Account-scoped, RLS-isolated.
Auth + database run on Supabase with row-level security policies that physically prevent one user's queries from returning another user's data. Not a promise — a constraint enforced by the database itself.
Personal Context IntelligenceYour knowledge is yours.
Your knowledge is yours.
We've made that mathematically true.
Personal Context Intelligence requires a different privacy posture than generic AI. Architecture, not promises. Here's what data exists, where it lives, which third parties touch it, what we do with it, and what Big Tech does that we won't.
Journal Genie is in early access. The core product is materially landed; broad public-launch hardening continues. We say so on every page so you can decide for yourself.
Account-backed sync with browser cache.
Your work syncs to your account for durability and hydrates into your browser for speed. Clearing local storage clears the local convenience layer, not your work.
Export anytime. Plain markdown. One click.
The same export the day you sign up, the day you cancel, the day you delete. We earn trust by making leaving easy.
The list is short because we mean every line.
- We do not train AI on your journals, sources, or chats. We have not opted in to model-provider training with our partners (OpenAI).
- We do not sell your data. Not to advertisers. Not to data brokers. Not to "research partners." Ever.
- We do not run ads. The product is funded by you paying us, not by your attention being sold.
- We do not ad-target you. We do not build behavioral profiles for marketing.
- We do not retain raw individual data for any "Insights API" — anything aggregate is thresholded; no individual rows ever.
The contrarian frame, with receipts.
Each row pairs a Big Tech behavior (substantiated by their public policies or product behavior) with ours. If anything below stops being true — for them or for us — we update this page within seven days.
Big Tech
Google's privacy policy explicitly permits using Gemini conversations to "improve services" — read in 2026, the policy includes training and personalization paths. Read the policy yourself: it's there.
Journal Genie
We have not opted in to training with our model providers. Your journal, your sources, your chats — never training data.
Big Tech
ChatGPT does not retain durable memory of your sources or context between sessions by default. Each conversation starts cold. The product surface that "remembers" runs server-side and is governed by OpenAI's data policy.
Journal Genie
Journal Genie persists your sources, journals, surveys, and Studio artifacts in YOUR account. We are the durable-memory layer ChatGPT doesn't have. And we hold ourselves to a "no training" stance with our providers.
Big Tech
Big Tech AI products are funded by ads, attention capture, or training-data extraction. The business model is the conflict of interest.
Journal Genie
Journal Genie is funded by subscriptions. Free has a real ceiling. Pro is $12/month. That's the entire business model.
Named third-party processors
Each row says what they hold, what they can't do, and how long they hold it.
Supabase
Authentication + the account-backed database. Holds your encrypted account data. Cannot read journal content as cleartext outside the RLS-scoped query path. Daily backups; 30-day point-in-time recovery on managed Postgres.
OpenAI
Model calls for grounded chat and question generation. We send the prompt + retrieved context for each call; we do not send your account identifier in conversational form. We have not opted into training. Conversation history is not retained by OpenAI per our API tier's data retention policy.
Stripe
Payment processing when paid billing is used. We never see your card. PCI scope sits entirely with Stripe.
Sentry + Better Stack
Observability. Error tracking and synthetic monitoring. Personally identifying fields in logs are redacted server-side via the logger's automatic-redaction layer.
Where we're early
The honest list. We'd rather you read it here than catch us off-guard.
- Counsel review on the formal Privacy + Terms text is in flight. The plain-language preface on this page is final; the formal policy text is provisional and marked as such.
- GENIE on-chain claims are pre-deployment. The points system works today; claiming as tokens waits for the on-chain distributor.
- Crisis-redirect safety surface in Journal is queued (separate run). Today, Journal is private + opt-in for Genie inclusion; the explicit safety affordance follows.
- The Insights API is on the roadmap. It will be aggregate-only with strict thresholds and never raw individual data — see /privacy for the language.
- Full incident-response runbook is being prepared with counsel input before public-launch broadcast.
Trust scales when the surfaces agree.
Pricing matches Trust. Privacy matches Terms. Security matches both. We hardened these together so the language doesn't contradict itself.