EDGAR accepted nothing on Monday — US federal holiday — so this edition reads Friday’s and this morning’s paper together. Both breaks are in the fields, not the dollar figures.
Edition 027 flagged this at 10.38% on the 5 October purchase. Four sessions later the gap is wider, not closed. Schedule 13G/A No. 4, filed 5 August for an event date of 30 June 2026, reports 27,185,941 shares and 8.8% of class. After this morning’s filing Drew Holdings Ltd holds 37,285,941 — +10,100,000 shares, +37.2% on that position in fourteen weeks, or 12.09% of class on a 308.51m share count. Inference the denominator is a market-data figure, not one stated in any filing.
The 9 October tranche is also the largest of the run and the most expensive: 2,500,000 shares at $4.7374, against $4.0245 on 13 August. He raised his bid on every tranche and paid the top price for the biggest block.
The consideration is $999,999.44 — 362,319 shares at $2.76 flat. Every other purchase and sale on this week’s tape carries a weighted average and a stated range, because a market order fills across a day. This one does not. $2.76 is the price at which the company’s $45.5m registered direct offering was struck on 5 October, closing 6 October — the transaction date on the form.
The stock closed Monday at $1.620, a new 52-week low against a 52-week high of $9.280, and 41.3% below the price on the form, four sessions later.
Nothing else in the window broke a person’s own pattern. Checked and found to be cadence: Ullal (Arista — two tranches, 6 and 7 Oct, plan dated 14 Nov 2025, field-level work carried in Edition 027), Tenev (Robinhood — 375,000 shares on 5 Oct, the fourth identical quarterly block of 2026; the Class A-to-zero artifact was Edition 024), Şahin (BioNTech — direct line to zero on 9 Oct, but 39,218,111 shares remain in Medine GmbH; 0.16% position change), von Ahn (Duolingo — Edition 030), Zatlyn (Cloudflare), Kurtz (CrowdStrike), Rothblatt (United Therapeutics), Dageville (Snowflake). Checked and found empty: Schedule 13D and 13D/A filings for 9 Oct, named-buyer prime property, Giving Pledge updates, domicile changes, dated auction results with a named principal.
A subscription filed as a purchase, an accumulation that has outrun its own disclosure, and a gift that is mostly a promise. Each is argued rather than summarised: what most readers will conclude, what I think instead, the specific fact that separates the two, and what would show me wrong.
Marcus Lemonis — Executive Chairman & CEO, director, Neighborhood Intelligence, Inc. (NASDAQ: NXH)
Acquired 362,319 common shares on 6 October 2026 at $2.76, consideration $999,999.44, reported under transaction code P. Holding after: 1,076,860 shares, all Direct — no indirect line, no footnotes, no Rule 10b5-1 plan reported.
On 5 October the company announced a $45.5m registered direct offering: 16,485,508 common shares or pre-funded warrants at $2.76, “priced at the closing price,” led by funds managed by Highbridge Capital Management LLC, placement agent Rodman & Renshaw, off a Form S-3 shelf effective 18 August 2026, closing on or about 6 October. Accompanying warrants for up to 16,123,189 shares at $3.45, unexercisable for six months. The release names the participants: the company’s “President’s List, including Executive Chairman and CEO Marcus Lemonis,” who “bought shares at the same price as other investors but did not take warrants.”
Form 4 — LEMONIS MARCUS, Neighborhood Intelligence Inc (NXH) · Director and Officer (Executive Chairman & CEO)
Accepted EDGAR 10/06/2026 16:02 · signed Christina Wheeler, attorney-in-fact, 10/06/2026
Table I: P, 10/06/2026, 362,319 (A), $2.76 → 1,076,860 D
Table II: none · Footnotes: none · Rule 10b5-1 plan: none reported
362,319 × $2.76 = $999,999.44 · prior holding 1,076,860 − 362,319 = 714,541 · increase +50.7%
Press release — Neighborhood Intelligence, 5 Oct 2026 (Business Wire): 16,485,508 shares at $2.76, ~$45.5m gross, closing on or about 6 Oct 2026; ATM facility reduced from $200m to ~$75m
The biggest insider purchase at this company in twelve months, and the CEO paid above where the stock now trades — the hardest version of the conviction signal to fake. It lands beside a business update guiding gross margin above 30% and synergies ahead of schedule. Insiders bought all year at an average of about $3.13 and none sold. Four outlets in this window read it exactly that way, and that reading is not unreasonable: the money is real and it is his.
This form carries no information about what Lemonis thinks the stock is worth, because he chose neither the price nor the day. It is the management-participation leg of a $45.5m registered direct offering that Highbridge led and that priced at the prior close. Read as a bid it is a $1m signal; read as what it is, it is the founder money that goes on the cover of a dilutive raise by a loss-making company. The signal is the raise, not the subscription — and the raise is roughly 17% of the company issued in a day. Inference share count derived from a $154.44m cap at $1.620.
1. The price field. This is the load-bearing fact. A market order for 362,319 shares fills across a day and is reported as a weighted average with a stated range. That is what every other filing on this week’s tape looks like: Trøim at $4.7374; Ullal at $217.1408, range $217.00–$217.18; Tenev in three bands, $113.2356 / $113.9682 / $114.9537; Control Empresarial at $16.9596, range $16.90–$17.06. Lemonis’s line reports $2.76, flat, with no range and no weighted-average footnote. One number is the signature of a subscription at a fixed price. It is not what open-market execution produces. Every price-driven story about this purchase predicts a range, and there isn’t one.
2. The price is the deal price, and the date is the closing date. The issuer says the offering was struck at $2.76, “the closing price,” on 5 October, and would close “on or about 6 October.” The Form 4 transaction date is 6 October. Those are the same event.
3. The consideration is a round dollar amount. $999,999.44 is what you get when someone wires $1,000,000 into a deal and receives whatever share count the deal price buys. A buyer working an order gives a broker shares or a limit, and lands on an untidy number.
4. The issuer names him as a participant in the offering. The Form 4 does not say he was in the deal; the press release does, by name and title, and specifies that he took no warrants.
Taken alone, any one of these is suggestive. Together, points 1 and 2 are close to dispositive: a single price and a date that equals the offering’s closing date leave no room for an open-market execution.
A Form 4/A, the 424(b)(5) prospectus supplement, or the closing 8-K showing that Lemonis’s 362,319 shares were bought in the market and were not part of the 16,485,508 shares sold in the registered direct offering. A broker-executed purchase reported with a weighted average and a price range for 6 October would do it on its own. So would a participant schedule allocating the full 16,485,508 shares to outside funds with no management tranche — that would mean his shares came from somewhere else, and the subscription reading fails.
His prior disclosed NXH purchases were small — one reported at $101,296, another of about 21,700 shares. This is roughly ten times either of them. Commentary in the window put the increase in his holding at 60%; the Form 4 arithmetic gives +50.7%, from 714,541 to 1,076,860. Behind the filing: NXH terminated a proposed acquisition of F9 Brands five weeks ago, terminated a merger with Fathom Holdings on about 5 October, appointed interim financial and operating leadership from Alvarez & Marsal in the same week, and closed Monday at a 52-week low of $1.620 with a market capitalisation of $154.44m, down 69.3%.
Tor Olav Trøim — director, Borr Drilling Ltd (NYSE: BORR) · buying through Drew Holdings Ltd, wholly owned by Drew Trust, a non-discretionary trust in which he is the beneficiary
Four open-market purchases: 1,500,000 at $4.1314 on 5 Oct; 1,500,000 at $4.2428 on 6 Oct; 1,250,000 at $4.31 on 7 Oct; 2,500,000 at $4.7374 on 9 Oct. Total 6,750,000 shares for approximately $29.8m, none under a Rule 10b5-1 plan. The 9 October tranche was accepted by EDGAR at 08:28 this morning and carries a deemed execution date of 12 October, the federal holiday.
| Date | Shares | Price | Holding after |
|---|---|---|---|
| 30 Jun 2026 — 13G/A event date | — | — | 27,185,941 |
| 13 Aug 2026 | +1,500,000 | $4.0245 | 28,685,941 |
| 5 Oct 2026 | +1,500,000 | $4.1314 | 32,035,941 |
| 6 Oct 2026 | +1,500,000 | $4.2428 | 33,535,941 |
| 7 Oct 2026 | +1,250,000 | $4.3100 | 34,785,941 |
| 9 Oct 2026 | +2,500,000 | $4.7374 | 37,285,941 |
| Since the 13G/A event date | +10,100,000 | +17.7% | +37.2% |
Form 4 — TROIM TOR OLAV, Borr Drilling Ltd (BORR) · Director — 10% Owner box not checked
Accepted EDGAR 10/13/2026 08:28 EDT · signed Temi Bankole, attorney-in-fact, 10/13/2026
Table I: P, 10/09/2026 (deemed execution 10/12/2026), 2,500,000 (A), $4.7374 → 37,285,941 I — by Drew Holdings Ltd.; 81,867 D
Footnote 1: “Drew Holdings Ltd. is wholly owned by Drew Trust, a non-discretionary trust in which the reporting person is the beneficiary.”
Schedule 13G/A No. 4 — Drew Holdings Ltd., filed 5 Aug 2026, signed 4 Aug 2026 by Erling Lind, director · event date 30 Jun 2026
27,185,941 shares · 8.8% of class · sole voting 27,185,941 · sole dispositive 27,185,941 · Item 10 certifications: Not Applicable
“may be deemed to be beneficially owned by Mr. Tor Olav Troim”
Form 6-K — Borr Drilling Ltd, accepted EDGAR 10/01/2026 06:02 · press release dated 30 Sep 2026, Hamilton, Bermuda
Norve: binding letter of award, three wells, ~245 days, start mid-Dec 2026 · Natt: Shell Nigeria exercises one-well option · Joro: UK accommodation contract, ~60 days from mid-Oct
A director with an already-large stake kept buying his own company — nearly $30m in four sessions, no trading plan, straight open-market risk. Jack-up day rates are firming, the company has just added three rig commitments running into August 2027, and the stock is still under $5. Insider buying at size, with the insider’s own money, after contract news, is about as clean a bullish signal as the tape produces.
The interesting field on this form is the one left blank. Drew Holdings’ standing beneficial-ownership filing says 8.8%; the arithmetic on the same vehicle this morning says above 12%. Two current documents disagree about how much of Borr this trust owns, and the gap has widened through five consecutive Form 4s. Separately, the shape of the buying is wrong for conviction-at-a-price: he raised his bid on every tranche and paid his highest price for his largest block. That is a buyer working toward a position size, on a schedule. The information here is about a target and a disclosure lag, not about day rates.
On the disclosure gap. The 13G/A event-date figure (27,185,941 / 8.8%, sole voting and sole dispositive) against the Form 4 post-trade figure (37,285,941) is a +10,100,000 share, +37.2% change in the same vehicle’s position with no amendment on file. Ten percent is not a round number, it is a status line — the level at which a passive filing has to be amended and at which an investor becomes an owner for several disclosure purposes. The Form 4 filed at 08:28 this morning still checks only Director. Inference my reading of which amendment deadline applies, not a statement in any filing.
On the shape of the buying. $4.0245 → $4.1314 → $4.2428 → $4.31 → $4.7374. Five tranches, five higher prices, and the 2,500,000-share block — 37% of the whole run — executed at the top of that ladder. A buyer optimising entry scales into weakness; the Fisher filings in Edition 027 showed what price-insensitivity looks like on the sell side, and this is the same signature inverted. Size rising with price is the fact that separates the two readings, because the conviction-at-a-price story predicts the opposite relationship.
And the cheap contrarian story is dead on the dates. The obvious suspicion — he was buying ahead of the rig announcement — does not survive the calendar. The Norve, Natt and Joro commitments were released on 30 September and furnished on a 6-K accepted 1 October at 06:02. Every October purchase is after that. He is not front-running news; he is paying up for stock that was already repriced on it.
A Schedule 13D or 13G/A from Drew Holdings covering the October purchases. If it appears and is within its deadline, then what I measured is a filing lag caught mid-window and nothing more — that is the most likely single outcome and I would be wrong to have called it anything else. The percentage claim fails separately if Borr’s actual shares outstanding are materially above the 308.51m market-data figure; the 8.8% in the 13G/A implies roughly 308.9m at 30 June, so a large issuance since then would move it. On the averaging-up claim: one tranche bought materially below a prior one, or a disclosed target-size mandate, reframes it as ordinary accumulation.
This is not a first purchase — he bought 1,500,000 shares on 13 August and has been adding through the autumn. What changed in October is the rate. Fourteen weeks of accumulation has taken the position up 37.2%, and 6,750,000 of those 10,100,000 shares were bought in the last four sessions of the window. His direct line is unchanged at 81,867 shares, which is 0.22% of what he controls.
Kenneth C. Griffin — founder and CEO, Citadel · personal philanthropy
7 October 2026: an additional $25m to Baptist Health South Florida’s neuroscience institute in Miami, announced as Baptist Health broke ground on a new building on the Baptist Hospital campus off Kendall Drive. The money funds the institute’s Transformational Impact Fund. It takes his cumulative support to $75m; the prior gift was $50m in 2024, which Baptist Health called the largest single donation in its history. The institute, formerly the Miami Neuroscience Institute, now carries his name.
Context, outside the window: on 30 September Carnegie Mellon announced that Griffin “has committed $3 billion” — $1bn to the Pittsburgh campus, split evenly between the university’s highest priorities and the School of Computer Science, which is renamed for him, and $2bn to launch CMU Miami on 35 acres in Wynwood. Construction is slated to begin in 2027, graduate enrolment in 2028, the first undergraduate cohort in 2032, all “subject to regulatory approvals.” Griffin joins the Board of Trustees.
The largest single gift in the history of American higher education, followed a week later by a further $25m to a Miami hospital. Griffin is converting a trading fortune into permanent civic institutions in the city he moved his firm to, at a scale that exceeds most universities’ entire endowments. The Florida framing is not a criticism of it; building a research university and a neuroscience institute where your people live is what civic capital has always looked like.
Only one of these is a transfer that has happened. The $25m is dated, tied to a groundbreaking, and names the fund receiving it. The $3bn is a commitment, and its larger part capitalises an institution that does not enrol an undergraduate until 2032, is conditional on regulatory approvals, and will be governed by a board the donor has just joined. The accurate description is a long-dated, milestone-linked capital commitment to build something, with naming and governance attached — which is a different object from a 2026 disposition of $3bn, and should not be counted as one.
The university’s own verb. CMU’s presidential letter of 30 September says Griffin “has committed $3 billion.” It does not say gave, transferred, or funded, and it publishes no disbursement schedule.
The use of funds is a physical build on a six-year clock. $2bn of the $3bn “launches CMU Miami”: 35 acres, construction from 2027, graduate enrolment 2028, first undergraduates 2032, plans “subject to regulatory approvals.” Money cannot be spent faster than the thing it builds, and a commitment contingent on approvals is not an unconditional transfer.
A governance term is attached. Griffin joins the Board of Trustees. That is a feature of a capital partnership, not of an unrestricted gift, and it is the kind of term that typically accompanies staged funding.
Against all of which, the Baptist Health item carries the markers of a completed transfer and the CMU one does not: a specific date, a named receiving fund, a cumulative total that moved from $50m to $75m, and an event — the groundbreaking — that the money was released against.
Where this stops. I could not find a published disbursement schedule, and the absence of one is not evidence that it is long. Two hypotheses are live: (a) funded now into a university quasi-endowment or restricted fund and simply not described that way, or (b) a multi-year pledge keyed to construction milestones. What distinguishes them is a single line item: in CMU’s FY2027 audited statements, a funded gift appears as contribution revenue and cash; a pledge appears as a discounted pledge receivable. Until that is published, nobody writing “largest donation ever made” knows whether they are describing a transfer or an intention with a number on it. I am claiming the second is unproven, not that the first is false.
CMU disclosing an irrevocable, already-funded transfer, or recording the $3bn as contributions received rather than pledges receivable in its FY2027 financial statements. A Griffin-side disclosure of a completed 2026 transfer would do the same. Any of those and the ordinary reading stands and this item was wrong.
The $25m is an increase on an existing relationship rather than a new one — second tranche, same institution, name already on the building. That is the quieter and more reliable shape: a donor who has already committed once, releasing more against a construction milestone.
One item I went after and could not overturn.
Control Empresarial de Capitales S.A. de C.V. — director and 10% owner of PBF Energy and Talos Energy · Form 4s signed by Marco Antonio Slim Domit as attorney-in-fact
PBF Energy: 100,000 Class A shares sold 1 Oct at $82.00–$82.005, about $8.2m, leaving 13,616,127; then 120,000 shares on 5 Oct at about $84.13, $10,095,096. Talos Energy: 480,000 shares sold 6 Oct at a weighted average $16.9596 (range $16.90–$17.06), $8,140,607, leaving 38,599,036.
Whether this is something other than trimming. Three candidates: a mechanical programme running on a calendar; a manoeuvre around a disclosure threshold on the 24.5% Talos stake; or the tell that would make it an allocation decision rather than a view — selling the winner and the loser in the same week, which is what a holder raising cash does and what a holder with a view does not.
There is no loser. Both names were sold near the top of their own 52-week ranges. PBF went at $82–$84 against a 52-week high of $85.08 after a 216% year to date. Talos went at $16.96 against a 52-week range of $8.97–$18.73 — 90.6% of its high. The asymmetry the liquidity story needs is simply not in the prices, and without it the story has nothing to stand on. I dropped it.
The pace says trim, not exit. Schedule 13D/A No. 4, event date 26 March 2026, reports 41,233,604 Talos shares, 24.5% of class, with zero sole and 41,233,604 shared voting and dispositive power across eight named Slim family reporting persons. The holding is now 38,599,036 — −2,634,568 shares, −6.4% of the position in six and a half months, leaving 22.9% of the company. Extrapolated, that is a fifteen-year exit. Nobody exits over fifteen years.
And the price-insensitivity signature is absent. The shape that gave away the Fisher wind-down in Edition 027 — sharply larger size at a materially worse price — is not here. Sizes are modest and the prices are the best available. That is what taking profit looks like, and it is what the commentary says it is.
A Schedule 13D/A taking the Talos stake through a round threshold, or a tranche sold at a materially worse price than the one before it at materially larger size. Either would reopen the question. Neither is on file.
The same structural fact sits underneath four of this week’s five names, and it is the reason each of them was misreported.
Not one of the large transactors in this window traded in their own name. Every move ran through a company or a trust, and in each case the individual’s own direct line is a rounding error against what they control:
The counter-example is what makes the point. Lemonis holds 1,076,860 shares reported Direct, with no indirect line on the form at all — and his company is worth $154m. The structure arrives with the zeroes.
This is also the seam every misreading in this edition opened along. “Arista’s CEO sold $118m” mistakes the filer for the owner. “Şahin’s holding went to zero” mistakes one pocket for the person. “Lemonis bought $1m” mistakes a subscription for a bid. In each case the error is at the entity boundary: who transacted, out of which pocket, on whose instruction. Four of five filings this week are unreadable without answering those three questions, and none of the headlines asked any of them.
What survives the trip down from a $5bn stake to one founder with a private Ontario corporation, and what plainly does not.
This week’s mechanic is not exotic and it is not a function of scale. Trøim owns Borr through a Bermudian company owned by a trust; Slim owns Talos through a Mexican S.A. de C.V.; the individual is never on the ticket. The Ontario equivalent is ordinary: the operating company held by a holdco, with the growth shares issued to a discretionary family trust.
What it actually buys, in this order: creditor separation between the trading risk and the accumulated cash; movement of retained earnings up to the holdco on a tax-deferred basis using the inter-corporate dividend deduction in section 112; and purification of the operating company so its shares can satisfy the asset tests for QSBC status and the lifetime capital gains exemption. The last one is the expensive one to get wrong, because it is tested over the 24 months before a sale, not on the day of it.
First step this week. Get a current, documented fair market valuation of your shares. Everything below is priced off that number, the number only goes up, and it is the one input you cannot backdate.
The reason Ullal’s $118m of selling tells you nothing about her view of Arista is that the shares were not hers to have a view about — roughly 25 million of them sit in trusts for her children and relatives, and her own line is 9,917 shares. That wealth left her estate years ago, while it was worth less.
The Canadian instrument is the estate freeze: exchange your common shares for fixed-value preferred shares that lock in today’s valuation, and issue new common to a family trust so all future growth accrues there. On a later sale, the gain in the trust can be allocated to multiple beneficiaries and sheltered against each of their lifetime capital gains exemptions, which is the structural reason founders do this early rather than late.
Three things bite in Ontario and all three need an accountant before anything is signed: TOSI restricts which family members can receive dividends at ordinary rates; the 21-year deemed disposition forces a decision before the trust turns 21; and attribution rules apply to a spouse and to minor children. The freeze is also irreversible in practice at the value you freeze at — which is the second reason step one is the valuation.
What gave the Lemonis filing away was a single clean price where a range should have been. The private-company version of that mistake is the one that actually costs money: when a corporation issues or transfers shares to an insider, the price is the filing, and CRA will test whether it was fair market value at the date of issue. The time to be able to prove it is the day it happens.
First step this week. For every share issuance or transfer your company has done, make sure the valuation that supported the price is on file, dated, with the method written down. If the number was picked rather than derived, that is worth fixing now rather than in an audit, and it is a conversation with your accountant, not a project.
The distinction the Griffin item turns on is the same one Canadian tax law draws. A pledge is worth nothing on a T1 or a T2 until it is funded: the donation credit (personally) or deduction (corporately) arises in the year the gift is made. Announcing is not giving, for the donor as much as for the recipient.
The part worth filing away for later: donating publicly traded securities in kind to a registered charity eliminates the capital gains inclusion on those securities, and where a corporation does it, the untaxed portion credits the capital dividend account, which can then be paid out to shareholders tax-free. That is a decision made before a sale, not after one — which is why it belongs in a plan now rather than in the week of a liquidity event.
Half of what the tape shows this week has no private-company analogue at all. A Rule 10b5-1 plan exists to create an affirmative defence against insider-trading liability for a person who trades in a continuously quoted market; the Canadian cousin, the automatic securities disposition plan, is a public-issuer instrument. Nothing in the Ullal, Şahin or Tenev items scales down to a company whose shares do not trade.
Trøim’s averaging-up does not travel either, and for a more interesting reason. It needs a continuously quoted price and an anonymous counterparty willing to sell. In a private company there is no market to buy from: increasing your stake means negotiating with one identified shareholder, and as Edition 030 set out, whether that is structured as a share purchase between individuals or a redemption by the company decides the tax bill — a subsection 84(3) deemed dividend on a redemption is often materially worse for the person leaving, and can cost them their capital gains exemption on those shares. The arithmetic of “my percentage went up” carries over. The economics do not.