An AI chat platform for roleplaying, storytelling, and learning by chatting with millions of user-created AI characters.
Conway AI
FreeStarting price
0Popularity
About Conway AI
Conway AI is a locally served AI assistant that learns from user interactions through a guarded learning loop. It allows users to teach it facts, preferences, or words, which are immediately stored in private memory and referenced in subsequent replies. The system filters sensitive or unsafe content and only accepts short, useful examples for learning. Background updates occur when the host machine is idle, ensuring live chat remains unaffected. Conway AI operates on a single 16GB iMac using local MLX inference, avoiding reliance on external APIs. It supports both free and paid plans, with paid tiers offering higher token allowances for longer conversations. The platform also provides an API for programmatic access to its models, Conway-Retrain 12B and Conway-Omega 188M, with pay-as-you-go pricing.
Key features
- Private memory for user-specific facts and preferences
- Guarded learning loop with public transparency
- Background updates during idle server periods
- One-tap memory forgetting capability
- Open-source model checkpoint available
- OpenAI-compatible API endpoints
- Token-based usage tracking with weekly resets
- Public learning dashboard for tracking progress
Use cases
- Personal knowledge assistant for private facts and preferences
- Educational tool for incremental learning and memory reinforcement
- API integration for custom applications requiring adaptive AI
Pros
- Locally served with no external API dependencies
- Immediate private memory updates for taught facts or preferences
- Guarded learning loop with public visibility of teaching and rejections
- Free tier available with no credit card required
- Supports both chat interface and API access
Cons
- Limited to local inference on a single 16GB iMac
- Paid plans required for higher token allowances
- Learning loop may introduce latency during background updates
- No mobile or cloud-based deployment options
Frequently asked questions about Conway AI
Does Conway AI really learn from my chats?
Yes, and users can observe the learning process in real time. Teaching a preference, fact, or word is saved to private memory immediately, and the next reply that uses it shows which memory was referenced. A small number of taught words can become shared knowledge for all users after independent confirmations and rewording.
How smart is Conway AI compared to other AI assistants?
It is a compact, locally served model, so frontier systems may outperform it on breadth and trivia. Its strength lies in the learning loop: confirmed teachings work immediately from private memory, and verified updates improve the model based on actual usage patterns.
Is Conway AI environmentally friendly?
The service runs on a single 16GB iMac, avoiding the need for large GPU clusters. Efficiency measures include short-answer token ceilings, repetition guards, and background learning that defers when inference is active. The team avoids publishing unverified carbon claims.
Can I build applications using Conway AI's models?
Yes, Conway AI offers a platform API with pay-as-you-go access to Conway-Retrain 12B and Conway-Omega 188M. The API provides OpenAI-compatible endpoints, allowing integration via standard SDKs with usage-based pricing.
What happens to sensitive or unsafe content in teaching?
Sensitive, unsafe, badly formatted, or down-rated replies are filtered out and do not enter the learning queue. Only short, useful, and safe examples are reviewed for learning, ensuring the system remains secure and reliable.
How does Conway AI handle background updates without disrupting live chat?
Background updates occur only when the host machine is idle, ensuring live chat remains unaffected. Eligible examples wait in a guarded queue, and training starts only after sufficient clean material is available and serving is idle.
