SPS 2026 Annual Meeting CSPT / SCPT

Second Cortex · data flow

What happens to your document

Workshop key · SPS 2026
Paste this OpenRouter API key into Second Cortex when it asks for one. Shared budget, capped, expires Sept 27 — nothing is billed to you.

Second Cortex is a web page, not a service. Reading, indexing and searching your documents all happen inside your own browser. The one thing that travels is a single request per question, and this page shows exactly what is in it.

Stays in your browser

Your files, the extracted text, the passages, their vectors, the search itself, your notes and your question history.

Leaves, once per question

Your question plus up to 10 excerpts of about 190 words each, to OpenRouter and on to the model, with zero data retention enforced.

Enforced, not promised

The page's security policy lets the browser connect to two places only: the app's own site and openrouter.ai. No server of ours, no accounts, no analytics.

  • Carries content from your documents
  • Goes out, but carries no document content
  • Comes in (download or answer)

Before anything

When you open the app

Downloads only

Your browser, on your laptop

The network

The internet

1

Open the web address

The browser downloads the app, the small language model used for indexing (about 129 MB, first visit only, then cached) and a ready-made index of the six ANX-4471 workshop files.

second-cortex.vercel.app · static files only

Comes in: app + indexing model + workshop index

Vercel, static hosting

Hands out files, like any website. There is no backend behind it and nowhere to upload to.

Receives an ordinary page request. Nothing from your documents, ever.

Phase 1

When you load a document

Nothing leaves your browser

Your browser, on your laptop

The network

The internet

3

Drop a file on the Library tab

PDF, Word, Markdown or plain text. "Upload" here means the browser reads the file from your own disk. A fingerprint of the file spots one you have already loaded.

File API · SHA-256 via crypto.subtle

Phase 2

When you ask a question

Question + excerpts leave

Your browser, on your laptop

The network

The internet

8

Turn the question into numbers

Same in-browser model as step 6. The question has not gone anywhere yet.

embedTexts(question) · 384 dimensions

Zoom in on the red arrow in step 10

What is inside the request

Sent

  • Your question, as typed
  • Up to 10 excerpts, each at most 190 words, labelled with document title and page
  • The instruction to answer only from those excerpts, in English or French
  • The chosen model name and the zero-data-retention rule
  • The OpenRouter key, as the authorisation header, and the app's name

Never sent

  • The original file, in any form
  • The rest of the document: every passage that was not in the top 10
  • Your earlier questions and answers
  • Your notes, unless a note is itself one of the 10 closest passages
  • Your name, email or any account of yours: the app has none
POST https://openrouter.ai/api/v1/chat/completions
Authorization: Bearer <OpenRouter key>

{
  "model": "anthropic/claude-sonnet-5",
  "provider": { "zdr": true },            // zero data retention, or refuse
  "temperature": 0.2,
  "stream": true,
  "messages": [
    { "role": "system",
      "content": "You are Second Cortex ... Answer ONLY using the numbered excerpts below.
                  If the excerpts do not contain the answer, say so plainly ...

                  [1] (ANX-SPH-003 ..., p. 14)  <excerpt, up to 190 words>
                  ...
                  [10] (ANX-TX-004 ..., p. 31)  <excerpt, up to 190 words>" },
    { "role": "user", "content": "<your question>" }
  ]
}

Why nothing confidential

Once a real document is loaded, excerpts of it go to the model provider with your question. The Library box says so. Published guidance, your own papers, a protocol template: yes. Anything under confidentiality: no.

"Nothing leaves your machine" would be the wrong sentence. The accurate one: the file never leaves; up to ten short excerpts per question do.

Where things rest afterwards

WhatWhere it livesDoes it ever leave the browser?
Files you loadIndexedDB in this browser, and your chosen folder if you turn that onNo
Passages and their numbersThe same in-browser databaseOnly the 10 closest passages, per question. The numbers themselves never do.
NotesIndexedDB in this browser. Notes are indexed like documents.Only when a note is among the 10 closest passages to a question
Question historyIndexedDB in this browserNo
OpenRouter keyThis browser's local storage, in plain textTo OpenRouter, with each request. "Log out" removes it from the browser.
Export fileWherever you save itOnly if you move it. It contains everything above, the key included in plain text, so treat it like a password.

Two tools, one morning: how the data paths differ

Second Cortex The two-prompt file
Runs fromA web address (static files on Vercel)One HTML file on your desktop
ReadsPDF, Word, Markdown, plain text, plus your notesPDF only
Finds passages byMeaning: an in-browser language model compares vectors, across English, French and GermanKeywords: the words you typed must appear on the page
Sent per questionQuestion + up to 10 excerpts of at most 190 wordsQuestion + the full text of up to 8 pages
ModelClaude Sonnet 5 by default; short menu of zero-retention-eligible modelsClaude Sonnet 5, fixed
Zero data retentionEnforced on every request; refuses rather than fall backNot requested; depends on the key owner's OpenRouter settings
Where the browser may connectIts own site and openrouter.ai only, enforced by the page's security policycdnjs (for pdf.js) and openrouter.ai; no policy restricts it
Kept between visitsLibrary, notes, history and key, in this browserThe key only. Documents are reloaded each time.
Both send the request straight from your browser to OpenRouter. Neither has a server of its own in between.