🧵
1/5
The NY Times reports FBI agents determined the Bexar voter file anomalies were most likely caused by a drag-and-drop error when county officials exported data from poll pad devices into Excel. I just published four mathematical proofs that no drag-and-drop error — at any stage, on any computer — could have produced this file.
@WestonMartinez @ZoomWalter @Lorionafarm @PeterBernegger 2/5
New finding: the first fake voter ID — 1,253,115,467.79993 — is no longer mysterious. Run a linear regression on the first 735 records of the original source file. The line at position 736 predicts that number exactly. Zero difference. Five decimal places. Reproducible by anyone in Excel in under a minute.

1/
"OK, but how many votes were affected?"
As a voter, I care.
As a researcher, I couldn't care less.
My research isn't about "who won?"
It's about " Why on earth was that election certified?"
No one can tell you how many people voted, let alone how many of those votes are legitimate. 2/
If a bank robber walks in, steals money, and corrupts the bank's official records of how much money they had — you no longer know how much was stolen.This is why, in accounting, the 'Threshold of Materiality' becomes extremely sensitive to any intentional falsehood. Once an official is willing to certify a statement containing known falsehoods, the specific dollar amounts no longer matter. The willingness to certify something false is itself material.This is what ultimately contributed to Enron's collapse — and many other corporate failures.

[1/3]
Did Hochul really win New York?
New York's voter rolls contain nearly twice as many illegal duplicate registrations — skewed Democratic — as votes that separated Hochul from Zeldin in 2022.
Did the clones vote?
I tried to find out. Here's what I found. 🧵 [2/3]
The rolls carry 2.2 million clone registrations — one in ten records. Illegal by definition. Same person, two or more different state IDs.
In the four suburban counties that decide every statewide race —
Nassau, Suffolk, Rockland, Westchester — the clones run 6–7 points more Democratic than the legitimate voters around them.
Statewide clone DEM surplus: 646,000 registrations.
Hochul's margin: 327,000 votes.
The surplus is 1.98× the margin. In the right direction.

The @nytimes recently published a story about Dan Bishop's investigation into Bexar County's recent Republican primary, and a mysterious file that surfaced during that election.
They attribute the creation of the file to a drag and drop error in Excel--something that is easy to do. In this context, they may as well have said that a train crashing into a Volkswagen creates a Maserati instead of a crushed Volkswagen.
Here's why they're wrong /1 A glitch, a crash, a drag-and-drop error — these all share one property: they damage or degrade what's already there. A train hitting a Volkswagen leaves recognizable Volkswagen parts. It does not manufacture a Maserati from the wreckage. Glitches destroy structure. They do not create it. Keep that in mind. /2

Arizona's voter database has a problem.
590,529 duplicate registrations. A hidden algorithm running in all 15 counties. And the federal law designed to fix it made things worse.
🧵 This is Pima County's voter registration database.
Each dot is a voter ID plotted against a registration ID. In a normal database these form a clean diagonal line.
This is not a normal database.

1/This is a table of voters named Meyers in Wisconsin.
Same last name. Same address. Same registration date.
Ten different ID numbers — ranging from 200 million to 1.1 billion.
One person. Ten identities. In an official government database. 2/That alone should be disqualifying.
But it gets stranger. Because those ten IDs are not random. They are part of a mathematical structure running through millions of Wisconsin voter records.
A structure that has no business being there.

New York's voter ID system contains a hidden algorithm. I reverse-engineered it. The complete solution is in this thread. 🧵 The Spiral maps County Voter IDs (CIDs) to State Board of Elections IDs (SBOEIDs) using a deterministic, reversible structure — predictable and built for data retrieval. Reversing it yields a serial marker I call the Algorithm ID (AID).

HAVA was supposed to help Americans vote. The data suggests it helped Boards of Elections generate millions of illegal registrations instead. A thread. 🧵 In New York — where I have done the deepest forensic work — clone records were 0.25% of the voter roll in 1990. By 2022: 17.36%. One in six records. The inflection begins precisely at HAVA implementation.

When New York Citizens Audit brought evidence of problems in New York's voter rolls to state officials, the State Board of Elections called them "wack jobs."
That word came from a board member. In an official meeting. On the record.
Here is what else happened in that room. 🧵 Board co-chair Peter Kosinski asked a reasonable question: if NYCA has made errors, why not meet with them and explain where they went wrong?
His own staff shot it down.
The reason they gave: NYCA "willfully misconstrues the truth."
No meeting. No rebuttal. No explanation.

A former NYPD detective looked at what I found hidden in New York's voter rolls and said one word:
"Plutonium."
The NY State Police Special Investigations Unit agreed.
Law enforcement and military intelligence both gave me the same instruction: publish fast.
This thread is why. 🧵 Every New York voter has a State Board of Elections ID — a 20-character number that is supposed to be a meaningless serial number.
999 quadrillion possible unique values per state.
It is not meaningless.
Someone engineered four hidden algorithms into New York's SBOEID number space. I found them.

1/4 Where are the dead voters hiding?
In New York, ask for their address. It's any voter ID assigned by the Shingle algorithm.
99.9% of Shingle records belong to purged voters — people who are deceased, moved away, or otherwise no longer eligible.
That is roughly 294,000 records. Over a quarter million voters who cannot vote, all concentrated in one algorithmic zone.
Purged records don't get audited. Nobody checks on dead voters.
A quarter million is enough to decide any election that isn't a landslide.
That's the point. New York's voter database uses 4 algorithms to assign ID numbers.
Spiral: 48.86% purged Metronome: 48.98% purged Tartan: 35.68% purged Shingle: 99.9% purged
One of these is not like the others. 2/4

I can look at a voter ID number in New York's database and tell you the record is purged — without looking at the status field.
I can tell you it carries a registration date inconsistent with its algorithmic placement. I can tell you it is almost certainly a clone — a duplicate registration capable of functioning independently in the system. And I can tell you it was assigned a brand new state ID number despite being ineligible to vote at the moment that number was assigned.
In a legitimate database, an ID number is administrative. It tells you nothing about status or authenticity. You check those things by looking at the relevant fields.
Unless someone built the status into the numbers themselves.

New York assigns voters two ID numbers: a County ID (CID) and a State Board of Elections ID (SBOEID). A collaborator — a programmer who has asked not to be named — flagged an unusual geometric pattern in Nassau County's ID structure and brought it to my attention before leaving the project. The analysis that follows is entirely my own.
When I plotted Nassau County's out-of-range SBOEID numbers against their CIDs, I found something that has no business in a voter registration database.
Geometry.
Not scatter. Not noise. Overlapping rectangular bands ascending in a precise staircase formation, each block stepping up and to the right like shingles on a roof. I named it the Shingle algorithm.
Real voter data does not produce geometry. When you see a pattern this clean in a dataset this large, you are no longer looking at registrations. You are looking at generation.

Nassau County's Shingle section contains approximately 176,090 records in the 2021 database. Nassau is not unique. The Shingle algorithm appears in all of the counties that use the Metronome algorithm for their in-range records — Nassau, Erie, and Westchester — plus at least two others including Onondaga, for a statewide total of approximately 700,000 Shingle records. Nassau simply has the largest concentration and is where the pattern was first identified. Here is what all of these records have in common.
Essentially 100% are purged. Not inactive — purged. Permanently removed from the rolls, ineligible to vote. And here is the critical point: every State Board of Elections ID number in this database was newly created when New York implemented the Help America Vote Act around June 2007. Before that, only county ID numbers existed. No one had a state ID. These were not legacy numbers carried over from a prior system. They were freshly assigned.
Which means someone made a deliberate decision to assign brand new, algorithmically structured state IDs to records that were already purged — or were purged at the exact moment of assignment. I call them born purged.
The Shingle records are not all of the pre-2007 purged records. They are a minority of them. That matters enormously. If the algorithm had simply swept up every old purged record, you might construct an administrative explanation. Instead it selected a specific subset, gave them a specific mathematical identity, and left the rest alone. That is not maintenance. That is classification.
They carry pre-2007 registration dates — making them appear to be historical registrations predating the algorithm that assigned their IDs. They are the only out-of-range records with pre-2007 dates. Every other out-of-range record, across all algorithms, carries a post-2007 date. The inference that these dates are inconsistent with the records' actual origin is difficult to avoid.
In the upper half of the CID range, nearly 100% are also clones — duplicate registrations with different ID numbers, each capable of operating independently in the system. A clone is not an administrative error. An error produces a duplicate with the same ID, which is immediately visible and non-functional. A clone has a different ID and can, in principle, vote.
Their purge status, clone relationship, and algorithmic identity can all be read from their ID numbers alone — before you look at any other field.

The Shingle records share ID space with a second algorithm I named the Tartan, which governs the majority of New York's out-of-range records. The two populations occupy the same numerical territory without overlapping. Whether that reflects deliberate coordination or simply the Tartan working around numbers that were already assigned, the result is the same: the Shingle records sit in a mathematically distinct space of their own, identifiable by algorithm alone.

I don't know the operational purpose of these records. The data does not settle that question and I will not speculate beyond what it supports.
What the data does establish: Nassau County's voter database contains approximately 176,090 records that were assigned fresh state ID numbers under a specific algorithm, carry purge status that appears to have been present from the moment of assignment, hold registration dates inconsistent with the algorithmic structure they occupy, and correlate heavily with clone status in the upper ID range. They represent a minority of all pre-2007 purged records, meaning they were specifically selected for this treatment.
No legitimate database process produces this. No administrative explanation accounts for assigning new, structured state IDs to records that cannot vote.
Which leaves one question the data cannot answer.
Why would you assign a new state ID number — one embedded in a sophisticated, hidden algorithm — to a record that is already ineligible to vote? A record that should, by any normal administrative standard, have been deleted or left with its county ID and nothing more?
That question has no innocent answer I have been able to find. It is a question someone with subpoena power should be directing at the people who built this system. This research is part of a multi-state forensic analysis of voter registration databases published in the Journal of Information Warfare. Full methodology, data, and citations are available at zarkfiles.substack.com and researchgate.net/profile/Andrew…
Check out my Kickstarter election fraud card game here: kickstarter.com/projects/paqph…