Transparency

🔍 The Verifier

Every claim we make about privacy is backed by code you can inspect. Below is the exact extraction logic that runs inside your browser when you upload a file — and a live demo you can run yourself.

How to verify independently: Open your browser's DevTools (F12) → Network tab → upload a file on /app → inspect the request to /api/generate. You'll see exactly 15 numbers in the payload. No message text. No topics. No names.

The extraction function

This is the core logic that runs in your browser. It scans your messages and produces aggregate numbers — it never transmits the messages themselves.

Loading...

Live demo — try it yourself

Paste a few lines from one of your AI conversations. The same logic that runs on your real export will extract metadata from this text and show you what gets sent.

What would be sent to the API
That's it. No message text, no topics, no personal details — just these numbers.

The 15 numbers that leave your device

Our API receives exactly these aggregate statistics. Nothing else is transmitted — no raw text, no topics, no names, no URLs.

1 Total conversations
2 User message count
3 Avg words per message
4 Question ratio (%)
5 Pushback ratio (%)
6 Deep-dive ratio (%)
7 Vocabulary richness score
8 Depth score
9 Long message % (>100 words)
10 Peak hour of day
11 Night usage %
12 Data span (months)
13 Integrity / confidence score
14 First & last date (month only)
15 Month-by-month activity shape

Claude (the AI) receives these numbers and writes your profile. It never sees your actual conversation content. The system prompt explicitly instructs it to work only from the numbers provided.


Satisfied? Generate your own profile — it takes under a minute.

Generate your profile →