CortexOS - Private AI Journal
The Story
I journal for my mental health. I write about anxiety, relationships, work stress, things I would never say out loud. And I kept asking myself: why am I handing this to a company that could get breached, subpoenaed, or just quietly sell my patterns to advertisers?
So I built something different. CortexOS runs a full language model directly on your phone. No cloud APIs. No data leaving your device. Your entries are encrypted with AES-256-GCM, and the keys are derived from a recovery phrase only you know. We literally cannot read your journal. Not because we promise not to, but because the math makes it impossible.
The AI detects over 20 emotions in every entry, tracks mood patterns across weeks and months, spots cognitive distortions like catastrophizing and black-and-white thinking, and gives you CBT-informed observations. It is like having a thought partner that remembers everything you have written but has zero ability to share it with anyone.
It also does things no journal app does. It groups your entries into 7-day chapters with AI-generated names. It builds a psychological profile that evolves with every entry. It generates a therapist-ready PDF report of your emotional patterns without including a single word from your actual journal. And it runs voice transcription on-device too, supporting 99 languages, with zero cloud processing.
I did not build CortexOS to compete with Notion or Day One. I built it because I needed a journal that could actually help me understand myself without making me trade my privacy for the privilege.
https://cortexos.app
Your Mind, Encrypted.
AI Overview
AI-generatedCortexOS encrypts every entry with AES-256-GCM and locks the key behind a six-word recovery phrase generated locally during installation. Because the phrase produces keys through Argon2id on-device, the developer cannot reproduce it even when served a warrant or data request. The same isolation philosophy powers an impressive technical feat: Meta’s Llama 3.2 language model—either 1 B or 3 B parameters—runs entirely within the app, slicing prompts and producing summaries without a single packet crossing the network. That means reflection, weekly chapter analysis, and 20-point emotion classification happen on silicon you control, not in some distant GPU farm. A Whisper-based speech engine follows the same rule: speak into the microphone and the transcription never leaves the handset, supporting 99 languages and demanding zero cloud credits.
The intelligence glows brighter because it feeds solely on your corpus. Each entry contributes to a living psychological profile; the AI highlights persistent cognitive distortions, charts mood against daily variables, and can export a clinician-ready PDF report stripped of any raw text. A still-optional encrypted cloud vault syncs ciphertext blobs that even the company’s servers treat as opaque, preserving offline-first peace of mind whenever you prefer it.
Crucially, the company never dangles paywalls over the core promise: the full on-device LLM and zero-knowledge encryption ship to every user on day one. Revenue talk is absent from the pitch, making the product feel less like a trial and more like a manifesto delivered in code. If journals are the R&D lab of the self, CortexOS finally treats that research as classified material—and keeps the clearance level where it belongs.
Key Features
AES-256-GCM Encryption
Every entry encrypted with AES-256-GCM, with keys generated locally using Argon2id and a six-word recovery phrase.
Local Language Model
Meta's Llama 3.2 (1B or 3B parameters) runs entirely within the app for summaries and analysis without network access.
Offline Speech Transcription
Voice input transcribed locally via Whisper-based engine supporting 99 languages.
Psychological Pattern Detection
Identifies cognitive distortions, charts mood against daily variables, and generates clinician-ready PDF reports.
Zero-Knowledge Cloud Sync
Optional encrypted vault syncs ciphertext that company servers cannot decrypt or read.
Use Cases
-
1
Mental Health Professionals
Therapy students and practitioners benefit from encrypted journaling without exposing sensitive patient data.
-
2
Legal Privacy
Individuals under legal or personal threat can journal knowing the developer cannot access entries even under warrant.
-
3
Confidential Reflection
Anyone seeking deep emotional journaling without uploading raw thoughts to external cloud servers.
-
4
Clinical Analysis
Healthcare providers can access patient-generated psychological reports in clinician-ready format.
FAQ
Is my journal data stored on company servers? ▾
Can you read my journal if I'm subpoenaed? ▾
Is the AI feature a premium add-on? ▾
What language is the AI available in? ▾
Tech Stack & Tags
Discussion
No comments yet — be the first!
Join the conversation — sign up to comment.
Sign up free