Argument

Nobody's Motion

Why American vexatious-litigant doctrine can count and cannot act — a named case study in the difference between permission to look and an obligation to.

This section argues; the rest of the site reports

Everywhere else, this site publishes records and labels what kind of evidence each one is. These 8 chapters are the one place it draws a conclusion from them — and they add no facts of their own. Every factual claim here is already published as a row on this site, and the case, dispute and update ids in the text link to the page that carries it. No court in these records has declared Frank Tufano a vexatious litigant, and nothing in this section says otherwise; the argument is about what the doctrine can and cannot do, not about what he should be called.

Chapters8
Words55,992
Originating actions98
Courts7

Abstract

American vexatious-litigant doctrine is usually described as docket-local: a court restrains a filer on the basis of repetition visible on its own docket. Reading the four regimes that govern the forums in this study — the All Writs Act as applied in the Second and Third Circuits, Pa. R.C.P. No. 233.1, 22 NYCRR 130-1.1, and Cal. Code Civ. Proc. § 391 — against their primary texts, this book finds that description false. Three of the four predicates reach a prior action in any court; California's does so by definition. What every regime shares is a different asymmetry: each is narrower on the remedy side than on the predicate side, each fires only on a private party's motion or not at all, and none obliges anyone to assemble the record its predicate is permitted to count.

The demonstration is a single named record. As of 11 September 2026 the corpus holds 97 originating actions, 93 of them commenced by one litigant across seven trial forums in three states since March 2018, de-duplicated from 107 trial-court docket records by a method that exactly reproduces the two counts judges have put on the record — 22 and 38, both in one federal district — before lifting the forum filter, at which point the same method returns 65 and 85. Against that record: 72 court findings by 27 judges, zero filing restrictions, zero monetary sanctions on any action he commenced, and two vexatious-litigant requests, neither adjudicated. A replay of every statutory counting threshold shows each was satisfied years before any judge counted, under every aperture that could have applied. Detection was never the binding constraint.

The finding is about the instrument, not the litigant — whom no court has adjudicated a vexatious litigant. The doctrine can count further than anyone has asked it to, and it cannot oblige anyone to ask. The reform that follows is a trigger and a consequence, not a wider aperture.

Revised 2026-09-11, per two external reviews. This block previously asserted that doctrine is "docket-local in operation," that "nothing in the system makes repetition elsewhere visible," and that "the invisibility widens as the filing rate climbs." The first is refuted on the text by chapter 2, the second by chapter 5 (the filer narrates the cross-forum record himself, continuously and in public) and by chapter 2 §VI (a defendant assembled it by hand), and the third by chapter 1 §VI, which concedes the share fell from 65.1% to 54.2% over the window the two footnotes bracket. Chapter 7 found detection was never the binding constraint and chapter 8 §IV retracted the keyhole image outright. The book's finding has been motion-dependence since chapter 2 was drafted; what changed on this date is that the front matter says so. The policy stake in the difference is not cosmetic: a blind-spot thesis points at registers and aggregators, which chapter 7 shows would have changed nothing here, while a motion-dependence thesis points at who is obliged to look.

Chapters

Chapter 1

Two footnotes

Two federal judges counted this plaintiff’s filings in footnotes, at 22 and 38. Both were accurate. Across every court he had filed in: 65 and 85.

3,578 words · as of 2026-08-28
Chapter 2

The four regimes

The All Writs Act, Pa. R.C.P. 233.1, 22 NYCRR 130-1.1 and Cal. C.C.P. § 391 read against their texts. Three of four already reach any court.

8,264 words · as of 2026-08-28
Chapter 3

Constructing the record

The method: dockets from seven courts in three states, removals de-duplicated into originating actions, party names canonicalised, limits stated.

5,222 words · as of 2026-08-28
Chapter 4

The distributed filer

Twenty-two dispute clusters, eighteen crossing a forum boundary. Why removal concentrates a record, and why the escalation ladder is not a ladder.

5,884 words · as of 2026-09-10
Chapter 5

The narrated docket

The aggregation no institution performed was being performed aloud: 1,691 public statements, 395 naming cases in two or more courts.

5,996 words · as of 2026-09-11
Chapter 6

The cost side

§ 1914(a) prices a docket while the duty to respond accrues per party: one $405 fee named 144 defendants, $2.81 a head. No monetary sanctions.

7,615 words · as of 2026-09-06
Chapter 7

Designing an aggregator

Four remedies replayed against the record. Every threshold fires well before action thirty, so detection was never the binding constraint.

8,359 words · as of 2026-09-03
Chapter 8

The null result

Ninety-seven actions, seventy-two court findings, no filing restriction anywhere. Nothing converts a visible pattern into a consequence.

5,881 words · as of 2026-09-06
Coding protocol A

The RefilingOfPriorCase rule

The published rule behind the RefilingOfPriorCase coding: the screen it was audited against, 80.4% agreement, and the stated limitations.

2,209 words
Coding protocol B

The escalation ladder

The rule for coding who a filing names — counterparty, counsel, platform, regulator, government, adjudicator — and why the values stay unpublished.

2,984 words

Corpus ledger

Each chapter was written against the corpus as it stood on its own date, and the corpus kept growing while the book was written. A figure in an earlier chapter is therefore smaller than the same figure in a later one, and neither is wrong. This table is what a reader needs to tell a dated figure from a disagreement; it is generated at build time from the same scan that fails the build when a chapter contradicts itself.

ChapterAs ofActionsRecordsForumsPartiesStatements
1 · Two footnotes2026-08-28921025
2 · The four regimes2026-08-285
3 · Constructing the record2026-08-281025
4 · The distributed filer2026-09-10328
5 · The narrated docket2026-09-111691
6 · The cost side2026-09-06953271667
7 · Designing an aggregator2026-09-0394104
8 · The null result2026-09-069510563271667
Live at this build2026-09-239810873411722

A blank cell means the chapter states no figure in that column — not that the figure was unknown to it. The corpus grows by discovery as well as by filing, so an earlier chapter’s total cannot be recovered by filtering a later database on dates.

A note on dates

The dataset is a moving target — 33 originating actions (38 docket records) were filed in the first eight months of 2026 alone, and the corpus moved from 92/102 to 94/104 between chapter 3 and chapter 5, three days apart, then to 95/105 when C105 arrived on 2026-09-06. Mind the grain: originating actions and docket records diverge, and quoting one figure for the other is the easiest mistake to make here. Every count in a chapter needs an explicit as-of date, and any figure should be regenerated rather than copied forward. src/seo-build-report.json carries the current totals after each build, and thesis/exhibits/figure_audit.py fails the moment a chapter dated today stops matching them.

An as-of date is not a filter, and this limits what a reader can check. The corpus grows by discovery as well as by filing, and the two are different clocks: C103 and C104 were filed in September and November 2025 but only found on 2026-08-30, and C105 was filed 2026-09-01 and found on 2026-09-06. So recomputing "as of 2026-08-28" against a later database returns 94/104, not the 92/102 chapters 1 to 3 correctly state, and nothing in the database records when a row was added, so the earlier states cannot be reconstructed from it. This is why chapters are revised for defects and not re-baselined: recomputing a dated figure would replace what was known with what is known now. Chapter 3's limitation 8 states it for readers.

One recorded exception, chapter 5, 2026-09-11. The rule holds for a figure. It does not hold for a distribution measured on half its data. Chapter 5 was drafted on 2026-08-31 against 817 statements from 333 videos; eleven days later the corpus held 1,691 from 620, because the archive feeding it was being built at the same time as the chapter. Two external reviews made the same objection and it is correct: "two-thirds of the statements carry no filing-conduct factor" is a claim about the shape of a population, and an as-of date does not rescue it if the population has doubled. That chapter was therefore recomputed and its earlier figures retired, with the superseded values kept in its Sources table so nothing is lost. The re-run strengthened the chapter rather than weakening it, and every dated narrative claim in it survived unchanged. Treat this as the bar for any future exception: a distribution, not a count, and only when the population it describes has moved by enough to change its shape.

MethodologyHow the evidence layers are kept apartThe recordsEvery docket this argues fromSourcesThe public-record audit trailCorrectionsHow to report an error