Fingerprintscore Chart form FS-1 Rubric 1.0.0 Priors: assumed · Counts: real Donut Browser ↗

Did the change move anything?

One total tells you almost nothing on its own. Two passes on the same stock tell you which pen moved, by how much, and — more usefully — which pens your change never reached.

This is the one comparison on the site that needs no assumptions about anybody else. A delta between two of your own passes is arithmetic on two readings taken under the same printed priors, so it survives every objection to the priors themselves: if the table is wrong, it is wrong identically in both passes and the difference is unaffected.

No passes stored in this browser

A pass is stored here only when you run one. Go to the chart form, press RUN, change one thing in your setup, then press ADVANCE — the second pass lands beside the first and this page will overlay them. Nothing is stored on a server, and clearing site data removes them.


How to read a delta

The baseline prints at 35% ink and the current pass prints over it in full, on the same form. A pen that lies exactly on its ghost did not move. A pen that jumped a printed envelope boundary changed category, not just value.

Watch the black summing pen last. It is the only line that carries the discount, so a signal that moved a lot but was discounted against a stronger groupmate barely bends it — which is the correct outcome and the one that surprises people.

What a delta cannot tell you

It cannot tell you whether a site would have identified you either time, because that depends on what the site collects and correlates. It cannot tell you that a lower total is safer: a value that is rare under these priors but internally consistent is often less interesting to a vendor than a common value that contradicts another signal.

And a delta across two different rubric versions means nothing at all, which is why every stored pass records the version it was scored under.


Four deltas worth taking deliberately

Before and after a proxy. The exit changes, and you find out whether your stack moved the time zone and the language list with it. If those two pens sit exactly on their ghosts while your exit country changed, nothing in your setup is following the proxy.

Before and after a fresh profile. Two profiles that are supposed to be different machines should not produce two identical traces. Identical readings across profiles is the failure mode that matters most to a multi-account operator, and it is visible here in one screen.

Before and after a browser update. Renderer strings and canvas output move on driver and engine changes without anyone intervening, so a baseline that is six months old is comparing against a machine you no longer have.

Before and after a spoofing tool is enabled. This is the one that settles arguments, because the columns that refuse to move are a direct statement about the level the tool operates at.

Reading the ghost trace

The baseline re-prints at 35% ink on the same form and the new pass draws over it. Three shapes are worth learning. A pen lying exactly on its ghost for its whole travel did not move at all. A pen that leaves its ghost and settles at a different height inside the same printed envelope changed value within the same price band. A pen that lands outside the band its ghost occupied changed category — the rule returned a different prior, not just a different value.

The fourth shape is a lift: a pen that drew last time and does not draw now. That is a rule declining to read, and it removes bits from the total without improving anything about the browser.

Which columns can be moved and what it costs · why a big per-column delta can produce a small total delta · how far a total has to move to cross an edge


Questions about comparing two passes

Comparison happens entirely in this browser. Nothing on this page sends a pass anywhere.

What counts as a meaningful delta?
Anything that changes which columns carry full weight, and anything above a bit or two on a column that was already dominant in its group. A total that moved by 0.4 bits usually means a value changed inside a bucket the rubric prices identically, which is a change to your machine and not to your reading.
Where are my saved passes stored?
In this browser, in IndexedDB, and nowhere else. There is no account, no server-side history and no second copy. Clearing site data for this domain deletes them, and a pass taken in one browser is invisible to another.
Can I compare passes taken under different rubric versions?
You can put them side by side, but the difference would be meaningless: a version change moves the priors themselves, so both readings would shift for reasons that have nothing to do with your configuration. Every stored pass records the version it was scored under so this is visible rather than silent.
Does re-running the check store another reading?
Not every time. The site stores one reading per browser, and again after a browser update or 30 days. Repeat passes in between are scored in your browser and displayed, so a comparison can be repeated as often as you like.
Why did a column disappear between two passes?
Because the rule declined to read it the second time. A blocked WebGL context, a refused audio context or an emptied language list makes the rule return null, the pen lifts, and the column leaves the sum entirely rather than being scored as zero.

Where the pens move

If the same pens refuse to move whatever you change, the change is not happening at the engine.

Engine level, not JS level. Donut Browser spoofs through CDP injection in its Wayfern Chromium fork, so the value a page reads is the value the engine reports — there is no patched accessor to catch.

It refuses to contradict itself. A profile whose proxy exit geography, time zone or language disagrees with its fingerprint does not launch. That is a pre-launch gate in the app, not advice in a manual.

Open, and quiet. AGPL-3.0, desktop app and sync server both in the repository, and no analytics SDK anywhere in the source. macOS, Windows, Linux.

Download Donut Browser Source on GitHub — AGPL-3.0 ↗