Home / evidence

Two findings, measured

Both come from live SEC EDGAR, a work of the U.S. government and not subject to copyright, so both are reproducible by anyone with an internet connection and no API key. Both are published because they are the kind of error that would otherwise never be visible to a reader.

7 of 15 fund-manager identifiers were wrong, and none of them threw

Every manager tracked by the Billionaire Consensus portfolio is checked against live SEC EDGAR by a script in the repository, and the report it produces is committed. All fifteen verify today. Seven did not, the first time the check was run — and six of those seven resolved cleanly to real 13F filers that were simply different firms. Not one would have raised an error. The portfolio would have tracked the wrong managers, published their holdings under the wrong names, and looked entirely healthy doing it, for years.

This is the quietest way this platform could mislead a reader, so it is published rather than fixed silently. The check now runs as a script, the corrected list is committed, and a test pins the wrong identifiers so they cannot return by copy-paste.

What the first run found — labelled versus what the identifier actually pointed at

Labelled in the mandate as	Actually was
Viking Global Investors LP	Third Point LLC
Third Point LLC	ValueAct Holdings LP
Soros Fund Management LLC	Southeastern Asset Management
Greenlight Capital Inc	Abrams Capital Management
Akre Capital Management LLC	Coatue Management LLC
Altarock Partners LLC	Ironsides Partners — which files no 13F at all
Lone Pine Capital LLC	did not resolve to anything

The list as it stands today — 15 of 15 verified

CIK	Name on EDGAR	Files 13F	Verified
0001067983	BERKSHIRE HATHAWAY INC	yes	ok
0001336528	Pershing Square Capital Management, L.P.	yes	ok
0001061768	BAUPOST GROUP LLC/MA	yes	ok
0001649339	Scion Asset Management, LLC	yes	ok
0001006438	APPALOOSA MANAGEMENT LP	yes	ok
0001167483	TIGER GLOBAL MANAGEMENT LLC	yes	ok
0001350694	Bridgewater Associates, LP	yes	ok
0001037389	RENAISSANCE TECHNOLOGIES LLC	yes	ok
0001061165	LONE PINE CAPITAL LLC	yes	ok
0001103804	VIKING GLOBAL INVESTORS LP	yes	ok
0001029160	SOROS FUND MANAGEMENT LLC	yes	ok
0001040273	Third Point LLC	yes	ok
0001079114	GREENLIGHT CAPITAL INC	yes	ok
0001631014	ALTAROCK PARTNERS LP	yes	ok
0001112520	AKRE CAPITAL MANAGEMENT LLC	yes	ok

Every identifier above is checked against data.sec.gov by a script in the repository, and the resulting report is committed. Changing the manager list without re-running that check is the failure mode; a test pins the identifiers that were wrong so they cannot return by copy-paste.

Companies restate their reported figures, a lot

For each filer and concept, how many reporting periods had their REPORTED VALUE change after first publication. Not how many were reported more than once — comparatives make that normal and uninteresting — but how many actually changed.

Filer	Net income periods revised	Revenue periods revised
GENERAL ELECTRIC CO CIK 0000040545	139 of 332	98 of 229
FORD MOTOR CO CIK 0000037996	41 of 203	18 of 141
WELLS FARGO & COMPANY/MN CIK 0000072971	35 of 175	8 of 43
MICROSOFT CORP CIK 0000789019	8 of 90	6 of 119
Apple Inc. CIK 0000320193	4 of 86	4 of 101
INTEL CORP CIK 0000050863	0 of 95	0 of 80

Keying a fundamentals table on (company, tag, period) silently overwrites the original figure the moment a restatement arrives. The original is the number that was visible on the day of the trade, and it is not recoverable from the API afterwards: the API returns the same superseded-and-current mixture. Every historical read on this platform therefore goes through a point-in-time view keyed on the filing date.

Reproduce it yourself: python scripts/build_evidence.py --ciks <cik>. No API key is required. A descriptive User-Agent with a contact address is mandatory; SEC returns HTTP 403 without one.

Measured 2026-08-06 against https://data.sec.gov/api/xbrl/companyconcept/ in 12 requests.

Why either of these matters to a reader

Neither error would have thrown. A portfolio tracking the wrong fund managers publishes their holdings under the wrong names and looks entirely healthy doing it. A fundamentals table keyed on the reporting period silently overwrites the original figure the moment a restatement arrives — and the original is the number that was visible on the day of the trade, which is the number a backtest has to use. Neither is recoverable afterwards.

Quiet errors of this shape are the only category that really threatens a site like this one. A crash is obvious and gets fixed. A portfolio that has been tracking the wrong managers for three years, with a clean audit chain and a plausible track record, is not.