Prepare

One cockpit, and what each tab is

The tab is in the address, so a link you send opens on the view you meant. Two of these used to have rows of their own beside the cockpit that already carried them — a surface reachable twice is a surface a reader has to work out is one place.

• Cockpit — the preparation checklist: four activities, each with a state word, and no composite anywhere.
• Posture — your recorded answers across the examination areas, and the gaps the readiness library found in them.
• Mock Examiner — what has actually been run against this institution, and the way into a session.
• Exam Binder — the examiner handoff document.
• Procedures — this one does not render a register of its own: it takes you straight to the examination procedure index in one click, because a page whose only content was a button to the real index was two clicks pretending to be a surface.

The checklist, and the words it is allowed to use

Three of the four steps ask "is there a record of you having done this", which none of the five threshold words answers. So they carry their own vocabulary — recorded, in progress, not started, not evaluated — where not evaluatedmeans what it means everywhere else: nothing ran, so the absence of a finding is not a finding. The one step that does sit against published lines, the engine's open findings, carries the engine's own words.

Progress is stated as steps complete out of steps total. There is no percentage ready, no percentage compliant and no "on track" — two composites were removed for being arithmetic over things that do not add up, and a binder either has been generated or has not. The liability statement above the checklist is not dismissible: completion reflects your documented preparation activity, not a determination of regulatory compliance or of an examination outcome.

Where a step reports nothing, it is because nothing ran. An institution the engine has never been run against does not read "no findings pending review": no gaps and no evaluation are different states and the cockpit distinguishes them.

The countdown to your examination has exactly one source — the examination date you typed into the binder form. It says so, rather than implying it knows your schedule.

The posture record

Posture holds 12 examination areas across 126 fields — BSA/AML, IT and cybersecurity, third-party risk, flood insurance, call report thresholds, interest rate risk, liquidity risk, safety and soundness, management, the Community Reinvestment Act, fair lending, and consumer compliance. They differ in size by more than fifty to one, which is why the register counts fields and areas separately and never reports a percentage.

These are notthe same slicing as the rule book's categories, and the two must never be reconciled. Posture slices by which examination programme the documentation belongs to; the rule book slices by what kind of number a rule measures. Both trace back to the same canonical taxonomy — they are different cuts of it.

An area can carry one of six words:

• Breach — a critical gap against the readiness library.
• Approaching — gaps, none of them critical.
• Passed — evaluated, and no gaps found.
• Not evaluated — no readiness evaluation has been run at all.
• Not covered — an evaluation ran, but no rule in the readiness library reads this area.
• Not started — no fields have been filled in.

The gap register sorts by severity in the ledger's own vocabulary — critical, material, minor, informational — and each gap carries the rule that produced it, its rationale, a closes when remediation sentence, its citation, and links straight to the fields behind it and to the procedures for that area. Open an area and you get its fields with the flagged ones marked; the header names the examination book, the canonical area and the raw storage prefix, because a field path in a trail is only readable if you know which area it belongs to.

Every edit is appended to the unified event stream. A correction is a new event naming the one it corrects; nothing is overwritten. See Dashboard for the trail and how to read it. The two posture areas that are unambiguously about one analytical domain — liquidity risk, and interest rate risk — also render beside the numbers on your own institution view.

The engine cockpit behind the findings step

The checklist summarises; the engine cockpit is the deep surface, and it is where you go when a finding raises a question the summary cannot answer. It runs when you open it — there is no button to press first. 50 rules across 12 categories evaluate against your own filing, and a further 5 loan-level rules run when you upload a portfolio, for 55 in the book. Of the institution-level set, 36 read filed figures and 14 read attestations you recorded.

• Every rule, in every state — including the ones that passed, the ones that could not be evaluated because the filing did not carry an item the rule needs (the row names the missing items, with their codes and schedules), and the ones that do not apply to this charter or election. Filter the run by any of them and read the counts before opening anything.
• The validation record behind each rule — the required label riding the collapsed row: not yet validated, provisional, self-attested, backtest did not support, recorded fact. A rule the validation programme will not let assert a verdict cannot render as an alarm anywhere in the product; the surface enforces that as a ceiling. Citations are honest on two separate axes, because they fail in two separate ways: whether the document resolves, and whether the cited document actually publishes the number. Where it does not, the basis says screening line.
• Recording a decision — open a finding and record it reviewed, resolved or dismissed, appended to a trail with your note. A decision never deletes the finding; that is the point of it being a record.
• Your numbers, beside the filing — the update lane lives here, and the Dashboard links straight at it.

Nothing in the cockpit writes to the ledger. Findings become obligations through a projection that runs on its own schedule and reaches a human as a next step on the Dashboard.

What a run turns into

• The examination binder — two generators, each publishing an honest register of the sections it actually prints, the feed behind each section, and the condition under which it appears. The certification page is blank by design: a human signs it, and nothing is attested on your behalf.
• The compliance calendar — your own register of dated compliance tasks, plus the remediation deadlines under your logged examination findings. Nothing here is measured from a filing and nothing here is a score; overdue is derived from the date, not from a stored status.
• The CAMELS self-assessment — a six-component workbook. Each step shows what the filing says and what the engine found, then asks for your rating and your narrative. It reports the arithmetic mean of the ratings you entered, labelled as exactly that. There is no composite, and it does not estimate the rating a regulator would assign.

The mock examiner

Proctor walks the same procedures a real examiner follows, taken from the FFIEC, OCC and FDIC manuals the area cites, one question at a time, and records what you answered. Nothing in a session scores the institution or estimates a finding. 10areas are available, and the cockpit's tab reports how many of them you have covered — against sessions started, which is deliberately not sessions finished and is not a score.

• Lite audit — top-level procedures only. The fastest pass that still surfaces the critical items, with a checklist as its output.
• Full exam — the complete hierarchical walkthrough with branching follow-ups, a gap inventory with remediation paths, and the posture delta, ending in a four-section report and a PDF.

Each mode card states its own retention plainly: whether the session is recorded in the audit trail and kept as institutional history, or shown in-session only and gone when you close the tab. Areas are grouped by examination book from the canonical taxonomy — which is why the picker cannot disagree with the procedure index or with your posture record — and not every registered area is offered, which is enforced in the registry rather than by a note somebody has to remember.

Selecting an area shows how complete your posture is for that area; a reading measured for a different area is not shown at all, because it would not be true of the one you picked. Each answer button says what it records, you can go back, and an unfinished session is either picked up where you left it or abandoned outright — both shown as their own state rather than as an error. On completion you get a register of what the session recorded, then the report.

Procedure libraries are updated. If a library moved after your last completed session in an area, the landing page tells you, and the report carries a header state saying it was read against a superseded library. A stale report that does not say it is stale is worse than no report. Sessions land in the unified event stream and are deep-linkable from the audit trail.

The examination procedures themselves

The procedure index is a reference view: the questions an examiner works from, area by area, each carrying the published document it came from. Nothing there starts a session or records an answer. Areas are grouped by examination book, and each row shows the area, how many procedures it holds, its wave and mode, and the agencies its procedures came from.

Three availability states, and they differ for a reason:

• Regulatory library — sourced from the ingested FFIEC and FDIC regulatory library, which takes precedence where it exists.
• Authored set — the authored mock examiner procedure set, the same procedures a session walks through.
• Not loaded — no procedures are loaded for this area in this environment. It is not a link, and it does not promise anything is coming: nothing failed, there is simply nothing to browse yet.

An area page opens on a register — how many top-level procedures, how many follow-ups, how many a lite session asks, how many map to a posture field, and how many citations resolve to a document you can actually open. The rest name their source without a link, and the register says so rather than implying every citation is clickable. Then the procedures themselves: sequence, severity, the question, its citation and that citation's completeness, what to do if it is a gap, the remediation path, and its posture-field mapping — or the plain statement that no posture field maps to it, in which case a session asks it every time. Areas that are prep sessions rather than full simulations carry a scope notice saying what is deliberately out of scope.

A first pass, in order

1. Open Prepare and read the checklist. It tells you which of the four activities has a record and which has never run.
2. Open the posture tab and fill in the area your next examination is about. Read the coverage line first — it says how much of what you enter is actually evaluated today.
3. Run the engine, then open the cockpit for the findings and their validation records.
4. Run a mock examiner session on the area you are least confident about, reading its procedures first.
5. Generate the binder last: it is built from the run and the posture record you have just brought up to date.