The Verification Record
Ledger closed 10 August 2026 · Two independent passes, both agreeing · The full row-by-row ledger is printed in the book's appendix
What this page is
AI systems are known to invent citations. Books about AI are not exempt, and readers have no cheap way to tell a real bibliography from a confident-sounding fake.
So before this book went on sale, every source in it was opened and read twice. Once by me, checking that each source actually says the thing the book claims it says. Once independently, against the primary sources, without my notes. This page is the result, including the five places the checks caught something.
A corrections log produced after complaints is damage control. One produced before anyone paid a dollar is the book holding itself to its own standard.
The tally
| Check | Result |
|---|---|
| Bibliography entries | 35 (36 ledger rows; one law and the audit of its enforcement are listed separately so each carries its own link) |
| Opened to the supporting sentence by the author, pre-publication | 36 / 36 |
| Independent re-check against primary sources, 10 Aug 2026 | 36 / 36 |
| Sources with a resolvable link | 36 / 36 |
| Fabricated, misattributed, or non-existent sources | 0 |
| Composition | 27 peer-reviewed research papers · 4 regulatory instruments · 3 security disclosures · 2 legal cases |
What the checks caught: five entries carried an edit
A verification pass that finds nothing is indistinguishable from a verification pass that never ran. This one found five things. One was material.
| Entry | What was corrected |
|---|---|
| EU AI Act (material) | Regulation (EU) 2026/1744, adopted 8 July 2026, postponed the Annex III high-risk employment obligations from 2 August 2026 to 2 December 2027. The postponement landed nine days before the original deadline and 14 days before this ledger closed, so the book's statement that the obligations were applicable was corrected before sale. The paragraph flagged in the text as the fastest-dating in the manuscript is the one that dated. |
| CVE-2025-32711 (EchoLeak) | 9.3 Critical is the vendor CNA score; NVD independently reassessed at 7.5 High. The book now cites both, so a reader who clicks the NVD link does not conclude the figure was inflated. |
| NY State Comptroller audit | Precise citation added: report 2024-N-6, issued 2 December 2025, covering July 2023 through June 2025, 13 recommendations. |
| EEOC guidance | "Removed from the website" replaced "withdrawn": there was no formal Commission rescission and no Federal Register notice. Disparate-impact liability survives in statute at 42 U.S.C. § 2000e-2(k). |
| Two venue upgrades | Durmus et al. cited at its peer-reviewed venue (COLM 2024) rather than the preprint; Fatemi et al., "Test of Time," cited at ICLR 2025. |
Spot checks that held exactly as cited
| Source | Status |
|---|---|
| Mata v. Avianca, 678 F. Supp. 3d 443 (S.D.N.Y.), sanctions opinion 22 June 2023 | Exactly as cited |
| Moffatt v. Air Canada, 2024 BCCRT 149, decided 14 February 2024 | Exactly as cited |
| NYC Local Law 144, effect 1 January 2023, enforcement from 5 July 2023 | Correct, with the two dates kept distinct |
| Liu, Zhang & Liang, generative search citation support | The book's "about half" corresponds to the paper's 51.5% fully-supported figure |
| All 27 research papers | Confirmed at the venue and year cited |
Method
Pass one: the author opened every ledger row to the specific sentence in the source that supports the book's claim, before publication. Pass two, 10 August 2026: an independent re-check against primary sources, run without reference to the author's notes. Both passes agree. The row-by-row ledger, with links, is printed in the book's appendix, and the working file ships with the Operator Toolkit records.
Claims on this site carry the same discipline. Every empirical claim is labeled RESEARCH REPORTED ANALYSIS MODELED or OBSERVED, and a modeled figure is never presented as a measured one.
The published files, by hash
The exact bytes that shipped are a matter of public record. Any copy that does not descend from these masters is not the file that was sold. Buyer copies are individually watermarked and will intentionally not match; these values identify the masters.
| File | SHA-256 |
|---|---|
| AI_Keep_Your_Career_Bible_FIELD_EDITION.pdf (205pp) | 28114f63e5842e9d407d1cf6ae269ee4eae1ace985fa65c96c5a17d64b2d1595 |
| The_Operator_Toolkit_FINAL_v3.pdf (18pp) | dfc46be4c7167ae3c60ffb86476602f7d85920ccbe74ea7345a888f05cf43cad |
| Operator_Toolkit_Templates_v2.xlsx (13 sheets) | 3de659e89a0a0c850a296d85fcd23856959d934484ba5c3f8434bab1797216a9 |
| IY_Verification_Ledger_7.xlsx (3 sheets) | 0937530b606faf921c4b5d9c7344d21587238fe988115281598e8e72ef322a66 |
Check a file on macOS or Linux with shasum -a 256 <file>, or on Windows with Get-FileHash <file> -Algorithm SHA256.
The 24 documented AI failure modes, the control that catches each one, and the 90-day plan that ends with a number your manager can see. Every source above is in it.
Get the Book · $29