully: more specific personas, more deliberate matching, honest AI disclosure, and controls that remain visible after onboarding.