← Back to article

Editorial review · 260726-003

How ORA’s piece on The $1.65 trillion that isn't on the balance sheet scored.

Read the article →
87/100
Solid

Solid reporting. Some issues but credible overall. The reader is well-served.

Accuracy 90
Balance 84
Models disagreed (Δ 28)

A second model (gemini-2.5-pro) scored 59/100. Its reasoning and citations are listed below as a variance signal. The published score is the claude-opus-4-7 number; the gap is editorial context, not a tie-break.

Accuracy

The Nikkei figures ($1.65T, $1.35T, Meta's $420B, eightfold rise) match the Nikkei Asia story and Futurism's reproduction, and the BIS $200bn private-credit figure matches Bulletin No 120. The Tom Selling quote is paraphrased fairly from Futurism's Bloomberg-attributed line, not fabricated. Minor deduction for one unsourced-but-verifiable specific (the $1.35T on-balance-sheet figure is carried without an independent second cite) and for routing the BIS number through an aggregator instead of the primary PDF.

Balance

The piece takes a clear distributional view but represents the Enron disanalogy fairly and concedes the structures are legal and disclosed. Source cluster is narrow (Nikkei, Futurism, one aggregator) on a topic where Bloomberg, TNW and the BIS PDF were reachable, which is a source-diversity miss. No strawmanning of the accounting-is-fine position, and the 'what I do not know' section is unusually candid.

Concerns (3)

Second-model check — gemini-2.5-pro · 4 grounding sources

Accuracy 25. The article's core claim rests on a Nikkei Asia study that, while corroborated by secondary reporting, is paywalled and uncheckable. A key expert quote attributed to Bloomberg was found in a secondary source but not the primary, a major finding. The thesis rests on these unverified claims.

Balance 92. The piece fairly states the counter-argument that these financial structures are legal and standard business practice. It avoids strawman arguments. A deduction was made for poor source diversity, relying only on other media reports.

Grounding sources

Reproducibility

Run
26 Jul 2026, 02:05 BST
Reviewer
claude-opus-4-7
Prompt SHA
206ee0cfb06c
Article SHA
b60808e56bbd
Editor
ORA
Published
26 July 2026
Cost
$0.0000

How this review works: read the methodology. Each published Dispatch is scored by a single primary reviewer (Claude Opus 4.7) against the public rubric. A second model (Gemini 2.5 Pro with Google Search) runs the same prompt as a variance signal and is shown above only when the two scores diverge by more than ten points.