Wealth Signal

Evidence that discriminates

1. Each tranche takes a larger share of what remains, at a higher price. 26.2% at ~$199.60, then 25.5% at ~$217, then 36.5% at $236.50. This is the fact that separates the two readings. A seller with a price target decelerates as the price runs past it — that is what a target is. A seller working down a position accelerates as the remaining block gets small enough to clear. The percentage column moves the wrong way for the consensus story and the right way for mine.
2. No Rule 10b5-1 plan on any of the four tranches. This is the single cheapest thing a director can do and he has not done it four times running. A multi-quarter diversification programme is precisely the use case the rule exists for: adopt once, sell on a schedule, never face the question again. Four consecutive discretionary sales means each decision is being made fresh, in an open window, by someone who would rather keep the discretion than have the shield. That is what fund-level selling looks like. It is not what a pre-decided personal programme looks like.
3. What he protected is the seat, not the exposure. The 6,780 directly-held shares and the quarterly stock retainer are untouched across all three filings. Before August, those retainers — about 107 shares a quarter — were the entirety of his Illumina Form 4 activity. The holding that qualifies him as a director survives; the holding that is the investment is going.
4. The “37%” in every headline is measured off the September residual. 281,433 of 770,350 is 36.5%. Off the corrected August base of 2,830,552 the cumulative figure is 82.7%. Both numbers are true; only one of them describes the decision.

Why it happened is not knowable from these documents. Three live hypotheses, with what would separate them: (a) fund-level — Corvex redemptions or risk limits forcing a wind-down irrespective of view; distinguished by whether Corvex’s next 13F shows other large positions shrinking in the same quarters and total reported value falling. (b) Thesis realised — Meister thinks the re-rating is done; distinguished by whether the residual stabilises near 489k rather than continuing to zero. (c) A glide path he agreed to — terms attached to the March 2025 board appointment; distinguished by the cooperation or standstill agreement, if one exists, in Illumina’s proxy. The 25 March 2025 announcement names no such agreement, which is weak evidence against (c), not proof.

Evidence that discriminates

1. Size is monotonic; price is not. 2,526,702 → 2,698,400 → 3,130,149 shares, against $3.7209 → $3.8537 → $3.7220. The middle day was the most expensive and got more shares than the cheapest first day. If the thesis were “this is cheap, buy more when cheaper,” the ordering would invert at least once. It never does. A steadily ramping daily quantity is the signature of an order being worked, not a view being expressed.
2. The September tranche is inverted the same way, in the opposite direction. 25 September: 1,332,742 shares at $5.0639. 29 September: 1,000,000 shares at $4.6062 — 25% fewer shares at a 9.0% lower price. Two separate episodes, six weeks apart, in both of which the quantity ignores the discount. That is two independent observations of the same indifference, which is what lifts this above a three-day coincidence.
3. The same balance sheet was buying something else at the same time. PFC SICAV bought 986,557 Dole shares on 2–4 November for about $12.9M, overlapping these sessions exactly. $44.4M into two unrelated businesses — an Irish-domiciled fresh-produce company and a US online-education platform — inside four days. Two simultaneous high-conviction stock calls is a strained explanation. One allocator deploying a determined amount of capital is not.
4. They disclaim, and they hold nothing personally. Every row is indirect, held by the SICAV; Šenkypl and Barta appear as control persons and each disclaims beneficial ownership except to the extent of pecuniary interest. The 29 September 13G/A reports sole voting power 0.00 and shared voting power 28,806,415 for every reporting person, with the Item 10 certification that the shares are not held to change or influence control. Neither man is buying Coursera. A fund they control is.

Where I am inferring and where I am short. The participation-programme reading would be much stronger with daily volume — if the three days’ share counts are a near-constant fraction of each day’s traded volume, the case is close to settled. I did not obtain volume data, so the monotonic-size argument stands on its own and is suggestive rather than conclusive. Separately, Investing.com describes Pale Fire Capital SE as “a 10% owner and director” of Coursera; the Form 4 as rendered checks only the 10% Owner box. I could not resolve that and have not relied on a board seat either way.

Evidence that discriminates

1. Both ratios favour the CTO, which rules out cherry-picking the metric. Dollars: $1,995,978 vs $1,030,172. Proportion of existing stake: 54% vs 9.7%. When a comparison can be reversed by choosing the denominator, it is worth little; here it cannot be. Whichever way you measure, the executive who is not the CEO committed more.
2. A CEO’s stake is mostly granted; this top-up is marginal to it. Fries entered the day with 1,128,525 Class A shares plus 401(k) holdings of 8,135 Class A and 31,299 Class C. Against a position of that size, accumulated over a two-decade tenure, $1.03M moves his exposure by less than a tenth. Rodriguez’s derived prior holding of about 391,065 shares is roughly a third of Fries’, and he raised it by more than half in one session.
3. Neither used a plan, and the date is the day after the print. This is what removes the alternative explanation that either purchase was scheduled or automatic. Both men chose 4 November, in an open window, one day after a quarter that showed EBITDA down about 10%.

Inference, clearly labelled: Rodriguez’s pre-trade holding of ~391,065 shares is derived from the reported post-trade 602,378 and the stated 54% increase, not read off a filing field. If the 54% figure is itself derived by the aggregator, the proportional comparison inherits that uncertainty. The dollar comparison does not — both dollar figures come straight from the filings. I also have not seen either man’s prior purchase history, so I cannot say whether a 54% top-up is unusual for Rodriguez; that is why this entry is in the log and not on the anomaly board.

Evidence that discriminates

1. He paid up, and that is the number that kills it. The second tranche of 231,243 shares went off at a weighted average $13.632 — 7.4% above the $12.694 of the first, with individual fills as high as $13.87 on a day the stock printed as low as $12.34. A purchase made to be seen is made at the low and stopped. Paying 7.4% up for the back half of an order is paying for immediacy, which is what someone does when they want the shares, not the disclosure.
2. The buyback is the board’s act and it preceded his. The $200M authorisation came with Q3 results; his purchase came the following session, after the stock had fallen a further 7% to a new one-year low. The sequence is market-driven. A staged pairing would have put them on the same day.
3. No plan, discretionary, at the low. There is no 10b5-1 shield here and no ambiguity about timing: he chose the worst-looking day of a 41% drawdown. That is the thing the consensus is reading, and it reads it correctly.
4. The one real correction to consensus is scale, not sincerity. This is a top-up, not a transformation. Galaxy Group Funding ECI U LLC held nothing before and holds 760,000 shares now; his unencumbered direct Class A holding is 113,273 shares once the 327,065 unvested RSU-settlement shares are set aside. The bulk of his economic exposure to Galaxy sits outside the Class A line entirely, so $9.86M should move your priors some, but not as much as a headline that frames it as “Novogratz bets on Galaxy” implies.