Skip to content

Content methodology

What is sourced, what is generated, and what still needs review.

This page describes the workflow implemented in the repository today. It is not a claim that every topic is complete or has a named technical reviewer.

Current workflow

From content record to maintained page

Independent technical review pending

1. Define the public record

A public topic or explainer is a source-controlled record with an explicit publication state. Public explainers live separately from the protected practice bank and carry no private question ID.

2. Attach sources to claims

Exam-owner facts and technical concepts list source IDs that resolve to visible source records. Product status and our own study advice are labeled separately from official exam facts.

3. Show the real review state

We check every page against its cited sources ourselves. Practice questions go live only after a review the owner runs and directs, with an automated pass that re-checks every answer key and the reasoning attached to each wrong option. Nobody outside Keiko has read any of it, so you will never see a reviewer name, a credential, or a review badge here.

4. Recheck and correct

Each content record has a last source-check date, a next review date, and maintenance notes. Correction reports carry the page path into the server-handled contact workflow.

Readiness evidence forecast

Observed evidence, not a predicted exam score

Version evidence-forecast-v1 uses only the learner's active exam and the published learning-model version. It reports skill observations, unique-question breadth, recency, explicit confidence when supplied, review retention, exact versus generic error recurrence, and timed-mock pacing and consistency. Missing evidence stays unavailable.

Domain bands are 80% Wilson intervals around the saved correct share. They describe the observed sample; they are not prediction intervals. Evidence is Insufficient below five skill observations or three unique questions, Early from that minimum, Useful only with 25 observations, 15 unique questions, three sampled skills, and a mock, and Substantial only with 100 observations, 50 unique questions, 60% skill coverage, two mocks, and five review observations.

Keiko does not publish an official-score range, percentile, mastery probability, or outcome prediction. A later calibrated model would require paired official outcomes, held-out validation, interval-coverage checks, drift monitoring, and a new explicit model version.

AI tutor boundary

Tutor responses are generated on request and may be incorrect. They are not publisher-reviewed study pages, official exam guidance, or a substitute for the cited source material. The workspace labels generated output and lets the learner choose which current context is sent.

Not official exam material

Keiko Study is independent and is not affiliated with or endorsed by an exam owner. Source links support stated facts; they do not turn our notes or original public questions into official test material.

Current source register

These are the source records currently used by published content. The license note describes how each record may be referenced; it is not a blanket license for exam-owner material.

Corrections and update policy

Use the correction form to identify the page and explain the issue. The server validates the public page path, applies length limits and a honeypot check, then stores the report in the server-only contact inbox when configured; an owner notification is best-effort. A report does not silently change published content. The source record and page are rechecked before an edit is shipped.

Report an issue with this pageGeneral contact