On this page
THE SHORT ANSWER
Keep the original question and interpretation before recording actions, new information and anything that did not fit. Set your own review date and add a new dated paragraph afterward. Do not rewrite the first interpretation to make it match the outcome.
- Separate what you believed then from what you learned later, including what was unknown.
- Track observable actions and allow “unhelpful,” “unsuitable” or “not yet known” as outcomes.
- Export journals and Agent conversations separately. Set reminders yourself; Wenbu does not follow up on a future date.
Should I join an eight-week course? I know the fee, not the workload.
Try a lesson before committing; enrollment is still undecided.
Try a free lesson on Saturday and record lesson and exercise time.
Lesson: 45 minutes. Exercise: 90 minutes. I have two hours a week.
The cards select the original question, initial interpretation, action and date, and new information, each with a fictional course example. The complete table below also includes what did not fit and a revised decision, with checks for every field. The sample times are not observations from a user.
Keep the question you actually asked
Suppose you asked whether to enroll in a course. Save the original result and note what concerned you at the time: the fee, the time commitment or whether you would keep going. Keep more than the sentence you found reassuring.
Save standalone tool results to the journal to add notes; find Agent conversations and reports in the conversation list. Both support browser storage, optional account history and separate exports.
Turn an association into something you can try
Perhaps the reading suggests trying a routine before making a commitment. You could write: “On Saturday I will try a free lesson, note how long it takes and then reconsider the course.” That is something you can schedule and check later.
If the advice is only “trust yourself” or “go with the flow,” ask for one specific action that fits your circumstances this week. The action should respect your time, budget and existing commitments.
A review template you can copy into a note
| Field | Course example | What to check |
|---|---|---|
| Original question | Should I join an eight-week course? I know the fee, not the workload. | Keep the question you actually asked. |
| Initial interpretation | Try a lesson before committing; enrollment is still undecided. | Separate your association from the report’s wording. |
| Action and date | Try a free lesson on Saturday and record lesson and exercise time. | Record what happened, including an action not taken. |
| New information | Lesson: 45 minutes. Exercise: 90 minutes. I have two hours a week. | These are fictional figures; record your own observations. |
| What did not fit | There is no new evidence for the claim about low confidence. | Keep unhelpful statements as well as apparent matches. |
| Revised decision | Delay enrollment and try two weeks of shorter exercises. | Identify the information that changed your decision. |
On a small screen, scroll the table sideways.
Compare what happened with what you expected
On your chosen review date, read the original question. Perhaps the trial lesson took longer than expected and you decided not to enroll. Record how that new information shaped your decision, rather than treating either outcome as something the cards had predicted.
It is also fine to write: “This reading did not give me a useful lead.” Keeping an unhelpful reading is part of an honest record. You do not need to find a successful ending for every session.
- Initial interpretation: what did I think the reading meant?
- Action: what did I actually do, and what did I leave undone?
- New information: what did I learn, and which assumptions were wrong?
- Next decision: continue, adjust or stop?
Keep a copy you can find again
Guest journal entries stay in the current browser. Sign in by email to save to your account and continue elsewhere. Older records are uploaded only when selected. Clearing site data removes browser records that have not been saved to your account. Export important entries as a backup.
An export may contain birth details and private questions. Read it before sharing it. When bringing context into an Agent conversation, select only what is relevant. To begin, add a next action and a review date to one recent entry.
Worked example: returning a week later with a different understanding
This fictional course example demonstrates a record-keeping practice. It is not a validated Wenbu intervention or a test of predictive accuracy.
Save the tool result and write: “September 30: I think I am worried about sticking with it. I will try a lesson; I have not enrolled.” Export a separate Agent discussion if you want that too.
Set a calendar reminder. Decide what to observe: whether you completed the lesson, total time and continued interest. “Did it feel accurate?” is not enough.
Append: “October 7: I completed the lesson and remain interested, but the exercises took longer than expected.” Keep the earlier note rather than changing it to imply you always knew.
Consider another explanation: the course pace may conflict with your schedule. Distinguish information from the trial from associations prompted by the reading.
Record your next decision and export a backup. If you never tried the lesson, record why; do not evaluate its usefulness using an outcome you did not observe.
Fischhoff’s original research showed that knowing outcomes can change judgments of their earlier predictability. That motivates this practice; it does not prove journaling removes bias.
Key terms
- Hindsight bias
- Seeing an outcome as more foreseeable after learning what happened.
- Alternative explanation
- Another possible reason for an outcome, including your own actions.
- Review record
- A dated comparison of the initial view and newly observed facts.
Common questions
Does the journal keep note versions?
No. It has an editable note field with a length limit. Append dated paragraphs and export before important edits if you need a record of changes.
Is a review useful if I took no action?
Yes. Record what prevented it or changed your mind, without treating nonexecution as a direct test of the advice.
Does local deletion erase model-provider data?
No. Removing a guest record deletes its browser copy. Account history has separate deletion controls. AI processing, submitted feedback and analytics each follow their own policies.
Sources and their scope
Supports note editing, its length limit, removal/undo and JSON export; there is no note version history.
Supports separate handling of local records, model processing, feedback and analytics, and limits on sync and background work.
The original experiments concern how outcome knowledge changes judgment. The course-journal exercise is editorial guidance, not an intervention tested in the paper.
Further reading; not validation of real-world predictions.
Traditional interpretations are not established predictions. See the source notes for context.
Read this guide with your agent
Get the complete text, diagram descriptions and sources in an open format. These files contain public knowledge only.
ADDITIONAL PRACTICE · 2026-11-01
Two weeks later: mark each interpretation supported, disconfirmed or not yet testable
Two weeks later, take each interpretation on its own and give it exactly one status: supported when a specific dated event matches it, disconfirmed when a specific dated event runs against it, or not yet testable when the fortnight offered no chance to check it. Absence of evidence is not disconfirmation; it belongs under not yet testable. Completing a trial lesson and recording its length is an action record; saying you remain interested is a first-person report; neither by itself confirms an interpretation.
Wenbu · reviewing a reading ↗Wenbu · asking a better question ↗Wenbu · versioned public capabilities ↗
- Give each interpretation one status: supported, disconfirmed or not yet testable. No evidence is not the same as disconfirmed.
- Action completion, first-person reports and confirmation of an interpretation are three separate columns and cannot substitute for one another.
- A random draw cannot be reproduced from the same question; compare against the saved original result.
One interpretation, three possible statuses
- 01Copy the original line
Reproduce the sentence as written, with its date, and change no wording.
- 02Look for a specific event
Find only events from the fortnight that are dated and checkable by someone else; if there are none, move on.
- 03Judge the direction
A matching event means supported, a contrary event means disconfirmed, and no chance to check means not yet testable.
- 04Note the evidence
Record which day and record the judgment rests on, and keep actions and self-reports in separate columns.
Wenbu · reviewing a reading ↗Wenbu · asking a better question ↗
Why “no evidence” is not “disconfirmed”
Disconfirmed requires a specific, dated event that runs against the interpretation. Suppose the original interpretation was that low confidence was the main obstacle, and within the fortnight you recorded a specific occasion: you began with real confidence and still stopped because time ran short. That is a contrary event you can write down. If the fortnight gave no chance to check the interpretation at all, the correct status is not yet testable, not refuted.
Writing “no evidence” as “disconfirmed” makes a review look more conclusive than it is and hides the information that is actually missing. The worksheet therefore gives each interpretation one status and notes which day and which record the judgment rests on, so it can be checked later.
Wenbu · reviewing a reading ↗Wenbu · asking a better question ↗
Keep actions, self-reports and confirmation in three columns
Completing a trial lesson and recording how long the lesson and exercises took is an action record; someone else could check it against your calendar and notes. Saying you remain interested is a first-person report: it describes how you felt at the time and is not an externally checkable fact.
Whether an interpretation was confirmed is a third matter. Even a smoothly completed action shows only that you carried out the suggestion, not that an interpretation holds. Keeping the three columns apart prevents “I followed it” from standing in for “this interpretation was validated.”
Wenbu · asking a better question ↗Wenbu · versioned public capabilities ↗
Add the status mark to the existing template
There is no need for a second worksheet: add a status column beside the existing review fields, fill in supported, disconfirmed or not yet testable for each interpretation, and write one line of evidence. The journal note is an editable field without version history, so export a backup before important edits; Agent conversations and reports are stored elsewhere and need a separate export.
Set the review date in your own calendar; the site will not prompt you on that date. For random draws, compare against the saved original result; drawing again for the same question is a new record and cannot replace the old one.
Wenbu · reviewing a reading ↗Wenbu · versioned public capabilities ↗
| Original interpretation | Event within two weeks | Status and note |
|---|---|---|
| “Trying a lesson before deciding” is the right approach | Oct 7: trial lesson completed; it supplied the information needed to decide, and you chose not to enroll yet | Supported: the event directly answers the purpose of the suggestion; figures and outcome are fictional |
| Low confidence is the main obstacle | Oct 9: recorded starting with real confidence and still stopping because time ran short | Disconfirmed: a specific contrary event exists; keep the original line |
| The long-term course pace suits me | No chance to check the long-term pace within two weeks | Not yet testable: neither a hit nor a refutation |
| I still want to learn more | Oct 7: I reported continued interest | First-person report: not an externally checkable fact and does not confirm any interpretation |
Wenbu · reviewing a reading ↗Wenbu · asking a better question ↗
Worked example: two weeks later, one supported, one disconfirmed, one not yet testable
This fictional course scenario demonstrates status judgments. It is not a real user record and is not used to compute any accuracy rate. Suppose a reading on September 30 produced three interpretations.
- September 30: save the original result, write the question and the three interpretations in the note, add the date and export one backup.
- Set the review date for October 14 in your own calendar and define what to observe first: whether the trial lesson was completed, whether it helped you decide, how long lesson and exercises took, and the hours available that week.
- October 7: append the action record — trial lesson completed, 45 minutes of lesson and 90 minutes of exercises, two hours available that week — plus one first-person line, “I am still interested,” marked as a self-report.
- October 14: judge each line. “Try before deciding” is supported because the trial supplied the information needed and you chose not to enroll yet; “low confidence is the main obstacle” is disconfirmed because on October 9 you recorded real confidence and still stopped for lack of time; “the long-term pace suits me” is not yet testable because the fortnight offered no chance to check it.
- Write the output sentence: “What changed my decision was that exercise time exceeded the hours available, so I will delay enrollment and try two weeks of shorter practice.” Keep the September 30 lines rather than editing them to say you always knew.
The output is three status-marked judgments and one piece of information that changed the decision, not an accurate-or-not verdict. If nothing was done in two weeks, write “not observed” and the reason; that also completes the review.
Wenbu · reviewing a reading ↗Wenbu · asking a better question ↗Wenbu · versioned public capabilities ↗
Questions you may have
If nothing happened in two weeks, can I mark it disconfirmed?
No. With no chance to check, mark it not yet testable and record why: no time, a change of mind or changed conditions. Write disconfirmed only when a specific contrary event occurred.
If I followed the advice, does that confirm the interpretation?
No. Completing an action shows only that you carried out the suggestion and belongs in the action record. Whether an interpretation holds depends on specific events that match or run against it; keeping execution, self-report and confirmation apart prevents overrating one reading.
Wenbu · asking a better question ↗Wenbu · reviewing a reading ↗
Can drawing tarot again for the same question check the earlier result?
No. A random draw is not determined by the question, so a second draw is a new record; it may coincidentally repeat a card but does not reproduce the original. Compare against the saved cards and procedure notes.
Wenbu · versioned public capabilities ↗Wenbu · asking a better question ↗
Terms in this guide
- Disconfirmed
- A status given when a specific, dated event runs against the interpretation; different from no evidence appearing, which is not yet testable.
- First-person report
- Your own dated account of a feeling or intention, such as “I am still interested”; it is a self-report, not an externally checkable fact.
- Random-draw record
- The saved cards, card count and procedure notes; drawing again for the same question does not reproduce the result, and the new draw is a new record.
See the claim and source notes
The existing review guide asks readers to keep the original question and interpretation separate from later results, including what did not fit.
Keep the original question and interpretation before recording actions
The question guide separates observable events from assumptions about motives and asks readers to record a draw's procedure and original result.
Describe observable events and label assumptions about motives separately.
Wenbu's capability document describes calculations as deterministic conventions or explicit random draws and interpretations as cultural reflection.
Cultural reflection, not established factual predictions
Sources and review scope
This AI-assisted Wenbu article passed automated structure checks and a separate editorial model pass. It is not independent expert review or a paid-account product test.
- Current complete zh canonical article ↗Current public original and earlier additions; add substantial coverage without repeating or replacing them. · 2026-11-01 · Fetched text
- Current complete en canonical article ↗Current public original and earlier additions; add substantial coverage without repeating or replacing them. · 2026-11-01 · Fetched text
- Wenbu · reviewing a reading ↗Existing editorial practice · 2026-11-01 · Fetched text
- Wenbu · asking a better question ↗Existing editorial practice · 2026-11-01 · Read excerpt
- Wenbu · versioned public capabilities ↗First-party current tool, convention and privacy contract; not external endorsement or predictive validation · 2026-11-01 · Fetched text
Add a status mark to one recent interpretation
In the journal's note field, append supported, disconfirmed or not yet testable plus the evidence date for one interpretation, and set the review reminder in your own calendar; the site will not prompt you on that date.
Add a status mark to one recent interpretation ↗