What we collect

Both lines below were produced just now by the same function your browser runs before anything is sent. The example is made up; the scrubbing is not.

You typecold email to sam.draper@acme.com about the Q3 invoice, cc @finance_team — call me on (555) 123-4567 or see https://acme.com/inv/88421

We storecold email to [EMAIL] about the Q3 invoice, cc [HANDLE] — call me on [PHONE] or see [URL]

Tally: {"EMAIL":1,"URL":1,"PHONE":1,"HANDLE":1} — counts only. The scrubber is built so it can never emit a value it removed. It recognises SSN, EMAIL, URL, CARD, PHONE and HANDLE.

rough

Your ask, scrubbed to typed placeholders.

refine

The optional second line, scrubbed the same way.

domain

Which of the 21 domains it was routed to.

moves

Each move: its key, its label, its value, and whether you supplied it, we added it, or we left it blank.

fills

Which blanks you filled and the words you used — scrubbed. The most valuable field here, and the only one that is volunteered rather than inferred.

engine

Which tier answered: template or AI.

redaction

How many redactions ran, by type — never the values. It counts passes, not people: your ask is echoed into the Task move, so one email in one sentence is scrubbed twice and tallied twice.

ts

When, to the millisecond. Set on the server; your clock is not trusted.

v / frameworkVer

Which capture and framework version produced the row, so old rows stay readable.

{
  "v": 1,
  "ts": 1789914600000,
  "rough": "cold email to [EMAIL] about the Q3 invoice, cc [HANDLE] — call me on [PHONE] or see [URL]",
  "refine": "",
  "domain": "Email & outreach",
  "engine": "template",
  "moves": [
    {
      "key": "task",
      "label": "Task",
      "value": "Cold email to [EMAIL] about the Q3 invoice, cc [HANDLE] — call me on [PHONE] or see [URL]",
      "source": "kept"
    },
    {
      "key": "audience",
      "label": "Recipient & relationship",
      "value": "Assume a cold recipient with no prior relationship; open by placing yourself (who you are + why them specifically) in one line.",
      "source": "added"
    },
    {
      "key": "context_hook",
      "label": "Why-you-why-them hook",
      "value": "Lead with one specific, true reason tied to the recipient, not a generic compliment.",
      "source": "added"
    },
    {
      "key": "tone",
      "label": "Tone & register",
      "value": "Warm, professional, direct; conversational not chummy; no corporate filler or hype.",
      "source": "added"
    },
    {
      "key": "format",
      "label": "Length & shape",
      "value": "Under 120 words, 3 short paragraphs, one clear ask, plus a subject line.",
      "source": "added"
    },
    {
      "key": "subject_line",
      "label": "Subject line",
      "value": "Offer 2-3 options: specific, under 6 words, no clickbait or all-caps.",
      "source": "added"
    },
    {
      "key": "cta",
      "label": "Call to action / next step",
      "value": "Close with a specific, low-effort next step actionable in a single reply.",
      "source": "added"
    }
  ],
  "fills": [],
  "frameworkVer": "v1",
  "redaction": {
    "EMAIL": 2,
    "URL": 2,
    "PHONE": 2,
    "HANDLE": 2
  }
}

Your raw text

Scrubbed in the browser before it is sent, and again on the server before it is written.

Your IP address

The collect endpoint does not read it. Not hashed, not bucketed, not stored — never touched.

Your user-agent

Not read.

Cookies or any session

The endpoint sets none and requires none. There is no account to attach a row to.

Anything that links two rows

There is no id field, no session and no account — a row carries no handle on the person who produced it. Nothing in the data points from one of your sharpens to another.

Everything above is about what a sharpen leaves behind. Simply visiting also registers a page view, through Vercel’s analytics. It is a different thing with a different answer, so it gets its own list rather than being folded into the one above.

What is counted

That a page was viewed, which page, and roughly where in the world from. Plus the usual shape of the device — mobile or desktop, which browser.

No cookies

None are set, for this or anything else. Nothing is written to your browser to recognise you later.

No profile

Views are counted, not followed. There is no identifier joining two page views into a person, here or anywhere.

Private mode stops it too

The same toggle. With it on, the page view is dropped in your browser before any request is made — not received and discarded.

Why have it at all: the capture log only ever sees people who typed something. It is blind to anyone who opened the page, didn’t know what to write, and left — and that person is the one worth knowing about.

Once in your browser, before anything crosses the network. Then again on the server, before anything is written. The second pass is redundant when everything works — which is exactly why it is there. It has to survive a bug in the first pass, a stale cached bundle, or someone posting to the endpoint by hand.

Tick it and the page makes no network call — not for the sharpen, not for the scrubbed row, and not for the page view. Not “we receive it and throw it away” — there is nothing to throw away, because nothing is sent. The setting lives in your browser, and if storage is blocked it defaults to private rather than to sharing.