Wealth Signal

Evidence that discriminates

1. Proposition 40 fixes residency as of 1 January 2026. Anyone who was a California resident on that date stays within the Act’s scope “even if the person later fully left the state.” Net worth is measured 31 December 2026; payment is due 15 April 2027. Moving in August, September or October 2026 does nothing. This is the fact that breaks the consensus story on its own terms — the threatened action could not achieve the stated purpose. (LAO official ballot analysis; Holland & Knight.)
2. The deadline he named was the election, not a tax-year boundary. On 4 August he said he intended to establish residence outside California before the 3 November election. A move driven by California income tax on the multi-billion-dollar liquidation he would need to fund the bill would be timed to the sale and to 31 December, not to a vote. A deadline set by an election is a political deadline.
3. The concession he accepted does not address tax at all. Class C gives him liquidity without vote dilution. It does not reduce his Proposition 40 liability, and it does not reduce the California income tax on the shares he sells to pay it. If tax were the motive, this package would not have satisfied him. It did — within five days of the 9 September dinner.
4. The recapitalisation predates the tax. The special committee was formed in February 2026; Proposition 40 entered the discussion in July. The Class C distribution proceeds whether or not the measure passes. So the share class is not caused by the tax, which means the tax cannot be what he traded the residency pledge for.
5. The unlock schedule is the structural inverse of what the committee asked for. A sunset terminates control when the founder leaves. The stapling release — 2.5% a quarter, conditional on remaining a full-time executive or executive chair — pays him to stay. He did not merely defeat the sunset; he obtained its mirror image.

Inference, labelled: points 1, 2 and 4 are filing and ballot-text facts. The reading that the residency threat was deployed as leverage rather than intended is mine, and it rests on the gap between what the threat could achieve and what he accepted in return.

Evidence that discriminates

1. The effect size against approved comparators in the same indication. RAP-219: 3.1 points placebo-adjusted on YMRS at week 3, p=0.045. Cariprazine’s registrational trial: LSMD −6.1 (3–6 mg) and −5.9 (6–12 mg), both p<.001. Olanzapine’s: 6.65 points, p<.001. RAP-219 delivers about half the benchmark on a p-value that would not survive a single pre-specified multiplicity correction. A cluster buy is consistent with both readings; the effect-size comparison is only consistent with the fade being informed. That is what makes it the crux rather than colour.
2. The CEO is absent from the bid. Abraham Ceesay, who had the earliest sight of the readout timeline, sold 4,522 shares on 6–7 October at $30.66–$32.03 under a plan adopted 27 March. He retains 452,245 shares directly, so he is not exiting — but he is not buying either, and his own CFO, who signed that Form 4 as attorney-in-fact on 8 October, bought two weeks later. The six buyers are four outside directors, a finance chief and a fund. The scientific principal is not among them.
3. 62% of the headline figure is not personal money. Steven M. Paul’s $974,277 was bought through Paul Capital Strategy III Fund LLC, reported as indirect ownership with a disclaimer of beneficial ownership beyond his pecuniary interest. The Form 4 ownership field says so; the headline “six insiders, $1.6M” does not.
4. The sizes cluster, the convictions do not. $98,460 / $99,636 / ~$100,000 / $144,000. Four of the six land within 2% of $100K. Independent valuation judgements do not converge on a round number. (Inference. The amounts are from the filings; the reading is mine.)

Also inference: the cross-trial effect-size comparison is not a head-to-head. Placebo response in mania trials has risen over the decades, which compresses modern deltas for design reasons rather than drug reasons — RAP-219’s 15.4-point absolute reduction implies a placebo arm near 12.3 points, which is high. That is the strongest counter to my read and it is a real one.

Evidence that discriminates

1. The discount did not move. Across four weeks and 1.54 million shares, the price went $12.50 (28 Sep) → $13.17 (9 Oct) → $12.45 (23 Oct). The campaign ended below where it started. If this were value buying into an irrational discount, tightening the float by that much should narrow it. It did not. A conviction buyer who is right gets paid; a buyer absorbing structural supply does not.
2. The buyer is the fee-earning entity, not the man. Every filing reports indirect ownership through Bluerock Asset Management LLC. Kamfar’s personal account appears in none of them. A principal who believes a dollar is on sale for fifty-five cents buys it himself; he does not route it through the company whose revenue is a percentage of the number being defended. This is the single field that separates the two readings and it is the field every headline drops.
3. The fund is bidding too, and says why. BPRE announced a repurchase programme on 16 December 2025, explicitly for buying shares back “when they trade below net asset value.” Two balance sheets on the same bid is a price-support operation. Two independent investors reaching the same conclusion do not coordinate their timing to the day.
4. The cadence. Sixteen trading days in four weeks, essentially continuous. Conviction accumulation is episodic and price-seeking; within the 6–9 Oct tranche he bought the most shares on the cheapest day and the second-most on the dearest. There is no price sensitivity in the pattern. (Inference from the Form 4 line items.)

Evidence that discriminates — there isn’t any, and I am saying so

I looked for a fact that separates the readings and could not find one. The structural change is real and documented: allocation authority moves from the donor to thirteen people he has no governance relationship with. The reason is not recoverable from anything public. So rather than dress a guess as a finding, here are the live hypotheses and what would separate them.

1. Prestige arbitrage. A Nobel-laureate-branded prize buys more reputational return per dollar than another named building. Separator: whether the trust’s communications foreground laureate names over recipient organisations.
2. Buying allocation signal. He is using laureates as scouts for where science money should go, cheaply, before committing larger sums himself. Separator: whether amounts scale after the first cohort, and whether the trust follows laureates’ picks with its own larger direct grants.
3. Succession design. A rules-based annual programme runs after the donor; discretionary mega-gifts do not. Wertheim is in his eighties and a Giving Pledge signatory. Separator: whether the trust documents commit to perpetual annual cohorts.

Hypothesis 2 is the one with a real wealth-mechanics claim in it and it is the one I would bet on, at maybe 40%. I am not presenting that as a finding.

What the fields imply next

The plan terminates on the earlier of 1,734,410 options exercised or 31 December 2026. She holds 195,410 options after this tranche, so the option ceiling is unreachable and the date governs. At 9,500 per tranche and roughly weekly cadence, about ten more tranches fit before year end — call it 95,000 options. That leaves on the order of 100,000 in-the-money options expiring 15 March 2027 with no plan covering them after 31 December, worth roughly $44M of intrinsic value at $563.

So expect one of three things before year end: a new or amended 10b5-1 plan, a sharp acceleration in tranche size, or a large block exercised in the open window in Q1 2027. This is arithmetic from the filing, not a view on the stock. If none of the three happens, I have mis-read the plan’s terms. (The projection is mine; every input is a filing field.)