The null result

It also introduces no new law. Every authority it restates was read from a free primary host by an earlier chapter and is listed, with that chapter named, in Legal authorities at the end — kept separate from Sources so that the rule this folder runs on, that no fact about the litigation originates in it, stays checkable by inspection rather than by trust.

Revision, 2026-09-11 (second). Section V gains the appellate-fee counterexample. An external review pointed out that the chapter claims "nothing prices the exit" while chapter 6 records three appeals dismissed outright for unpaid fees of $91.25 and $605 — the only gate in the whole record that has repeatedly stopped a proceeding, and one that needed no motion from anyone. It is a counterexample to the sentence and the proof of concept for the reform, and it belongs in the conclusion rather than only in chapter 6. Section III also restates chapter 4's and chapter 6's verdicts as those chapters now report them.

Revision, 2026-09-11. Section III used to read as a confession, because the book's front matter still advertised the four propositions this section reports as refuted. The front matter now states the finding the exhibits support, so section III is what it always should have been — a record of method — and its opening says so. Section IV no longer retracts chapter 1's keyhole image, because chapter 1 no longer deploys it: that chapter now poses the question this section answers. No figure changes.

Revision, 2026-09-06 (later the same day). Three things moved after this chapter was drafted. The canonical-entity count fell from 335 to 327 when eight duplicate party records were merged — the same defect chapter 4 names, plus six more found by rewriting a gate to recompute the pattern rather than remember its known instances. The appellate layer became first-class and is now measured, so section VII's limitation 4 is retired. And the claim in section I about what chapters 1 to 7 say about forums was wrong: chapters 4, 5 and 6 state no forum count. Every other figure is unchanged.

I. Where the record stands

As of 6 September 2026 the corpus holds 95 originating actions, de-duplicated by union-find over SameActionAsCaseID from 105 docket records. Ninety-one of those actions were commenced by this plaintiff; four were filed against him and are excluded from every figure about his own filing. They are distributed across six courts in three states:

The action that opened the sixth forum was filed five days before this chapter's as-of date. Chapters 1, 2, 3 and 7 say "five trial forums"; that was true when each was written, and section VI returns to the case that changed it. Chapters 4, 5 and 6 state no forum count at all, so nothing in them needed changing.

Against those actions the corpus records 66 court findings involving 26 distinct judges, 327 canonical party entities, 22 dispute groups, 35 filing-restriction indicators, and 1,667 timestamped public statements drawn from 616 in-scope videos.

The two figures the whole argument rests on are unchanged in kind and larger in degree. On 3 February 2026 Magistrate Judge Latella's footnote in C055 counted "at least twenty-two separate actions in this Court"; restricted to that forum and that date, the same union-find returns 22, and unrestricted it returns 65. On 13 July 2026 Judge Munley's order in C057 recorded 38 M.D. Pa. lawsuits since December 2023; the same method returns 38 in forum and 85 across all forums. The exact reproduction is the method's licence for the second number, and detection_gap.py exits non-zero if it ever stops reproducing them.

The gaps are 43 and 47. The highest number any court has stated about this record is 38.

II. The null result

Set beside that record, the outcome column is empty, and it is empty in a specific way worth stating precisely.

• Zero pre-filing orders. No filing restriction of any kind has been entered against this plaintiff in any tracked case, in any forum.
• Zero monetary sanctions on any action he commenced. None has been verified anywhere in the corpus.
• Two sanctions in total, both non-monetary, both terminating, both entered by the same judge in the same department of the Los Angeles Superior Court on that court's own order to show cause, for failing to appear (C101, 13 October 2025; C102, 14 July 2026).
• One fee award in the whole record, and it runs the other way — it sits on C100, one of the four actions filed against him.
• Two vexatious-litigant requests, neither ever decided. In C046 a defence motion was filed on 31 March 2026; the docket was read through entry 158 and records no ruling. In C102 the court held in July 2026 that the designation was beyond the scope of the motion to quash then before it and that a separate motion would be required. That motion was never filed.

The last of those matters more than the others, because it forecloses the easy explanation. It is not the case that nobody asked, and it is not the case that nobody assembled the record across forums. In C066 — a Lackawanna action removed to the Middle District of Pennsylvania on 2 April 2026 — the defence exhibits put material from two California Superior Court matters against the same defendant onto a Pennsylvania federal docket. This project had tracked neither; it now holds them as C101 and C102, and they are the two actions in which the only sanctions in this record were entered, neither of them for anything to do with filing volume. Six days after the removal, that docket and its companion C063 both carried show-cause orders concerning improper venue and possible transfer. The one place in this corpus where the cross-forum record was gathered, it was gathered by a private party, and the institution's answer was about where the case belonged.

Two requests, no decisions. That is not a doctrine that considered restriction and declined it on the merits. It is a doctrine that was asked and did not answer.

III. Four propositions this project began with, and what its own exhibits did to them

The front matter of this book states a finding about motion-dependence. That is not the finding the project set out to demonstrate, and the distance between the two is the most useful thing a conclusion can record — it is the part a reader cannot reconstruct from the chapters, and the part a future researcher will otherwise repeat. Four of the propositions this project began with did not survive contact with its own data. Each was replaced by something narrower, and each replacement is better evidenced than the thing it replaced.

"Docket-local by construction" is false on the text. The proposal's thesis statement asserted that every restricting mechanism triggers on repetition visible to the court being asked. Chapter 2 read the four regimes against free primary sources and found otherwise. 28 U.S.C. § 1651 states no predicate at all, so every federal pre-filing standard is judge-made gloss. The Second Circuit's first Safir factor is "the litigant's history of litigation" — unqualified, naming no court. Pa. R.C.P. No. 233.1's predicate is "a prior action," also unqualified, and Coulter v. Ramsden applied it to a prior federal action. Cal. Code Civ. Proc. § 391(a) defines litigation as any proceeding "in any state or federal court," solving aggregation by definition. The replacement finding is sharper than the one it replaced: every regime is narrower on the remedy side than on the predicate side. California may count federal filings and § 391.7 restrains only "the courts of this state." Safir permits unlimited history and the approved injunction reached federal actions. Rule 233.1 will count a federal prior and bars only same-defendant, same-claim, pro se refilings. Seeing further than you can reach produces nothing.

"Widen the aperture and the threshold fires" is false here. Chapter 7 built an aperture lattice — same court, same state, all federal, all forums — and tested every counting rule against the replay. Every threshold fires early under every aperture that could have applied. Run § 391(b)(1) as written and it clears inside a single court between the seventh and ninth actions he commenced, on New York Supreme Court filings, no later than 11 March 2024 — nearly two years before the first judicial count. At the thirtieth action he commenced (C006, Lackawanna County, 23 April 2025) the priors visible were 5 in that court, 16 across Pennsylvania, 0 federal and 29 across all forums. The apertures are also a lattice, not a ladder: "all federal" is incomparable to "same state," and on this record the comparison is lopsided — 282 pairs where a federal aperture sees what a state one does not, against 3,487 the other way. Those two figures are this chapter's own re-run of chapter 7's exhibit, not a quotation of it: chapter 7 reports 188 and 3,463 as of 2026-09-03, and C105 moved both. Detection was never the binding constraint.

The escalation ladder is not a progression. Chapter 4 pre-registered four propositions about a filer moving outward from counterparties to counsel, platforms, regulators, government and finally adjudicators. One was refuted outright and two more failed to find any trend at all — a distinction that chapter now draws explicitly, because a non-significant rank correlation is a failure to find rather than a refutation. The one that was refuted, E1, depends on a definitional choice the chapter tests rather than assumes: C099 reaches the top rung only if a private arbitral institution counts as an adjudicator alongside a sitting judge, and under the alternative reading E1 holds. What does not depend on the choice is the observation itself. C099 — the sixth action he commenced, filed 11 January 2023 — names an opposing party, his own former attorney, the arbitrator and the arbitral institution in one complaint, spanning the bottom, middle and top of the supposed ladder two and a half years before he first sued a judge. What the record supports is that the ladder is additive rather than substitutive: rungs open and never close, 74 of the 91 actions still name a counterparty, and what the later period adds is volume rather than height. Volume is what chapter 1's two footnotes were counting.

The cost story ran the opposite way from the outline, and its one surviving proposition is the one that limits it. Chapter 6 expected a filer minimising his own cost. Where the record speaks — 20 of the 101 records he commenced carry any statement about how the filing was paid for — it says the fee was paid three times as often as it says it was waived, including seven actions filed on a single day, 29 July 2026, under one receipt for $2,835. That proposition is recorded as unsupported rather than refuted, because fee status entered this corpus where a court had reason to remark on it and payment is the remarkable case; what the record affirmatively establishes is a floor of $4,050 paid at intake, which is enough to dispose of "near zero" without supporting any rate. Meanwhile the proposition that cost imposition is a marginal public theme held: 17 of 1,667 statements, 1.02%, carry it. That number limits the chapter that reports it, and is stated before the quotations rather than after them.

A book whose exhibits refuted its own outline four times is not thereby a weaker book. It is the reason the remaining claims are worth anything.

IV. What survives

Strip away what the exhibits refuted and the residue is narrow, and it is not what the project expected to be arguing.

The failure in this record is not a perception failure. Judges of the Middle District of Pennsylvania noticed, said so in strong terms, and told each other: Carlson's "prodigious, but prodigiously unsuccessful, pro se litigant" characterisation of November 2025 was echoed in a companion report and expressly ratified by Munley in July 2026, after she reviewed the cases it rested on. Nine months of that, and the characterisation travelled while the count never left the district.

It is not an authority failure. Three of the four predicates chapter 2 read are drafted broadly enough to reach the whole record already: the Second Circuit's first Safir factor, which names no court; Pa. R.C.P. No. 233.1's "a prior action", held to include a federal one; and Cal. Code Civ. Proc. § 391(a), which says so in terms.

It is not an information failure. The dockets are public and this project found them with free tools. Beyond that, chapter 5 established that the filer narrated his own litigation continuously and in public across 1,667 timestamped statements from June 2022 onward. On this project's own classification of them, 445 describe filing conduct under a closed vocabulary and 47 describe repeat filing on the same dispute — figures chapter 5 §II reports as illustrative rather than as counts, because that classification has one coder and no reliability evidence.

The claim here does not need them, which is why it survives the caveat intact. What carries it is the part that rests on records rather than on readings: statements naming cases that sit in two or more different courts, and cases that sit on both sides of the state–federal line. Which cases a statement names and which court each sits in are recorded facts, and on those facts the cross-forum aggregation no institution performed was being performed out loud, continuously, for four years. The information was free, public, timestamped and in his own voice throughout.

And it is not a detectability failure, because chapter 7's replay fires every threshold years before anyone looked.

What is left is the thing none of those covers. Nothing in this record converts a visible pattern into a consequence across a sovereign boundary. Noticing and restricting run on separate tracks, and only the second requires a count. A characterisation travels between judges and requires no count. Preclusion requires a prior judgment between the same parties and disposes of one case — Judge Mehalchick's res judicata holding in C054 in February 2026 is the closest any court in this record came to treating one action as a fact about another, and it was preclusion. A filing restriction requires a pattern, a pattern requires a count, and no institution here was constituted to keep one.

Chapter 1 ended by posing the question this section has now answered, and the answer is not the intuitive one. The intuitive answer is that the aperture is too narrow — that a judge asked how many actions this plaintiff has filed simply cannot be shown more than his own docket holds. Chapter 2 disposed of that on the text and chapter 7 disposed of it on the record. The doctrine's problem is not the size of the aperture. It is that there is no one whose job is to look, and no rule saying what follows if they do.

V. The reform that follows, and the one that does not

Chapter 7 tested four designs against the record and reached a narrower conclusion than it set out to reach: an aggregator is necessary and demonstrably insufficient. Necessary, because Ringgold-Lockhart requires that an adequate record for review list all the cases and motions, and nothing in this system produces that listing. Insufficient, because on this record the listing was already producible — a New York clerk could have counted to five in 2023 — and no consequence attached to its producibility.

So the load-bearing reform is not the aperture. It is the trigger and the consequence: who is obliged to look, and what follows when they do. § 1915(g) is the proof that Congress can write a cross-forum counting rule when it wants one — it counts dismissals in any "court of the United States" — and the proof that the omission everywhere else is a choice rather than an impossibility. Its reach is prisoners proceeding in forma pauperis, in federal court, and its consequence is the loss of fee waiver. It is small, it is thirty years old, and it is the only one.

One counterexample belongs here before those recommendations, because it is the only thing in this record that behaves the way the reform above asks an instrument to behave, and it was not designed for the purpose at all.

Three appeals in this corpus were dismissed for an unpaid fee. The Third Circuit dismissed No. 24-3329 "for failure to timely prosecute insofar as appellant failed to pay the requisite fee as directed" — $605 — six months after the district court had certified that the appeal was not taken in good faith, and nothing else disposed of it. The Superior Court of Pennsylvania dismissed Nos. 711 and 723 MDA 2026 for an unpaid $91.25 apiece, in each case having also quashed the appeal sua sponte on jurisdictional grounds, so there the fee is an independent and sufficient ground rather than the only one. Chapter 6 §II separates the three.

Set that beside everything else in this chapter. Across 93 actions, four restriction regimes, 72 court findings and two requests for a restriction, the one price that has repeatedly stopped a proceeding is a ninety-one-dollar appellate filing fee.

Two properties make it the proof of concept rather than a curiosity. It required no motion. No defendant assembled anything, asked for anything, or bore any cost; a clerk applied a schedule and the court applied its own rule — and in the two Pennsylvania appeals the court reached the jurisdictional defect sua sponte as well, which is the same property showing up twice in one order. That is precisely the trigger the rest of this chapter says the doctrine lacks. And it carried a consequence that attached automatically to a condition, rather than to a judgment about the filer. Nobody had to decide anything about his litigation history for it to fire.

It is also, deliberately, a small point. An appellate filing fee is not a vexatious-litigant remedy, prices nothing about the pattern, and would stop a meritorious appeal by an impecunious litigant exactly as readily — which is why § 1915 exists and why chapter 6 declines to recommend making entry more expensive. What it demonstrates is narrower and more useful: a rule that fires on a condition rather than on someone's initiative does operate in this system, at scale, today. The reform this book points at is not a new kind of instrument. It is that kind of instrument, pointed at the pattern instead of at the fee.

Two cautions the exhibits impose on any such design, both of which cut against the argument and both of which belong here rather than in a footnote. A pure velocity flag set at three actions in ninety days fires on this corpus's first day, 9 March 2018 — a false positive in our own data, generated by three same-day filings. And chapter 5's caution runs the other way: an aggregator counting only dockets would have run a week behind filings this record announced in advance, while one counting announcements would have counted a suit in every state he said he would sue in, most of which were never filed. Neither the docket nor the narration is sufficient alone, and a threshold cheap enough to be automatic is cheap enough to be wrong.

VI. A sixth forum, five days ago

The strongest argument against a thesis about structural invisibility is that the structure might have changed while the book was being written. On this record it did not.

Dispute group DSP006 concerns a failed 2025 wagyu shipment. It began as C015 in Lackawanna County on 22 July 2025. A defendant removed it to the Middle District of Pennsylvania on 21 August 2025, where it became C057 — and C057 is the case Judge Munley dismissed on 13 July 2026, in the order whose footnote produced the second of this book's two judicial counts. An appeal from that dismissal was docketed at the Third Circuit as No. 26-2838. Then, on 1 September 2026, an action naming the same canonical defendant on the same dispute was filed in the Eastern District of New York as C105, a court that has never seen any part of it, while the appeal was pending.

Three trial forums and an appellate court on one dispute. The judge who wrote the largest number any court has put on this record dismissed the case fifty days before it reappeared in a district that had never seen any part of it, and nothing in the intake of that district connects the two.

Two things must be said about C105 rather than left to implication. It is coded RefilingOfPriorCase = Unclear, not Yes: no document on the docket is obtainable through RECAP, the Internet Archive or GovInfo, so the complaint could not be read, and the protocol requires Unclear over a guess when the third condition — substantially the same claims — cannot be established. The indicator recorded against it is Moderate for the same reason. And this is one action, not a demonstration. The demonstration is the other 94.

VII. What this record cannot settle

Stated plainly, because these bound the claims above rather than decorating them.

1. Intent is not in the corpus. Nothing establishes that actions were distributed across forums in order to avoid aggregation. Ten of the cross-forum pairings exist because a defendant removed the case, which is the defendant's choice. The filer narrates his own litigation constantly and in public, which is close to the opposite of concealment. Distribution is a property of the record; motive is not in it.
2. This is one filer. The mechanism generalises only as far as the argument about the regimes does, and the regimes were read as texts rather than sampled across filers.
3. ProSe is blank on 45 of 105 records — 54 Yes against 60 recorded values. It blocks no chapter, but it means no quantitative claim is made about how many of these actions were filed without counsel, and Pa. R.C.P. No. 233.1 reaches only pro se plaintiffs.
4. The appellate layer became first-class hours after this chapter was drafted, and was verified four days later. thesis/exhibits/appellate_layer.py records 30 appellate dockets arising from 22 trial-court records, none of which crosses a sovereign boundary: the second de-duplication axis multiplies records without widening reach, which is this thesis's own finding one level up. 19 of the 30 are now checked against the reviewing court's own docket, up from 7; 8 carry a number whose status is unconfirmed and 3 an appeal whose number was never recovered, so posture counts drawn from that table remain a floor rather than a census. The asymmetry the pass exposed belongs in this list too: Pennsylvania publishes a free appeal docket sheet carrying the disposition text itself, while the federal circuits publish no free docket at all — every federal disposition here had to be read from the district docket below, and eight of the eleven rows still unresolved are federal on that account.
5. No coded variable in this project has an inter-rater check. The repeat-filing code had a rule-versus-judgment audit at 80.4% agreement across 18 adjudicated disagreements standing in for one; the statement classification, now by a wide margin the largest coded variable here, has nothing equivalent. As of 2026-09-11 the consequence is carried in the text rather than only admitted in this list: chapter 5 §II separates figures that rest on a record join, which are counts, from figures that rest on the classification, which are reported as illustrative — and this chapter's arguments are built on the first kind wherever the record allows. Closing the gap needs a deterministic screen of the sort thesis/protocols/01-refiling.md describes, written up as a third protocol and run against a stratified sample.
6. No defence cost is recorded anywhere. Not one invoice, fee petition or cost declaration for any of the 91 actions he commenced. Every dollar figure on the burden side is a party's estimate.
7. Ohio and Texas are unverified. Both appear to maintain published vexatious-litigant registers, which would refine chapter 2's hedge that California's is the only one this project has verified, and would change chapter 7's third remedy. Their statute hosts refuse automated connections, so no claim is made about either.
8. The corpus is still moving. It gained a docket record and a sixth forum on the day this chapter was written, after two other chapters had already been drafted against the older totals. Every number above carries its as-of date for that reason.

None of these makes the null result contingent. Zero pre-filing orders and zero monetary sanctions across 91 actions is not a figure that a better ProSe column or a first-class appellate table would revise.

VIII. The finding

A litigant filed 91 actions across six courts in three states over eight years. Twenty-six judges wrote 66 findings about them, in language that includes frivolousness, jurisdictional manipulation, futility of amendment and claims identical to claims already dismissed. Four restriction regimes were available. Two parties asked for one. Every counting threshold that any of those regimes states was satisfied years before any judge put a number on the record, under every aperture that could have applied.

And the highest number any court has stated is 38, of 95.

That is not a finding about this litigant, and this project has been careful throughout not to make one — no court has determined that he is a vexatious litigant, and the whole point of the argument would collapse if a research classification were allowed to stand in for the determination whose absence is the subject. It is a finding about the instrument. American vexatious-litigant doctrine can count, in three of its four regimes, further than anyone has ever asked it to count. What it cannot do is oblige anyone to ask.