Private Zero-Knowledge AI Journal
Private Zero-Knowledge AI Journal
CortexOS is a zero-knowledge AI journaling application designed for iOS and Android. Its primary function is to provide a private, on-device writing experience where an AI language model, Llama 3.2, reads and reflects on journal entries directly on the user’s phone. No data ever leaves the device; there are no cloud calls, no third-party AI services, and no external servers involved. The app is built around the principle that users can prove their entries remain encrypted and inaccessible to anyone, including the developers, through open-source encryption code published under an MIT license on GitHub. The app offers several key features. Every entry receives a written reflection that summarizes the content in the user’s own voice, identifies cognitive patterns with suggested reframes, and links to past entries that share similar themes. Paid tiers unlock Deep Analysis, which runs asynchronously after each entry, producing a detailed reflection within seconds. The AI also tracks emotional states (over 20 emotions on premium tiers), behavioral patterns, and builds an encrypted personal summary that updates with each entry. Users can choose between standard and enhanced AI model sizes (approximately 1 GB or 2 GB) and switch between them at any time. Typical workflows involve writing daily or occasional journal entries, then receiving written AI reflections that highlight recurring themes, emotional trajectories, and connections to older entries. The app also provides nudges and prompts shaped by the user’s actual emotional history. The free tier allows up to ten encrypted entries per month, along with voice journaling and basic emotion detection. Premium tiers (with a 7-day free trial) include full AI capabilities, deeper emotional analysis, behavioral pattern detection, and an encrypted cloud vault for backup. A lifetime purchase option is also available. Technically, CortexOS operates entirely on the user’s device. The AI model runs in the app’s own process without any network calls, ensuring that all analysis happens locally. Encryption keys are derived from a recovery phrase, making it impossible for anyone else to read the entries. The app is available on iOS and Android, and its encryption code is publicly accessible for independent audit.