Kin is a personal AI assistant for social and personal growth, built around personal privacy and knowledge. In public beta since fall 2023, and spanning across a Chat and Journal function, its 3 most distinct features are:
For a more in-depth guide to these features, Kin’s other features, and how to use them, head over to our article explaining How to start with Kin.
Kin’s features allow it to be a powerful, private support system - and not just because, like other AI tools, it’s always available.
With a long-term memory it can draw on at any time, Kin can easily recall and analyze past feelings, situations, and the patterns within them.
This makes it uniquely positioned to help you learn about yourself and make personal decisions, because it can see and advise on patterns in your life that may be difficult for you to recognise.
Kin currently runs on the latest Llama 3 70B open-source LLM, enhanced with custom prompting and reasoning improvements through its long-term memory system. Kin can also reach out to the internet to provide additional information and insights, if you allow it to in the settings.
While Llama isn't the most advanced AI available, it's among the best accessible to regular users, and is ideal for our on-device, privacy-focused approach.
Through two-way Chat discussions and just-you Journal entries, Kin builds a private memory of facts you share. Kin can access this memory during any conversation, meaning you never need to repeat yourself, and Kin knows to ask how your friend is doing, or to remind you to bring water to an interview (because you forgot it that one time).
Kin takes a local-first approach, meaning most data processing and storage happens purely on your device. This uses minimal storage space - but also means you must make your own back-ups: we as developers cannot access or read your interactions, and have no rights to your content.
You can further enhance privacy by managing data processing settings in the app.
Yes, you can customize Kin's "Tone of Voice" in the "Your Kin" section. This will change its tone and attitude in responses. While there's no strict limit, keeping the prompt under a few thousand characters is the most efficient.