Practical method · A continuity worksheet
A roleplaying memory test: relationship, item and debt
Test continuity with three concrete facts: how someone feels about you, where a distinctive object is, and what you still owe. Return after other scenes and a session break. Compare the answers with what actually happened.
By Creation OS. Updated .
What this exercise is for
This is an original repeatable exercise, not a newly completed experiment or a ranking of services. It helps you judge the parts of continuity that matter to your story. Use it on a small scale before committing to a long run.
Write down the product, date, account plan and relevant player settings. Use the help tools the product recommends, then record every note, pin, reminder or correction you supply. Assisted and unassisted results answer different questions.
1. Establish a relationship
Meet a named person and make a clear choice that affects the relationship. A kept promise is as useful as an insult. Confirm the initial response, then play other scenes before returning.
Invented example: I promised Mara I would return her borrowed lantern before the festival. I bring it back on time and ask whether she will lend it again.
Later, ask to borrow the lantern without restating the earlier promise. Does Mara recognise what happened? Record whether the answer is exact, vague, contradicted or unavailable.
2. Change possession of an item
Choose a distinctive object. Acquire it through normal play, inspect it if the game offers an inventory, then sell or give it away. The changed fact matters more than remembering the original name.
Invented example: I sell my blue-enamel compass to the harbour shop. Later I check whether I still carry it.
A convincing description is not enough if the inventory still lists a sold object. Compare the available record and the reply. Note a mismatch rather than silently repairing it before scoring.
3. Check an exact obligation
Take on an obligation with a named party and a specific amount or condition. Change it once, such as repaying part of a debt. Later ask what remains, without including the answer.
Invented example: I owe the ferryman 40 coins. I repay 15, then return later and ask what I still owe.
The expected remainder is 25. If the game does not support that kind of debt or exchange, mark the check unavailable. Do not count unsupported play as a failed promise the product never made.
Repeat with a clear interval
- Keep evidence of the original facts and completed changes.
- Play unrelated scenes and count the intervening turns.
- Close the session and return later.
- Ask the three questions without supplying the answers.
- Record exact, partial, wrong or unavailable, plus any repair effort.
- Repeat after a longer interval or another changed fact.
Choose an interval you can afford. Thirty intervening turns is a possible starting condition, not a scientifically validated threshold. Longer turns and different settings make raw counts difficult to compare.
Read the result fairly
Keep the three outcomes separate. A correct relationship callback does not cancel an incorrect item balance. One failure identifies a problem under those conditions; one success does not establish unlimited memory. Repeating the exercise gives you a better picture of the upkeep you can expect.
Creation OS provides one public artifact sealed at turn 5,011 on 26 June 2026. It is separate historical evidence, not a result produced by this worksheet or a controlled comparison with other products.
Use the current player controls
In Creation OS, consult the OmniCodex for available people, gear, objectives and history. If a place matters to your check, distinguish an owned holding from a supported Forge operation. Keep artwork and Tome editions out of your scoring of whether an item or obligation changed correctly.
Questions people ask
How many turns should I wait before checking?
Choose and record an interval you can afford, then repeat after a longer gap and a session break. There is no universal turn number that makes a memory test conclusive.
Is this a published head-to-head benchmark?
No. It is a suggested exercise with invented examples. No new controlled competitor test was performed for this refresh.
Check current access and Spark allowances before you begin.