The narrated docket

This chapter cites law as well as corpus rows, so it carries two closing tables. Sources is the corpus; Legal authorities is everything that is not. Keeping them apart is how a reader checks the project's no-fact-originates-here rule by inspection rather than by trust.

Re-baselined 2026-09-11, and this is the project's one exception to a standing rule. Every other chapter here is revised for defects and never re-baselined, because a dated figure is correct as of its date and recomputing it would replace what was known with what is known now. This chapter was drafted on 2026-08-31 against 817 statements from 333 videos. The corpus now holds 1,691 from 620 — it doubled in eleven days, because the archive that feeds it was being built at the same time as the chapter. Two external reviews made the same point, and it is right: a distribution measured on half its data is not a dated figure, it is a chapter built on a superseded corpus, and no as-of date rescues a claim that two-thirds of the statements carry no filing-conduct factor if the real figure might be different. So sections I to IV are recomputed here and the earlier figures are retired rather than preserved.

Two things are worth recording about what the re-run did. The argument got stronger, not weaker: cross-forum narration went from 147 statements to 395, and the announcing statements that precede a filing they name went from 32 to 76. And every dated narrative claim survived verbatim — each sequence in section IV was re-checked against the database and each still holds to the day. The corpus ledger in the front matter records both the old and the current figures.

I. A corpus that should not exist

Vexatious-litigant doctrine has an inference problem it almost never names. Several of its tests turn on why a person is filing. Safir v. U.S. Lines makes the second of its five factors "the litigant's motive in pursuing the litigation, e.g., does the litigant have an objective good faith expectation of prevailing?" New York's frivolous-conduct rule, 22 NYCRR 130-1.1(c)(2), reaches conduct "undertaken primarily to delay or prolong the resolution of the litigation, or to harass or maliciously injure another." California's § 391(b)(3) reaches tactics "solely intended to cause unnecessary delay." Each of those is a question about a state of mind.

Courts answer it, when they answer it at all, by inference from the docket — from the number of filings, their repetitiveness, the disposition of the earlier ones. They do this because the alternative is unavailable. A litigant's purpose is not ordinarily observable, and a litigant asked about it in the proceeding has every reason to describe it favourably.

In this record it is observable. As of 2026-09-11 the corpus holds 1,691 timestamped statements drawn from 620 videos on the subject's own YouTube channel, published between 2022-06-01 and 2026-09-09. Every one is transcribed verbatim from the video's own caption track and carries a start and end timestamp, so a reader can open the clip and check it. Three numbers describe the video population and they should be kept apart: 625 videos are catalogued, 624 of them are in scope — one is coded Not litigation-related and is excluded from this project entirely, along with its statement rows — and 620 have produced at least one statement. The gap between 624 and 620 is not an exclusion; it is four in-scope videos nobody has yet found anything quotable in.

What follows is a chapter about that corpus, and the first thing to say about it is what it mostly is not.

Of the 1,691 statements, 1,234 — nearly three-quarters — carry no filing-conduct factor at all. They are a man describing his disputes: a surgery that went wrong, a shipment that was destroyed, a tax assessment, a copied protein bar. The single largest classified factor in the corpus is Judicial or forum bias alleged, at 157 statements, which is an allegation about courts rather than a description of filing. The corpus is not a confession. It is a very long public diary that happens, in about a quarter of its entries, to describe litigation conduct.

That proportion is worth pausing on, because it moved the wrong way for this chapter when the corpus doubled and it is reported anyway. At 817 statements the no-factor share was two-thirds; at 1,691 it is 73%. The larger the archive gets, the smaller the fraction of it that is about filing at all.

The 457 statements that do carry a filing-conduct factor are the subject of this chapter. It is worth being exact about what a classification on one of them means. pattern_relevance — Strong, Moderate or Context — and pattern_factor are this project's research classifications, assigned under a rubric published on the site. They are not judicial findings, they are not adjudications of anything said in a video, and they are not a determination that anyone is a vexatious litigant. No court in this record has made that determination, which is the whole point of the thesis.

II. The taxonomy, and why it is closed

Each statement carries a pattern_factor naming which filing-restriction factor the speaker's own words describe. The vocabulary is closed at eight values plus one sentinel, and each value corresponds to a criterion recorded in the project's vexatious_indicators data — or, in three cases, says openly that it corresponds to none.

A statement may carry more than one factor; 457 carry at least one, and 72 carry one of the three core factors. Those three are core because they are exactly the conduct named in the published Strong rubric, so only they can qualify a statement as Strong. Sixty-five statements are classified Strong, 1,271 Moderate, 355 Context.

Two design decisions are worth stating because they cut against the classification's reach.

First, the vocabulary hard-fails. An unrecognised factor value, or the No filing-conduct factor sentinel combined with a real factor, aborts the site build outright rather than being silently dropped. narrated_docket.py reimplements the same check and exits non-zero on the same conditions, and additionally on a statement citing a case id the corpus does not contain. A silently-dropped classification is the failure mode the column exists to remove, so it is made loud in both places.

Second, a direct appeal is deliberately not Filing after an adverse decision. That factor means a new or repeat action; the ordinary appellate remedy is not evidence of anything except that a party lost and used the procedure available to them. VS657 is the clean illustration of the line: the speaker says "we're going to file a new lawsuit in the federal court. And we also appealed the state lawsuit." The new federal action is coded; the appeal is not.

This is the same discipline chapter 3 described for RefilingOfPriorCase, and it carries the same limitation: there was no second coder. The tiers and factors here were assigned by a single coder against a published rubric, without an inter-rater check. For the refiling variable a rule-versus-judgment audit ran in place of one, at 80.4% agreement; no equivalent audit has been run on the statement classification, which at 1,691 rows is now by a wide margin the largest coded variable in this project.

Two external reviews in September 2026 made the same point about that admission: stating a limit is not the same as respecting it, and a chapter that concedes it has no reliability evidence and then reports its classification as counts has not really conceded anything. They are right, so this chapter draws a line through its own figures.

Two kinds of number appear below, and only one of them is a count.

A figure resting on a record join is a count. Which cases a statement names, what court those cases sit in, and when a video was published against when an action was filed are recorded facts — RelatedCaseIDs, cases.Court, cases.Filed — and a second coder would return the same answer from the same rows. Section III's cross-forum narration figures are all of this kind.

A figure resting on the classification is illustrative rather than a count. How many statements carry a given pattern_factor, how many are tiered Strong: these describe how one coder read 1,691 passages against a published rubric, and nothing here establishes that a second reader would have drawn the lines in the same places. The rubric is published and the vocabulary hard-fails, so the classification is reproducible — and chapter 3 §VII sets out at length why reproducible is not the same as valid.

The factor table below, the 457, the 72 core factors and the tier split are all of the second kind and are offered as illustration. So, necessarily, is section IV's restriction to "announcing" statements, since that restriction is made by factor. Section III's figures are of the first kind. Where this chapter's argument carries weight, it is built on the first kind wherever the record allows — which is why the cross-forum narration finding, and not the factor distribution, is what chapters 7 and 8 take from here.

III. He narrates the distribution

Chapter 1 described two judges, five months apart, each counting the actions before them. Magistrate Judge Latella counted twenty-two in the Middle District of Pennsylvania on 2026-02-03, when sixty-five originating actions existed across all forums (C055). Judge Munley counted thirty-eight on 2026-07-13, when eighty-five existed (C057). Both counts were exactly right and both were forum-bounded, and nothing in the system was positioned to cross that boundary.

Something was crossing it the whole time.

Three hundred and ninety-five statements name cases sitting in two or more different courts. Three hundred and eighty-four of those span the state–federal line. One hundred and thirty-six name cases in two or more different states. The aggregation no institution performed was being performed out loud, on a public channel, continuously, for four years.

Those three figures are the ones the re-baselining moved most — they were 147, 141 and 38 when this chapter was first drafted against half the archive. The direction is worth noting because it is the opposite of what a sceptic would predict. If cross-forum narration were an artifact of a few unusually discursive videos, doubling the corpus would dilute it. It concentrated instead.

Sometimes it is a bare self-report. On 2025-10-23, in a short about a platform ban, the speaker says (VS725, linked to C014, C043 and C053):

"You know, I filed multiple lawsuits in different courts. These judges don't care. They're all bought and paid for by all of these elite freaks that don't care about you."

Sometimes it is an itemised account of a single dispute distributed across two court systems. On 2025-08-16 (VS076):

"There's the main lawsuit against him — there's one in state court, one in federal court. The one in federal court actually got dismissed, I had to appeal it because I was incapacitated... The second lawsuit I filed against him was with the TikTok stuff. And then I filed a third lawsuit against him because of the Google reviews and the [Real Self] reviews."

And sometimes it is a complete narrated procedural history of an escalation, delivered in forty seconds. On 2026-08-06, about the county tax dispute (VS250, linked to C017 and C067):

"When I sued them in the state court, the judge wiped his tushy with it like my uh legal complaint was toilet paper and then I filed an appeal. They threw it out again. I went to the federal court, same thing. The federal court said, “Oh, well this is the state court's jurisdiction.”"

That is, in the speaker's own compressed idiom, an accurate description of a state action (C017, Lackawanna County, filed 2025-07-25), an appeal, and a subsequent federal action (C067, M.D. Pa., filed 2026-04-06). It is precisely the cross-forum sequence the doctrine would need to see and has no instrument for assembling.

Eighty of the 107 docket records — 73 of the 97 originating actions — are named in at least one public statement. Not summarised, not alluded to: named, with a case or dispute identifiable well enough for a researcher to tie the statement to a docket row. Roughly three-quarters of this litigation record has a public, timestamped, first-person commentary track attached to it — up from five-eighths when this chapter was first drafted.

There is a complication here that cuts against reading the narrator as a reliable aggregator, and it should be conceded before it is found. He undercounts too. On 2025-07-17 he referred to "all 20 of my other lawsuits" (VS057). On that date the corpus records forty-three originating actions across all forums, and sixteen in the Middle District of Pennsylvania. His own figure sits between the forum-local count and the aggregate, closer to the former. The published record is far more complete than any docket, but it is not a register, and nobody — including the person keeping it — was holding the total.

IV. Announcement precedes filing

The stronger claim available from this corpus is temporal, and it needs to be made carefully because the obvious version of it is partly circular.

Of the 1,691 statements, 847 name at least one tracked case, and all 847 have both a video publication date and a filing date to compare. Of those, 423 were published on or before a filing they name — 175 within thirty days of it, 108 within seven, 17 on the same day.

That 423 is an outer bound and should be treated as one. Some case links in the corpus were established partly by subject-matter and date correspondence rather than by anything in the words. VS622 is the honest example: a 2025-12-19 short describing a car accident, linked to C025 (filed the same day) because the facts described match three pleaded elements — but nothing in the speaker's words identifies a filing at all. Counting that as "published before a filing" measures the linking method, not the speaker.

So restrict to statements whose own words announce or threaten a filing — those carrying Filing announced or confirmed or Threatened or prospective filing. That leaves 76 statements published on or before a filing they name: 49 within thirty days, 39 within seven, 8 on the same day.

Those 76 are the defensible core of the claim in one respect and its weakest point in another, and both belong in the same breath. The dates are records. The restriction is a classification, made by one coder, with no reliability evidence behind it — so the 76 is an illustrative figure in the sense §II sets out, and a second coder drawing the announcing line differently would return a different number. What does not depend on the classification at all is the handful of individual sequences below: each was read from the statement's own words against the docket, and each can be checked by anyone with the video and the case number. They are the reason this section matters, and they are why the argument rests on them rather than on the 76.

Two filings announced at once, in different court systems, both landing. On 2025-06-06 (VS042):

"I'm filing is against the law firm, Tabon, and Tik Tok for deplatforming me for no legitimate reason ... And then I'm filing a completely new separate lawsuit against Tik Tok, the law firm, as well as Tibon for trying to destroy my business. Tortious interference."

C014, Tufano v. TikTok, was filed in Lackawanna County that same day. C053, Tufano v. TikTok et al, was filed in the Middle District of Pennsylvania twenty-one days later. The two are recorded as the same originating action. A viewer of that video on the day it went up had, in one sentence, the fact of a second suit against the same defendants that no docket would show for three weeks.

A defendant added between announcement and filing. On 2026-03-06 (VS649): "I'm glad I didn't file my lawsuit yet because I guess this is another defendant." C036 was filed in Lackawanna County three days later. Ten days after that, on 2026-03-16 (VS490): "I filed my lawsuit last week and now I got to file a whole new separate one because I miss these guys." C036 had indeed been filed seven days earlier; the companion federal action C068 followed on 2026-04-13, twenty-eight days after the statement. The speaker's own "last week" checks out against the docket, which is a small but real validation of the corpus's date handling.

A suit against a Commonwealth and a county, announced five days out. On 2026-03-18 (VS118):

"and they just don't allow prosay [pro se] litigants to participate in the court system. So I either have to appeal or file a lawsuit against the state of Pennsylvania and the county of Lacwan [Lackawanna] explaining hey I have followed the civil procedure they are refusing to acknowledge me as a prosay litigant and acknowledge my motions for default."

C038, Tufano v. The Commonwealth of Pennsylvania, was filed on 2026-03-23 — five days later, naming both intended defendants.

And the same-day case. On 2026-04-06 (VS657): "but we're going to file a new lawsuit in the federal court." C067 was filed in the Middle District of Pennsylvania that day.

Three limits on all of this, stated plainly.

A publication date is not a statement date. The date recorded is the video's, and the speaker may have recorded it earlier. For same-day and one-day intervals that uncertainty is the whole margin.

Adjacency is chronology, not intent. A statement that precedes a filing establishes that the filing was contemplated publicly beforehand. It establishes nothing about why the filing was made, and this project's video_filing_chronology rows say so on their face: the assessments there are recorded as "chronology/context, not a finding."

Not every announcement becomes a filing. A hundred and twenty-seven statements threaten or contemplate a filing; far fewer are followed by one. The most vivid threat in the corpus is a counter-example. On 2026-06-22, about the freight dispute already in suit in two forums (VS697):

"And then what? I'll file the lawsuit in I'll file the lawsuit in New York tomorrow before you even dismiss it. Like, what kind of BS is that? Like, just jerking me around trying to waste my time. don't want to pay me. If they want me to file I'll file a lawsuit in New Jersey, too. I'll file a lawsuit in every state I can drive to"

No New York action on that dispute appears in the corpus, and no New Jersey action appears anywhere in it. The plainest available statement of a strategy of forum multiplication is also, on this record, a statement of something that did not happen. Treating stated intention as accomplished conduct is exactly the error the factor vocabulary separates Threatened or prospective filing from Filing announced or confirmed to avoid.

V. The evidentiary question

Which brings the chapter to the question the outline promised and no court in this record has had to answer: may a court considering a filing restriction take account of a party's own published statements about why he files?

Three bodies of law meet here, and they do not point the same way.

Evidence law offers no obstacle. A party's own out-of-court statement offered against him is not hearsay at all: Fed. R. Evid. 801(d)(2)(A) excludes from the hearsay definition a statement "offered against an opposing party" that "was made by the party in an individual or representative capacity." A YouTube video published by a plaintiff, offered against that plaintiff, is the paradigm case. Authentication and relevance would be contested; the hearsay rule would not be the fight.

The restriction standards positively invite the inquiry. Safir's second factor asks about motive in terms. Section 391(b)(3) asks whether tactics were "solely intended to cause unnecessary delay." Rule 130-1.1(c)(2) asks whether conduct was "undertaken primarily to delay or prolong the resolution of the litigation, or to harass or maliciously injure another." These are subjective-intent tests written into the operative text, and a court applying them ordinarily has nothing to work with but inference from the docket. A published statement of purpose is directly responsive evidence.

The corpus contains statements that answer those questions in terms. VS263, on 2026-08-11: "we'll just file another lawsuit stating the simple facts of the conspiracy just to continue to expose how crooked this system is. And as I've stated before, I never expected to win these lawsuits." Safir asks whether the litigant has "an objective good faith expectation of prevailing." That is a published answer to it. VS737, on 2026-08-01, addresses the cost question the same way: "whether I have to file this lawsuit every year for the rest of my life, state and federal court... I'll bury them in legal fees." Seventeen statements in the corpus are classified as stating cost imposition as a motive — 1.0% of it, a share chapter 6 states before it quotes any of them.

The First Amendment complicates it, and the complication is structural rather than absolute. Litigation is petitioning: "the right of access to the courts is indeed but one aspect of the right of petition" (California Motor Transport). Petitioning is immune unless it is a sham, and Professional Real Estate Investors fixed the order of that inquiry: "First, the lawsuit must be objectively baseless in the sense that no reasonable litigant could realistically expect success on the merits," and "[o]nly if challenged litigation is objectively meritless may a court examine the litigant's subjective motivation." If an objective litigant could think the suit reasonably calculated to elicit a favourable outcome, "the suit is immunized under Noerr." Bill Johnson's Restaurants had already paired the two elements in the labour context — a baseless suit prosecuted with retaliatory intent — and BE&K Construction later held that a standard reaching reasonably based but unsuccessful suits on motive alone was invalid, reasoning that "the genuineness of a grievance does not turn on whether it succeeds."

The structural consequence for a statement corpus is precise, and it is a limit rather than a prohibition: published statements of purpose are second-step evidence. Under PRE's ordering they are reached only after an objective assessment of the filings themselves, and they cannot substitute for that assessment. A court could not restrict a filer because he said he wanted to impose costs; it could consider that he said so, once the filings had independently been found objectively meritless.

Two caveats keep that from being a tidy conclusion. PRE is an antitrust-immunity case, and the sham-litigation framework it orders is not formally the same inquiry as a court's supervisory power over its own docket under § 1651 or § 391 — chapter 2 traced how thin the textual footing of the federal standard is. And California Motor Transport already contemplated the aggregate: "a pattern of baseless, repetitive claims may emerge which leads the factfinder to conclude that the administrative and judicial processes have been abused." That sentence is about pattern, which is what this record has and no court has seen.

One court in this record has already used the videos. On 2026-08-07, in C046 (filed in the Middle District of Pennsylvania as 3:25-cv-00011, now before the Central District of California as 2:25-cv-00147), Judge Stephen V. Wilson excluded expert testimony under Rule 37 and made bad-faith and credibility findings. Among the grounds recorded is "public video activity inconsistent with that claimed incapacity." The channel is already in evidence somewhere in this litigation — used, entirely conventionally, to test a party's account of his own medical condition on a discovery motion.

What no court has done is use it on the aggregation question. That is not because the question was answered against the proposition. It is because the question was never reached.

VI. What it would take, and what this is not

Two vexatious-litigant requests exist in this record and neither has been adjudicated. In C046 a defence motion for a vexatious-litigant declaration or Rule 11 relief was filed on 2026-03-31; the docket was read through entry 158 as of 2026-08-28 and records no ruling. In C102 a party argued in July 2026 that the plaintiff had become a vexatious litigant; the court held the designation beyond the scope of the motion to quash then before it and said a separate motion would be required. No separate motion was ever filed.

So the honest close is a negative one. None of the 1,691 statements has been received in evidence on the question this chapter is about, because nothing has put that question before a court. The corpus exists; the doctrine has no procedural slot that reaches for it.

That gap is not primarily evidentiary. If a court were to reach the aggregation question, this material would be available to it on ordinary principles, subject to the ordering PRE imposes. The reason it has not been reached is the reason chapters 1 and 2 gave: no motion, no instrument, no artifact in which the record across forums is assembled. A published statement corpus does not fix that. It does something narrower and, for the argument, more useful — it shows that the invisibility in this record is not an information problem. The information was free, public, timestamped, and in the filer's own voice for more than four years. What was missing was an institution positioned to look.

Chapter 7 takes up what such an institution would have to be. This chapter's contribution to that design question is a caution: an aggregator that counted only dockets would have run a week behind the filings this record announced in advance, and an aggregator that counted announcements would have counted a suit in every state he could drive to, which was never filed.