Receipts
Authority should be earned from a dated record, not asserted. This ledger is append-only: each entry records a public claim or reading, the date it was published, and its outcome as the evidence arrives — including the times we were wrong, which stay on the page. It is the mirror of the falsifier ledger: there we publish what would disprove us; here we log how it actually went. The desk is young; the ledger is short. It only grows.
2026-06-25
The market-vs-ground-truth gauge inverted hard in 2026 Q2 — M(t) +2.83 against G(t) −1.23 — and we published it with its caveats (n=4, descriptive) rather than waiting for confirmation. The reading stands on the Tape; it resolves as the ground-truth series confirm or reverse.
OPEN — resolves with the data2026-07-01
Our circular-financing work argues the build-out's capacity ran ahead of external demand. On July 1, Bloomberg reported Meta is building a cloud business to sell excess AI compute — capacity hunting demand after the fact, the pattern the recycling ratio measures and the telecom chapter documents ("dark fiber"). One data point, not a verdict; logged because it was publishable before the news, in the paper.
CONSISTENT — one observation2026-07-01
Two pages carried non-reconciling versions of the Amazon useful-life canary. We unified to the filed figures (~$1.4B run-rate step-up + $920M one-time, FY2025 10-K Note 1) and recorded the revision, dated, in the falsifier ledger. Mistakes get fixed in public here, not silently.
CORRECTED — self-reported2026-07-02
Walk the Loop (54 pp) and The Recycling Ratio (12 pp) went live with every financing-edge accession independently re-verified against SEC EDGAR — 15 of 15 — and the falsifiers stated in advance. The claims in those papers are now this ledger's obligations.
ON THE RECORD2026-07-02
The Earnings Desk published its per-name watch list before the prints: exactly which line items we pull from each 10-K/10-Q, July 28–30. Whether we deliver is checkable against this dated entry.
OPEN — resolves Jul 28–302026-07-02
We re-read every load-bearing figure in the papers against the filing text itself. Eight confirmed verbatim. Three corrections, applied across every surface today. First, our ledger had missed Amazon's Q1 2026 investment of $15.0B in OpenAI Series C preferred stock plus a $35.0B commitment letter (10-Q, accession 0001018724-26-000014). Adding the funded $15B widens the outside-equity base to ~$35B and compresses the funded-cash recycling ratio from 26× to 15.5× — while structurally tightening the circle: both top clouds now hold equity in the lab committed to their own datacenters. This is not the arm's-length capital our falsifier asks for, and we say so. Second, Microsoft's OpenAI paper gain is $5.9B of investment gains (primarily the recapitalization dilution gain), not $4.5B — the $4.5B we carried was the same filing's after-tax net-income impact, mislabeled. Mark-to-model total: $16.8B → $18.2B. Third, Amazon's $920M one-time charge is real but sits in the FY2024 10-K (Q4 2024), not the FY2025 10-K we cited. Papers re-rendered, toolkit regression tests updated, all site surfaces swept the same day. A correction that cuts our own headline number is the system working.
CORRECTED — self-reported, same-dayAppend-only. Entries are never edited after the fact; corrections get new entries. Where a claim resolves against us, the entry says so and stays.