[ 00 / LATEST WRITING ]
2026-08-08A Green Light That Proves NothingA check that has never rejected anything has told you nothing. It might be a correct rule that has simply never been violated. It might be a rule that…2026-08-06The Aggregator Is a Quarter Behind, and It Doesn't Say SoThe question that started this was mundane: why don't we have last year's figures for a company we cover?2026-08-04Forty Days of Looking HealthyOur engine kept marking companies. Our scanner kept ingesting filings. Our public case count stayed comfortably non-zero and the site returned 200 to …2026-08-02Visibility Only Goes One WayWe shipped a page that showed anonymous visitors the full list of gray flags. Subscribers paid for that list. It was readable without logging in.2026-07-31The Ledger Refused Its AuthorWe tell subscribers that our records are never rewritten. The interesting question is not whether we mean it. It is what happens the first time the pe…2026-07-30We Publish the DisagreementThree annotation experts. None sees the others' work. If they disagree — the disagreement goes into the record.2026-07-28Three Blind Readers and a State MachineThree annotation experts work each historical case. None sees the others' work. If they disagree, the disagreement goes into the record.2026-07-26Field Evidence, Not Field AnalysisWhen a desk investigation reaches the limit of what filings can tell you, the next step is the physical world. Somebody goes and looks. A facility eit…2026-07-24What May Enter the CorpusOur reasoning engine learns from one thing: signed annotations written by credentialed forensic accountants on historical filings. We have written abo…2026-07-23The Key Was Not the VerdictFor a stretch of time our public case pages told institutional readers that certain detection dimensions had fired on certain companies. On some of th…2026-07-21The Graph That Learned the Wrong ThingWe spent a stretch of engineering time building a graph neural network over corporate relationships, ran it properly, and archived it. We have written…2026-07-19Counting One Witness Four TimesThe previous post ended on an admission: the running-product construction we use to accumulate evidence is valid when the pieces are independent, and …2026-07-18Evidence That Survives Being WatchedA scanner has a property that breaks classical statistics, and it breaks it quietly enough that you can ship for a year before noticing.2026-07-16Two Kinds of Evidence, and Only One of Them Can Raise a FlagWe have written before about why we look for contradictions rather than outliers(/writing/contradictions-not-anomalies). That piece made the case in t…2026-07-15The Record Is the ProductA red flag on this platform is not an opinion. It is the output of a trained machine — and the thing we actually sell is the record of how that machin…2026-07-14Accuracy Is Not a Number. It's a Discipline.If you ask most teams how accurate their system is, they will quote you a metric. A precision figure, a recall figure, an area under some curve. We ha…2026-07-11Judgment Belongs to Experience: Why the Engine Only Produces LeadsThere is a structural decision at the center of how we build, and it is easy to get wrong because the wrong version is so tempting. The question is: w…2026-07-07Who Is Qualified to Teach the MachineThere is a seductive shortcut available to anyone building a reasoning system today. You need training examples — cases where an expert examined a sit…2026-07-04The Discipline of Time: Why a Judgment Must Live in the Moment It Was MadeThere is a quiet way to lie to yourself when building a system that learns from history, and almost everyone does it at least once. You take an event …2026-07-02The Scale Trap: When 'Big' Masquerades as a SignalThere is a failure mode in financial machine learning that is so common, and so well-disguised, that we've come to treat encountering it as a signal i…2026-06-30Grey Flags and Red Flags: Designing a Two-Layer Judgment SystemWe've argued that the right unit of detection is a contradiction inside a company's own disclosures, not a statistical outlier. That commitment raises…2026-06-28Why We Look for Contradictions, Not AnomaliesMost quantitative approaches to fraud detection start from the same instinct: find the outliers. Rank every company on some metric — accruals, revenue…2026-06-26The Value of Negation: How to Rigorously Abandon a MethodMost engineering writing is about things that worked. This one is about deciding that something didn't, and treating that decision as a real result ra…2026-06-26We're Looking for People Who Can Tell a Real Signal from a Plausible OneA financial contradiction is easy to detect and hard to judge. A company reports cumulative profit alongside negative operating cash flow; a related-p…2026-06-23Reading list: the forensic accounting canonYou cannot detect what you don't understand, and you cannot understand financial misconduct by intuition alone. The people who pull it apart for a liv…2026-06-20Anatomy of a health checkThe most dangerous data problem is not the one that crashes the pipeline. It's the one that doesn't — the silent drift that lets bad numbers flow thro…2026-06-18Reading XBRL like a forensic accountantA company never lies in one place. It lies in the gaps between places — and XBRL is where those gaps live in plain sight, if you know the dialect.2026-06-14Tearing out the vote: an architecture post-mortemWe've written about the principle: the engine produces clues, never verdicts(/methodology/feature-factory). This is the story of how we learned it — b…2026-06-13The 33x lesson: bulk loading at filing scaleSome lessons you read about and nod. Others cost you a night. This one cost us a night, and it's the reason a one-line rule now sits in the documents …