Posts
How We Did This
The Methodology Behind THE CHOCOLATE MACHINE — What the Word "Again" Actually Means, What Surprised Us, What We Almost Got Wrong, and What 35 Posts Across Four Series Produced
THE CHOCOLATE MACHINE — Series Capstone C | February 2026
Why Series 4 Is Different
The first three series investigated extraction from diffuse public resources. NFL stadium subsidies harm taxpayers generally. Defense contractor capture harms federal budgets generally. University endowment tax exemptions benefit wealthy institutions at the expense of the general public generally. The victims are real, but they are not named. They cannot be counted precisely. They are "the public" or "future generations" or "communities" — categories, not individuals.
The Milton Hershey Trust is different in one structural way that makes it harder to write about and more important to document: the deed names the victims. Not categories. Specific children — poor, lacking adequate parental care, in Pennsylvania, ages 4-15 — who qualify under criteria established in 1909 and apply to a school that receives more applications than it accepts. The children who applied and didn't get in because enrollment is fixed at 2,100 are specific human beings. They exist. They are not hypothetical. Their names aren't public. Their number isn't disclosed. But the criteria that produced their rejection are published on mhskids.org in the school's own words.
That specificity — the named victim class, the mandatory deed language, the sworn court filing with the word "again" — is what makes this the most morally direct investigation of the four. It is also what makes it the most difficult to write fairly. The school is genuinely extraordinary for the children inside it. The board is not uniformly corrupt. The Catherine Hershey Schools are real and serving real children. The fair account is genuinely complicated. We tried to hold all of it.
The Research Process
STEP 1 — THE 1909 DEED FIRST:
The user provided the actual deed of trust — the founding legal
document — before research began. That document set the standard
against which every subsequent finding was measured. Four words
in the deed define the entire investigation.
STEP 2 — PRIMARY SOURCES ONLY:
Court filings (In re Milton Hershey School Trust, 807 A.2d 324).
The school’s own website (mhskids.org). The trust’s own case studies.
IRS Form 990 tax filings. Orphans’ Court records. Attorney General
settlement documents. Every claim sourced before it was written.
STEP 3 — COUNTERARGUMENT FIRST:
Every post documents the strongest case for the trust’s position
before making the structural argument. The perpetuity concern is real.
The quality-vs-quantity tension is real. The CHS expansion is real.
The fair account is presented honestly, not as a fig leaf.
STEP 4 — LABEL WHAT WE DON’T KNOW:
We don’t know how many children apply and are rejected annually.
We don’t know the size of the waitlist. We don’t know what
“additional factors” in the school’s “sole discretion” include.
We said so. Clearly. In every post where it mattered.
STEP 5 — THE RECEIPTS OR NOTHING:
The standard that has held across all four series: if we can’t
show the primary source, we don’t make the claim.
What Surprised Us
The most devastating finding of the series was a single word in a single court document. January 20, 1999. Robert Vowler, President of the Hershey Trust Company, and Dr. William Lepley, President of the Milton Hershey School, filed a verified petition saying the surplus had “again grown to the point where it is more than sufficient.” Not grown. Again grown. Under oath. In court. Preserved permanently in the Pennsylvania Commonwealth Court record at 807 A.2d 324.
We expected to find documented evidence of the gap between mandate and practice. We did not expect to find the trust’s own leadership acknowledging it in sworn testimony in 1999 — while predicting it would “continue to grow further beyond the amount necessary” without structural change. They were describing the mechanism. From inside it. Under oath. Twenty-five years before this investigation began.
The Pennsylvania Supreme Court’s 2006 ruling denying the Alumni Association standing to enforce the deed is the single most consequential legal decision in the trust’s modern history — and it received almost no sustained attention in the journalism covering the trust. The ruling did not just deny the Alumni Association standing. It established permanently that only the AG can enforce the deed. That ruling, combined with the AG’s consistent preference for negotiated settlements over litigation about enrollment numbers, means the question “does 2,100 satisfy ‘as many as possible’?” has never been asked in any courtroom and probably never will be. That’s not a policy gap. It’s a structural closure. The door was documented, litigated to the Pennsylvania Supreme Court, and shut. In 2006. We didn’t know that before we started.
In September 2002, Slate published an article titled “How Pennsylvania officials screwed poor kids out of $1 billion by stopping the sale of Hershey.” The argument: the politically motivated AG intervention blocked a $89/share sale — and Hershey stock fell back to $65 the same day the sale died. The children’s trust absorbed a paper loss of hundreds of millions in a single afternoon of political theater that was driven by electoral calculations, not financial analysis. A trust that sold at $89/share in 2002 and reinvested in a diversified portfolio might have ended 2024 with assets comparable to or larger than what the current concentrated Hershey stake represents — with dramatically more liquidity and less concentration risk. The trust has never publicly made this counterargument. The community that marched to stop the sale has never engaged with it. The argument is serious. We documented it. In full.
What We Almost Got Wrong
Early drafts of Posts 2 and 5 treated the deed’s income-only restriction as if it were the same as the principal restriction on private foundations. It isn’t. The deed restricts spending to income — not to 5% of assets as required of private foundations. The trust is not a private foundation and is not governed by that federal requirement. Pennsylvania’s 1998 law allowing 7% of assets annually applies to charitable trusts — but only if the board adopts it by vote. The board chose not to. We clarified this distinction carefully throughout the series to avoid conflating two different legal regimes.
Early drafts of Post 7 came close to characterizing the Catherine Hershey Schools as straightforwardly cynical — a redirect designed to deflect criticism without expanding MHS enrollment. The primary sources don’t support that characterization. The buildings are real. The program is serious. The early childhood research backing it is robust. The two-generational model is innovative. The Orphans’ Court reviewed and approved it. The correct framing — which we landed on — is the distinction between two legitimate questions: Is CHS good? (Yes.) Does CHS satisfy the deed’s enrollment mandate? (No court has ruled it does.) Those are different questions. We held the distinction.
Bob Heist’s lawsuit is the most dramatic governance moment in the series. Early framing treated it as evidence of active concealment — the school hiding something from its own chairman. The available primary sources are more limited than that framing suggests. We know Heist requested records. We know the school resisted. We know a judge pressed both sides. We know the case was resolved privately without public findings. We do not know what the records contained, whether they revealed wrongdoing, or why the school initially resisted disclosure. We adjusted the framing to document what is confirmed — the request, the resistance, the lawsuit, the judge’s questions — without characterizing the outcome or intent beyond what the record supports.
The Human-AI Collaboration: What It Looked Like in Series 4
Identified the Milton Hershey Trust as Series 4 subject — based on its structural similarity to Series 3 but with a more specific victim class
Provided the actual 1909 Deed of Trust — the primary source that defined the entire investigation
Recognized immediately that "as many as possible" was the four words the investigation would live in
Set the directional thesis for each post
Made every editorial judgment: what to include, what to cut, what to flag
Posted every post. Received reader responses. The work is public.
Held the standard across 35 documents: show the receipts or don't make the claim
Recognized when something was truly surprising (the word "again")
Recognized when something needed to be held more carefully (the CHS framing)
WHAT CLAUDE DID IN SERIES 4:
Executed searches targeting court records, school websites, IRS filings, AG settlements
Found the 1999 sworn filing with “again” — In re Milton Hershey School Trust, 807 A.2d 324
Found the 2006 PA Supreme Court ruling that closed the enforcement door
Found the Start Strong PA 72% / 128,485 figure cited in CHS’s own documentation
Found the “source of pride” language on mhskids.org
Synthesized findings across 8 posts maintaining sourcing discipline throughout
Distinguished consistently between documented facts and inferences
Built the complete structural comparison across all four series
WHAT NEITHER OF US DID:
Started with a conclusion and worked backward
Made a claim without a source
Presented the trust as uniformly corrupt or the board as villains
Ignored evidence that complicated the argument
Overstated what the documents support
What 35 Posts Actually Produced
We want to be honest about this. Because the series ends with a finding, not a solution. The structure persists. The surplus grows. The four words have never been litigated. The children the deed names have no mechanism to enforce it. The AG is watching. The board is serving. The machine runs.
What 35 posts produced is documentation. Not change — documentation. The word "again" is now in 35 posts of public record. The $900 million rainy day fund is documented. The 2006 Supreme Court ruling that closed the enforcement door is documented. The IQ floor that isn't in the deed is documented. The 0.7% of the documented Pennsylvania need that CHS addresses is documented.
Documentation is not nothing. The mechanisms we've investigated across four series operate most efficiently in obscurity. The railroad barons needed the public not to understand the land grant system. The defense contractors need the public not to understand the revolving door. The university endowment machine needs the public not to understand the closed loop. The Hershey Trust needs the public not to understand what "again" means in a 1999 sworn court filing.
Now it's documented. Permanently. Anyone can find it. The AG can find it. Journalists can find it. Lawmakers can find it. Researchers can find it. Future investigators can find it.
The series ends when the mechanism does. The mechanism is still running. So the investigation continues.
THE LAND GRAB (Series 1): 8 posts + 3 capstones = 11 documents
THE ENDLESS FRONTIER (Series 2): 8 posts + 3 capstones = 11 documents
THE ENDOWMENT MACHINE (Series 3): 8 posts + 3 capstones = 11 documents
THE CHOCOLATE MACHINE (Series 4): 8 posts + 3 capstones = 11 documents
TOTAL: 44 documents
ESTIMATED WORDS: ~185,000
UNSOURCED CLAIMS: Zero
COUNTERARGUMENTS OMITTED: Zero
FINDING: One
THE FINDING:
The structure matters more than the intentions. Always.
Across 200 years. Across four series. Across 44 documents.
The same mechanism. The same script. The same result.
And the same question at the end of every investigation:
Now that you can see it — what do you do with that?
"Well, I have no children — that is, no heirs.
So I decided to make the orphan boys of the United States my heirs."
He meant all of them. As many as possible.
He gave away everything he had to make it real.
He signed a deed that said "must."
He meant it.
The structure he created to honor those words
has been accumulating the surplus those words should have spent
for 116 consecutive years.
"Again." One word. Sworn court testimony. January 20, 1999.
The most important word in this investigation.
And the trust's own president wrote it.
THE CHOCOLATE MACHINE is the fourth in an ongoing investigation into how public resources — and in this case, one man's private gift — generate private accumulation through institutional structures that claim public benefit status. The first three series investigated diffuse public resources. Series 4 investigated a named private gift to named beneficiaries, governed by a specific mandatory deed, with a documented 116-year gap between the mandate and the delivery.
The investigation continues because the mechanism continues. The series ends when the mechanism does. 35 posts. One finding. The structure matters more than the intentions. Always.
No comments:
Post a Comment