An entry doesn’t reconcile. The senior clerk has gone home. You are a junior auditor at Pennworth, Hargrove & Sloane, a Victorian-era counting house on Threadneedle Street, three days before the annual settlement. The brass-shaded oil lamps are still lit; the ledger books are open; the quill is in your hand. The cognitive shape is audit — noticing what should be there and isn’t.
The Inside the Firm chassis handles modern transactional finance — deals, models, term sheets. It cannot hold audit work because the cognitive shape is different. Audit demands the slow, methodical discipline of cross-referencing two sets of books under time pressure, the patience to trace a single transaction backwards through three ledgers, and the moral clarity to report what you find even when the senior partner would prefer you hadn’t. The Counting House carries that.
The firm is real-shape: Pennworth, Hargrove & Sloane, a mid-Victorian counting house handling insurance underwriting, property rents, and merchant accounts. The senior clerk, Mr. Threlfall, has been with the firm thirty-one years. Each chapter is a real moment of audit work — daily totals, missing receipts, cross-referencing, quarter-day reconciliation, insurance claims, capital movements, external audit, and year-end close — with the cognitive load shifting as the stakes compound.
Opening · daily reconciliation · 1.1 hr
Monday morning. The cashbook from Friday is open on the high desk, and the day’s column doesn’t balance. Seven shillings, fourpence. A small number — small enough to post to sundries and move on. But Mr. Threlfall left a note: “Find it.” The cognition is close reading under volume — forty-three entries, one error, and the discipline to trace rather than adjust.
Tracing · transaction reversal · 1.2 hr
A payment of fourteen pounds, six shillings to a coal merchant appears in the disbursements journal but has no corresponding receipt in the voucher file. The coal merchant says he was paid in cash. The cashbook says cheque. Somewhere between the journal, the voucher file, and the bank ledger, fourteen pounds vanished into the gap between what was recorded and what happened. The cognition is backwards tracing — starting from the absence and reconstructing the transaction.
Cross-reference · dual-book audit · 1.4 hr · ◆ FEATURED PLAYTHROUGH
Mr. Threlfall keeps a private ledger. He calls it his “memorandum book” — a running tally of entries he wants to verify before posting. You’ve been asked to reconcile it against the general ledger. Fourteen entries don’t match. Some are timing differences — he recorded them early. Some are rounding. But two are neither: they’re entries in Threlfall’s book that never appear in the general ledger at all. The cognition is pattern recognition across two data sources — distinguishing the innocent discrepancy from the material one.
Rent collection · reconciliation under pressure · 1.3 hr
Lady Day, March 25th. Rents are due from eleven properties. Three tenants have paid in advance, two are late, one disputes the amount, and one has paid the wrong landlord entirely. The rent roll, the receipt book, and the bank statement must agree by close of business — because Mr. Hargrove presents the quarterly summary to the partners tomorrow morning. The cognition is reconciliation under time pressure with incomplete information.
Valuation · inventory verification · 1.3 hr
Fire at the Bermondsey warehouse. The client, a tea merchant, claims losses of two hundred and forty pounds. The inventory ledger says he had sixty-four chests in store. The warehouse manifest says forty-one. The insurance policy covers “goods in store at time of loss.” Someone’s numbers are wrong — the question is whose, and whether the discrepancy is negligence or fraud. The cognition is forensic evaluation — reading documents against each other to find the lie.
Capital accounts · irregularities · 1.2 hr
Mr. Sloane has drawn forty pounds against his capital account — unremarkable, except the drawing predates the quarterly distribution by three weeks and no corresponding entry appears in the capital ledger until after the distribution date. The books show the withdrawal posted retroactively. You are a junior auditor, and the man whose books don’t add up is one of the three partners. The cognition is professional judgment under hierarchy — what you owe the books versus what you owe the firm.
External scrutiny · preparing the accounts · 1.3 hr
Mr. Cavendish of Price, Holyland & Waterhouse is here for the annual inspection. He is methodical, unhurried, and he has been through the firm’s books before. You have three hours to prepare the accounts for his review. Everything you’ve found over the past seven chapters — the missing receipt, the two phantom entries, Sloane’s retroactive drawing — is either in the books cleanly or it isn’t. The cognition is disclosure judgment — what to surface, what to annotate, what to let the auditor find himself.
Year-end close · final reconciliation · 1.2 hr
December 31st. The general ledger, the cash book, the rent roll, the capital accounts, and the bank statement must agree. Every discrepancy you traced, every receipt you found, every judgment you made flows into this final page. The balance either reconciles or it doesn’t. There is no arguing with arithmetic. The cognition is synthesis at apex — the entire year compressed into one clean, defensible set of accounts.
Litigation has adversaries; deal-making has counterparties. Audit has the ledger — and the ledger doesn’t argue back, but it doesn’t lie either. The cognitive shape of noticing what should be there and isn’t cannot be carried by an engine designed for argument or negotiation.
Double-entry bookkeeping means every transaction appears twice — debit and credit, journal and ledger, source and summary. The engine presents both sides and measures whether the player checks. An engine designed for argument doesn’t have this structure; the Ledger engine is built around the expectation that verification is the primary cognitive act.
Most engines measure what the player does. The Ledger engine measures what the player notices is absent. A missing receipt, a phantom entry, a gap in the sequence — the skill being measured is the discipline to spot a hole in a wall of numbers. This is the signature of audit cognition and it requires an engine that can score the detection of absence, not just the selection of action.
Audit is never purely technical. Every discrepancy implicates a person — the clerk who posted it, the partner who authorized it, the client who reported it. The Ledger engine tracks the player’s professional-judgment decisions alongside their quantitative ones. Forensic accounting is arithmetic plus ethics; the engine must score both.
The Ledger engine uses Item Response Theory to calibrate discrepancy difficulty. A seven-shilling rounding error in Chapter I is a warm-up item; a retroactive capital-account entry in Chapter VI is a high-difficulty item. The engine adjusts the next discrepancy’s complexity based on the player’s demonstrated ability — measuring the boundary of competence, not just confirming what the player already knows.
Across the eight chapters, weighted by chapter hours. The Counting House’s signature is D2 + D6 — quantitative reasoning plus domain knowledge — with strong D4 (logic/rule-application) and D5 (judgment under uncertainty). D7 emerges heavily in the final two chapters as accumulated findings must be synthesized into a defensible set of accounts.
Tuesday evening. The gas lamps have been turned up. Mr. Threlfall left at four o’clock — early, for him — and his memorandum book is on his desk, open to this month’s page. Mr. Hargrove has asked you to reconcile it against the general ledger. Fourteen entries between the two books. Most will match. The question is what you do about the ones that don’t.
| Date | Description | Memo. Book | Gen. Ledger | Variance |
|---|---|---|---|---|
| Oct 3 | Mercer & Co. rent | £12. 0s. 0d. | £12. 0s. 0d. | — |
| Oct 7 | Briggs coal delivery | £14. 6s. 0d. | £14. 6s. 0d. | — |
| Oct 11 | Whitmore policy prem. | £8. 15s. 0d. | £8. 5s. 0d. | 10s. 0d. |
| Oct 14 | Stationery — Hatchards | £1. 2s. 6d. | £1. 2s. 6d. | — |
| Oct 19 | Cuthbert warehouse fee | £6. 10s. 0d. | — | £6. 10s. 0d. |
| Oct 22 | Hargrove — drawings | — | £25. 0s. 0d. | £25. 0s. 0d. |
Six of fourteen entries shown. Oct 11: ten-shilling variance — likely a transposition. Oct 19: entry in Threlfall’s book, absent from the general ledger entirely. Oct 22: entry in the general ledger, absent from Threlfall’s book entirely.
D2 + D6 reading: the choice between A, B, and C reveals the player’s audit philosophy. A is textbook-correct and credits D6 (domain knowledge of proper audit procedure); B credits D5 (materiality judgment); C credits D2 (quantitative pattern-seeking). The model reads which option the player picks against their Chapter I behavior — a player who found the seven-shilling error methodically and now picks (C) is evolving their approach; a player who guessed in Ch. I and picks (C) here is consistently impatient.
This is the professional-judgment decision point the Ledger engine is built around. All three are defensible; the differentiator is the player’s read of their position. (A) is safe but may look negligent if the irregularity turns out to be real; (C) is courageous but may damage working relationships if the explanation is innocent. The engine doesn’t penalize (A) or (C) absolutely — it reads the player’s choices across all chapters and credits consistency of professional posture.
This is the D4 forensic-logic beat — the mirror of the Cuthbert entry, now involving a partner. (A) credits D4 (consistent application of the same analytical framework regardless of who is involved). (C) credits D2 and D6 (the instinct to verify one more data source before concluding). (B) credits D5 (professional courage) but the engine tracks whether the player has earned the right to escalate — a player who picked (A) on the Cuthbert entry and now picks (B) on the partner’s entry is applying a double standard, which the engine detects and penalizes on D4.
The chapter has 28 measurement events, of which three are shown above. The reconciliation ends with your report placed on Hargrove’s desk. Whether it contains fourteen notes or two flagged irregularities depends on the choices. The morning will tell you whether the books agree — and whether Threlfall’s thirty-one years of clean record survive what you found.
Eight chapters. Ten hours. The discipline of noticing what should be there and isn’t. The Counting House is waiting.
Begin the audit →