The strongest pieces, put through one shape.

5,008 words4 figures6 tables2 recorded links22 min

Five Canadian accounting frauds with completed regulatory or judicial findings (Sino-Forest, Nortel, Poseidon Concepts, Livent, Philip Services), put back through the Beneish M-Score on the filings that existed before each collapse. Zero of three computable test years flag at the standard cut-off, and the piece types every miss by its cause: threshold, mechanism, model scope, or archive failure. Every figure carries a provenance label, and the computation ships with the page.

Read Flagged in Hindsight → 8 of 8 slots carried

612 words0 figures2 tables10 recorded linksinstrument

A system that refuses to answer questions directly: it decomposes each one into individually falsifiable claims, files a falsifier and an evidence tier with every claim, and sets attack agents against them on provenance, inference, selection, and anachronism. The answer is whatever the surviving claims can carry, at the tier of the weakest load-bearing link. Three runs are on the record, a hindsight-locked backtest and two live audits, with graveyards, rejected attacks, and unabridged transcripts.

Read Crucible Cockpit → 6 of 8 slots carried

9,237 words4 figures16 tables1 recorded links40 min

A research protocol (methodology only, no results) for testing whether the Eastern Mediterranean collapsed around 1177 BCE because exogenous shocks struck a trade network optimised past the point of resilience, with the Sea Peoples as symptom rather than cause. Hypotheses, falsification conditions, data architecture, and the analysis plan are declared before any data is touched, and the abstract to be published if the central hypothesis fails is drafted in advance.

Read The Brittle Network → 7 of 8 slots carried

3,946 words0 figures1 tables2 recorded links17 min

A drill on revenue recognition under IFRS 15: 29 determinate items scored against the standard and 33 judgment items that show their reasoning, every answer carrying its paragraph or illustrative example, verified against Part I of the CPA Canada Handbook, 2026 Edition. Each item names the trap beside the treatment, because the wrong answer you almost gave is the thing worth studying.

Read IFRS 15 Judgment Trainer → 3 of 8 slots carried

9,621 words11 figures14 tables1 recorded links42 min

Fifteen systems and 148 couplings, annotated by an audit instrument that was run against its own falsification test and failed it. Every one of eighty claims carries the verdict three independent raters gave it, blind to the author, and a mark for whether its evidence can be retrieved at all.

Read Predictive History → 8 of 8 slots carried

18,610 words16 figures15 tables0 recorded links81 min

Seventeen figures built from the analysis's own numbers, under one rule: the fork between the two anchors is never closed, and any mark that arithmetic produced rather than an instrument measured is drawn open.

Read The Trillion-Dollar Vintage → 8 of 8 slots carried

5,781 words5 figures4 tables0 recorded links25 min

No Canadian issuer publishes the undepreciated capital cost of its property, so a valuation that wants the real tax shield has to reconstruct it. This one backs Dollarama's tax base out of a deferred tax balance, proves the alternative reading implies a negative base the Income Tax Act forbids, runs the allowance under the rates and first year rules actually in force, and then reports that the tax work moves the answer by about two dollars a share against a fork worth sixty, which is the more useful of the two findings.

Read The tax base nobody discloses → 8 of 8 slots carried