The cost side

This chapter cites law as well as corpus rows. The two are kept in separate tables at the end — Sources for the record, Legal authorities for the statutes, rules and opinions — so that the rule this folder runs on, that no fact about the litigation originates in it, stays checkable by inspection.

Revision, 2026-09-11. Two external reviews found four defects here, and all four are repaired. P1 was reported as refuted on a selected sample: the chapter said in one paragraph that its fee table is "a floor, not a rate" and in the next that the table refutes P1. It does not — fee status is recorded where a court had reason to remark on it, and payment is the remarkable case — so P1 is now unsupported, with the narrower claims that are established set out beside it. The exposure bias now has a direction: the 45 excluded records skew long, so the 149-day median is a floor. A units slip in §VI said defendants are named in exactly one action "in 38.8% of cases"; 38.8% is a share of entities. And section II's "last two rows" pointed at a pair the table as printed does not make identifiable, so the rows are now named. Separately, this chapter carries material added on 2026-09-10 — the appellate fees — under a 2026-09-06 as-of date; the body figures remain as of 2026-09-06 and the appellate rows are dated where they appear.

Revision, later on 2026-09-06. Eight duplicate party records were merged that evening — one person recorded twice, surname-first on a state docket and first-name-first on the federal docket it was removed to — so every entity figure in section III moves: 335 canonical entities become 327, 680 party-slots become 673, the mean parties per action falls from 7.23 to 7.08 and the median from 2.5 to 2, and the one-shot share falls from 40.3% to 38.8%. None of it disturbs the argument; section III's point is that burden accrues per name while the fee is charged per filing, and a slightly smaller denominator sharpens rather than softens it. Two stale counts are also corrected: P3 was pre-registered over "the ninety actions he commenced" and section IV repeated it, while five other passages already said 91 — 91 is right, and C105 is why. Section VI's fee-paid count is brought into line with the table in section II, which says 12.

Revision, later on 2026-09-06. C105 entered the corpus after this chapter was first drafted and immediately tripped the exhibit's fee-closure gate, which is what that gate is for. Two things changed as a result. C105 is recorded fee-paid, so §II's table gains a row. And the dollar arithmetic was narrowed to the records that state the amount: it had been multiplying $405 by every fee-paid record, including C073, which says only that he "paid the full filing fee" and names no sum. A fourteenth gate now checks that base in both directions. No claim in this chapter turned on the affected figure — §II quotes the sourced $2,835, not the aggregate — but the aggregate was wrong and is now $4,050 over the ten records that name the amount.

I. Four propositions, before the numbers

This is the third of four demonstrations that nothing converts a visible pattern into a consequence. Chapter 4 took the shape of the litigation and chapter 5 took its narration; this one takes the money, which is the currency in which a consequence would actually have to be denominated if one ever landed.

Chapter 4 stated its claims as falsifiable propositions before reading its exhibit, and reported on each. That habit is worth keeping, so here are this chapter's, written down first.

P1. The filer's marginal cost per action is near zero, because he proceeds in forma pauperis.

P2. What a filing costs him scales, at least loosely, with what it costs the people it names.

P3. Somewhere across the ninety-one actions he commenced, the system recovered some of that cost from him — a monetary sanction, a fee award, or both.

P4. Cost imposition is a marginal theme in what he says publicly: incidental to the disputes rather than characteristic of them.

P2 and P3 are refuted. P1 is not refuted — it is unsupported, and on a sample selected by the very mechanism that recorded it; §II states why the difference matters. P4 survives, and it is the one that limits this chapter rather than supporting it, so it is reported plainly in §V and not argued away.

II. What a filing costs the filer

28 U.S.C. § 1914(a) prices a federal civil case at "a filing fee of $350," payable by "the parties instituting any civil action, suit or proceeding in such court, whether by original process, removal or otherwise," with § 1914(b) authorising such "additional fees only as are prescribed by the Judicial Conference of the United States." The corpus records the sum actually charged in this district as $405, and it records it the same way each time.

P1 is unsupported. Of the 101 docket records this plaintiff commenced, 20 carry any recorded statement about how the filing was paid for — 19 originating actions, once the union-find collapses one state-and-removal pair. Where the record speaks, it says the fee was paid roughly three times as often as it says the fee was waived:

Two of those twelve — C073 and C105 — record that the fee was paid but name no sum, so they are outside the dollar arithmetic even though they are inside the count. Ten records name $405, and $4,050 is the whole of what this corpus can be made to say about money paid at intake.

The IFP sought on appeal and fee ordered, not paid rows were added on 2026-09-10, when the appellate verification pass read dispositions that turned on money rather than on merits. They are appellate fees, which is a different price from the one this section has been pricing: $91.25 in the Superior Court of Pennsylvania, $605 in a federal circuit. Three appeals in this corpus were dismissed for an unpaid fee, and the three are not alike, so they are separated here.

In one the fee was the whole of it. The Third Circuit dismissed No. 24-3329 "for failure to timely prosecute insofar as appellant failed to pay the requisite fee as directed," six months after the district court had certified that the appeal was not taken in good faith and denied leave to proceed in forma pauperis — the C042 row above. Nothing else disposed of it.

In the other two the fee was an independent and sufficient ground rather than the only one. The Superior Court quashed Nos. 711 and 723 MDA 2026 sua sponte on jurisdictional grounds — in the first because the order appealed from was not final, claims against one defendant never having been adjudicated; in the second because an order denying recusal and a stay is interlocutory, and because appellate jurisdiction over tort claims against a local agency lies exclusively in the Commonwealth Court under 42 Pa.C.S.A. § 762(a)(7) — and separately dismissed each for the unpaid $91.25, recording in each case that the $202.91 Lackawanna County card transaction he produced "does not correlate with the filing fee for this appeal."

That is worth stating precisely, because it cuts against the shape of this chapter's argument rather than with it. Everywhere else here, the fee is the one cost the filer does pay and the defendant's costs are the ones nobody records. On appeal the fee bites the filer instead — once on its own, and twice more as a ground that would have sufficed by itself. It is the only price anywhere in this record that has repeatedly stopped a proceeding, and it stopped them without anyone having to ask.

That table is a floor, not a rate, and the reason matters. Fee status is not a column anyone set out to populate; it entered the corpus incidentally, because a report and recommendation happened to mention it in a footnote or a docket entry happened to name a receipt. Eighty-one records say nothing either way. The honest statement is therefore narrow: among the actions whose fee status the record establishes, paying the fee is the common case and waiver is the exception — not that he pays 12 times for every 4 waivers across the corpus.

Why that is "unsupported" and not "refuted"

An earlier draft of this chapter concluded that the table above refutes P1, and it does not. The reason is a selection problem sitting inside the recording mechanism, and it is worth setting out because it recurs wherever a corpus is built from whatever the record happens to mention.

Fee status is selected toward fee-paid by the thing that records it. Nobody set out to populate this column; a value exists only where a court or a docket entry had some reason to say so. And the reason is usually that the fact was remarkable. The clearest case is C073: the report and recommendation notes that he "paid the full filing fee rather than proceeding in forma pauperis" precisely because that fact carries a procedural consequence — no § 1915(e)(2) screening track attaches to a fee-paid complaint. An IFP filing, by contrast, generates its own paperwork and then stops being notable. So the mechanism that put twelve fee-paid records into this corpus is a mechanism with a reason to notice fee payment, and the 81 silent records are not a random sample of the rest.

What can be said, then, is narrower than a refutation and still fatal to P1 as written:

• Where the record speaks, it says paid, roughly three times as often as it says waived.
• At least $4,050 was paid at intake, across the ten records that name the amount, and $2,835 of it in a single day. That figure is a floor with a document behind every dollar.
• Only four records he commenced sit on an IFP track at all, so the premise that the marginal cost is near zero because he proceeds in forma pauperis has no support anywhere in this corpus.

A marginal cost of "near zero" is falsified by $4,050 in receipts. What is not established is the rate — that he pays three times for every waiver — and this chapter makes no such claim. The distinction is the difference between a proposition that died and a proposition nobody could test, and the second is what happened here.

The clearest instance of what the record does establish is the seven actions of 2026-07-29. C077 through C083 were filed on a single day, at 3:26-cv-02132 through 3:26-cv-02138, and the record states they "were fee-paid at $405 under a single shared receipt number 333120943, not filed in forma pauperis." One receipt. Seven actions. $2,835 paid in a day. They were assigned across five different magistrate judges with no consolidation and no common screening order — the intake system treated them as seven unrelated cases, which is chapter 1's finding arriving again at the cashier's window.

Two further recorded facts cut the same way. The report and recommendation in C060 records in its own footnote 1 that when he filed C073 — "raising the identical claims brought in" C060 — he "paid the full filing fee rather than proceeding in forma pauperis," with the consequence that "no 28 U.S.C. 1915(e)(2) screening track applies to it." And in C012 he has repeatedly asked the Lackawanna court either to rule on his pending motions or refund his filing fees so he can pursue the claims elsewhere. A litigant asking for his fees back is a litigant to whom the fee is a real number.

The ability to pay is itself docket-local

The single most telling row in this section is a pair. C011 and C052 are the same originating action. C011 was filed in Lackawanna County, where in forma pauperis status was granted on 2025-05-27 by Judge Nealon. Fenix Internet LLC removed it on 2025-06-26, and in the Middle District of Pennsylvania in forma pauperis status was denied on 2025-08-27.

Neither court was wrong. They were applying different instruments — Pa. R.C.P. No. 240, which routes fee waiver through Pa.R.J.A. 1990, and 28 U.S.C. § 1915(a)(1), which asks for an affidavit that "the person is unable to pay such fees." But the two determinations are about the same person, the same action and the same three months, and they came out opposite ways because each court asked the question inside its own docket. The thesis's central claim is usually stated about counting filings. Here it applies to the poverty determination itself.

III. What one payment obliges

P2 is refuted, and not marginally. § 1914(a) prices the case. The obligation to appear, answer or move is owed by each party named. Nothing connects the two.

As of 2026-09-06 the corpus records 327 canonical entities across the 95 originating actions, filling 673 party-slots. The distribution is extremely uneven:

The five largest actions by parties named:

C054 named 144 entities for the same $405 that C078 paid to name one. That is $2.81 per party. Across the 2026-07-29 block, $2,835 named 21 distinct entities — $135 per party. The filing fee is not a weak proxy for the burden a filing creates; it is uncorrelated with it by construction, because it is charged per docket number and the burden accrues per name in the caption.

The dispersal is as notable as the concentration. 127 of the 327 entities — 38.8% — are named in exactly one originating action, and 104 more in exactly two. Whatever this record looks like from the inside, from the position of well over a third of the people in it, it is a single lawsuit that arrived once.

How long it stayed is measurable on part of the corpus. Sixty of the 105 docket records carry both a filing date and an outcome date; the other 45 lack one or both, and every figure here is over those 60 and not over the corpus. The direction of that exclusion is knowable and should be stated rather than merely flagged: a record lacking an outcome date usually lacks one because the case has not ended, so the excluded 45 skew toward the longest exposures and every figure below is biased downward. The median is a floor on how long a defendant is kept in, not an estimate of it. Median 149 days open, first quartile 80, third quartile 326. Seventeen resolved inside 90 days and 34 inside 180 — so roughly half the measurable records closed within about six months, and a quarter ran past eleven. The longest is 2,021 days, and it is C100, which is one of the four actions filed against him.

What this chapter refuses to say about the defendants

00-proposal.md's outline described this chapter's subject as "336 defendants, overwhelmingly small." The count is 327 as of this date, at the canonical-entity grain. The characterisation is withdrawn, and the reason is a rule rather than a stylistic preference.

Entity size is not in the corpus. There is no column for it, and coding one would mean attaching a judgment — small, institutional — by name to living people and going businesses. Chapter 4 established what governs here: build_seo.py gives every canonical party with 2+ originating actions its own /defendants/ page, so a coded value written to research.db becomes a published label on a named person. A one-coder classification with no second rater does not earn that, which is why chapter 4's rung labels stayed in its exhibit. This chapter goes one step further and does not code the variable at all, because unlike the ladder it does not need it: the argument runs on party counts, one-shot share and exposure, which are all recorded facts.

One figure in this area is easy to over-read, so it is stated with its limitation attached. The corpus records counsel for exactly 6 entities. That number is not "only six defendants had lawyers." The attorneys table exists because those six lawyers and law firms were themselves sued and became parties; the corpus has no field for defence representation and does not track it. What the six support is a narrow point made in chapter 4 — that counsel appear in this record as opponents — and nothing about how the other 329 responded.

IV. The recovery, and which way it ran

P3 is refuted, and this is the chapter's principal finding.

Under the American Rule, as Alyeska Pipeline Service Co. v. Wilderness Society put it in 1975, "the prevailing litigant is ordinarily not entitled to collect a reasonable attorneys' fee from the loser." Three instruments could displace that here, and each has a trigger. Fed. R. Civ. P. 11(c)(1) permits a sanction "on any attorney, law firm, or party that violated the rule," but 11(c)(4) allows payment of the other side's fees only "if imposed on motion and warranted for effective deterrence," and 11(c)(2) requires the motion to be served and then withheld for 21 days while the challenged paper can be "withdrawn or appropriately corrected." 28 U.S.C. § 1927 reaches "[a]ny attorney or other person admitted to conduct cases" who "multiplies the proceedings … unreasonably and vexatiously" — by its terms, a counsel-facing provision. And the vexatious-litigant regimes chapter 2 surveyed restrain future filing rather than price past filing.

Across the 91 originating actions this plaintiff commenced, the corpus records:

• zero monetary sanctions;
• zero fee awards to any defendant;
• two sanctions of any kind, both non-monetary, both in Los Angeles Superior Court Department 205 before Judge Edward B. Moreton, Jr., both entered on the court's own order to show cause rather than on a defendant's motion, and both consisting of dismissing the complaint without prejudice (C101, 2025-10-13; C102, 2026-07-14);
• one vexatious-litigant motion still undecided. It was filed in C046 on 2026-03-31 and no ruling appeared on the docket read through entry 158 as of 2026-08-28 — five months. A second such argument, raised by ASOPRS in C102 on 2026-07-08, was held beyond the scope of the motion to quash it was raised in; the separate motion the court said would be required was never filed.

A terminating sanction without prejudice is worth pausing on. It ends the case, which is relief; it returns nothing, and it does not bar the next filing. In cost terms it is a refund of future expense, not a recovery of past expense.

The one time a court priced litigation, it priced his

There is exactly one measured legal-cost figure anywhere in this corpus, and it runs the other way. On 2026-05-21 final judgment was entered in C100 — Saladino, MD v. Frank Tufano, S.D.N.Y. 7:20-cv-09346 — for "$1 nominal damages, $52,500 punitive damages, $100,000 statutory damages under 15 U.S.C. Sec. 1125(d), $76,514 in attorney's fees, and $4,867.55 in costs."

Four things about that judgment matter here.

It is a defendant-side action: C100 is a suit brought against him, one of only four in the corpus. The plaintiff there was represented, and obtained a default judgment on 2025-01-30 after the court held on 2022-07-12 that the defaults were willful and no meritorious defence had been presented.

The fee award rests on a substantive fee-shifting provision, not a procedural one. 15 U.S.C. § 1117(a) provides that "[t]he court in exceptional cases may award reasonable attorney fees to the prevailing party," and the $100,000 is § 1117(d)'s statutory maximum for a cyberpiracy violation — "not less than $1,000 and not more than $100,000 per domain name." Nothing in that machinery has anything to do with how many actions anyone has filed. It attaches to a cause of action, and it fired because someone pleaded that cause of action and won on it.

It is larger than everything the vexatious-litigant regimes produced in this record combined, which is zero.

And C100 is the corpus's longest-running docket at 2,021 days — longer than any of the 91 actions he commenced.

That is the cost-side restatement of chapter 7's conclusion. Chapter 7 found that detection was never the binding constraint, and that the load-bearing reform is the trigger and the consequence rather than the aperture. Here the same shape appears in the money: the only consequence that landed in six years came from a statute that does not count filings at all. Every instrument that does count filings — Rule 11, § 1927, § 391, Rule 233.1, § 1651 — either needs a motion from a defendant, or reaches counsel rather than a party, or restrains the future without pricing the past.

V. What he says the other side pays

P4 survives, and the arithmetic that upholds it is stated before the quotations rather than after them. Of 1,667 statements in the corpus, 17 carry the factor Cost imposition as stated motive — 1.02% of the corpus, and 3.8% of the 445 statements carrying any filing-conduct factor at all.

Those three figures rest on this project's own classification of the statement corpus, which has no second coder and no reliability audit, so chapter 5 §II now reports figures of that kind as illustrative rather than as counts, and this section inherits the label. It cuts in the direction of caution here rather than against it: P4 claims the theme is marginal, and a classification that is wrong in either direction by a factor of two or three leaves it marginal. Nothing in this chapter turns on 17 being exactly 17. What it turns on is that the set is small, and that the seventeen passages in it are quoted verbatim and can be read. By volume it is a minor theme. Anyone reading the passages below and inferring that this is what the four-year statement record is mostly about would be reading it wrong.

What makes the 17 worth a section anyway is not their share. It is that fourteen of them are tiered Strong, that Cost imposition as stated motive is one of the three core factors in the published rubric, and that these are the statements in which a reason for filing is given in the same breath as the filing. Eleven of the 17 link to a tracked case.

The passages below are verbatim video_statements.StatementText values. Auto-caption mis-transcriptions are preserved rather than silently corrected, which is chapter 5's convention, and the [ __ ] markers are YouTube's own profanity masking.

Two of the seventeen are reproduced here and the rest are cited by id. That restraint is the point of putting the 1% first: a section that quoted at length from all seventeen would read like the corpus's dominant theme, and it is not.

He prices his own input in hours and the other side's in dollars, explicitly, and more than once — VS003, VS035, VS049, VS065, VS189 and VS1199 all do the same arithmetic. VS059 is the clearest:

But at least if I file the lawsuit, they're going to have to pay — if I file the lawsuit, technically my money is already avenged because each of those organizations are going to have to pay a lawyer like a $10,000 retainer minimum. So just me taking an hour out of my day to write up and file a lawsuit is already going to screw them over for more money than they've screwed me over. — VS059, linked to C015 and C057

Across the eight statements that name a figure the estimates run from $10,000 to $150,000: a $10,000 minimum retainer per entity (VS059, VS065, VS189, VS1199), $10,000–$15,000 (VS003, VS035), $20,000–$30,000 for a large platform (VS049), and $150,000 already spent by one set of opponents (VS737). The second quotation is that last one, and it earns its place because it is the only statement in the set that pairs the arithmetic with an open-ended filing horizon:

which means whether I have to file this lawsuit every year for the rest of my life, state and federal court, these people have already spent at least $150,000 in attorney fees. So whether these people want to pay me for this idea or not that they stole from me, I'm going to make sure they got to spend their money on something. They don't want to pay me, that's fine. I'll bury them in legal fees. — VS737, linked to C036 and C068

One further passage grounds the mechanism in the other side's inability to buy a defence (VS1069, linked to C040), and the remaining ten — VS001, VS023, VS037, VS501, VS738, VS828, VS1582 and VS1655 among them — are listed in the Sources table rather than quoted.

These are estimates by a party, not measurements. No row in research.db records what any defendant actually spent; there is no invoice, no fee petition and no declaration of costs anywhere in the corpus for any of the 91 actions he commenced. The retainer figures above are public statements, classified by this project under a published rubric, and they belong to the same evidence layer as any other thing a party says about his own case. The one number in this chapter that a court put on litigation cost is the $76,514 in §IV, and it was awarded against him.

VI. Why "restrict more" does not follow

Both fee-waiver regimes in this record exist to keep courts open to people who cannot pay, and both already carry their own filter. 28 U.S.C. § 1915(a)(1) authorises commencement "without prepayment of fees or security therefor," § 1915(d) puts service on the state — "[t]he officers of the court shall issue and serve all process" — and § 1915(e)(2) then requires that the court "shall dismiss the case at any time" if the action "is frivolous or malicious," "fails to state a claim," or "seeks monetary relief against a defendant who is immune." Pennsylvania's Rule 240 does the same job in two lines: subdivision (c) preserves the court's power to dismiss sua sponte "an underlying action, which is filed contemporaneously with an application to waive fees and costs, as frivolous."

Those screens work. C056 shows one working: it sits on an in forma pauperis screening track under § 1915(e)(2)(B) with service held in abeyance, and on 2026-03-16 the magistrate judge entered an order to show cause directing him both to pay the $405 fee or justify keeping IFP status, and separately to show cause why the action should not be dismissed on res judicata and issue preclusion. That is a fee-waiver screen reaching a prior case.

But it reaches one prior case, from inside one docket — and on this record the fee-waiver track is not where the volume is. Tightening § 1915 screening would have touched 4 of the 19 records whose fee status is recorded and none of the seven filed on 2026-07-29, because those seven paid. An access restriction aimed at people who cannot afford the fee does not reach a filer who is paying it.

Meanwhile the instruments that would price the conduct sit behind a trigger that this record shows is rarely pulled. Rule 11 fee-shifting requires a defendant's motion and a 21-day safe harbor. Pa. R.C.P. 233.1, as chapter 2 established, fires only on a defendant's motion. The one vexatious-litigant motion actually filed here has been pending five months. And of the people who would have to file those motions, 38.8% are parties named in exactly one action — that share is of entities, per §III, not of cases — facing a median exposure of at least 149 days, for whom the cheapest available response to a $405 complaint is very often to do nothing that costs more than the complaint did.

The Supreme Court has described the tension without resolving it. In re Sindram denied leave to proceed in forma pauperis to a repeat filer in 1991 on the ground that the goals of fairly dispensing justice are compromised by the processing of "repetitious and frivolous requests" — the Court applying, to its own docket, precisely the aggregation this thesis says nobody performs elsewhere. It could do that because every one of those filings arrived at the same clerk's window.

So the cost side does not produce a recommendation to restrict access. It produces a sharper version of chapter 7's negative result. The binding constraint is not the price of entry — it is that nothing prices the exit. A regime that made filing more expensive would fall hardest on the 4 records here where the fee was waived and hardly at all on the 12 where it was paid — on what the record says, with the selection caveat of §II attached — while the mechanism that could have made 91 actions expensive to their author went unused because it depends on a motion from whichever one-shot defendant happened to be named that week.

VII. What this chapter does not measure

Stated plainly, because several of these bound the claims above rather than decorating them:

• No defence cost is recorded anywhere in the corpus. Not one invoice, fee petition or cost declaration for any of the 91 actions he commenced. Every dollar figure in §V is a party's estimate.
• Fee status is unrecorded for 81 of the 101 records he commenced, and the 20 that are recorded are not a random sample of the 101. A value exists where a court had reason to remark on it, and fee payment is the remarkable case because it removes the § 1915(e)(2) screen. §II's table is therefore a floor and a direction, never a rate, and P1 is reported as unsupported rather than refuted for that reason.
• Exposure rests on 60 of 105 records, and the bias runs one way. The 45 excluded lack a filing date, an outcome date, or both, and are disproportionately still open — so they are disproportionately the long ones. Every exposure figure in §III is therefore biased downward, and the 149-day median is a floor rather than an estimate.
• pattern_relevance and pattern_factor have no inter-rater check. They were assigned by a single coder against the published rubric. The refiling variable at least had a rule-versus-judgment audit standing in for one, at 80.4% agreement across 18 adjudicated disagreements; the statement corpus, now the largest coded variable in the project, has nothing equivalent. Chapter 5 states this limit and it applies unchanged to §V.
• ProSe is blank on 45 of 105 records — 54 Yes against 60 recorded values, 3 No and 3 Mixed. Chapter 3 published this gap and it is not closed. The chapter therefore makes no quantitative claim about how many of these actions were filed without counsel, and the fee-asymmetry argument above is built on filing fees, party counts and the recovery record rather than on representation status.
• Entity size is not coded, deliberately, for the reason given in §III.
• Counsel is recorded for 6 entities and only because they were sued. The corpus does not track defence representation.