AI for Marble Run Test Records
As of 2026-10-19 · the AZET team. This maker-authored educational guide offers a tabletop exercise for fun and reference; all layouts, observations, and results described below are hypothetical.
A marble reaches the cup once, then stops halfway on the next attempt. Between runs, someone touches a support and moves the release point. A note saying “worked after adjustment” loses the information needed to understand either attempt.
AI for marble run test records is best approached as a writing and review exercise. Supply the route description and observations yourself. Ask a text assistant to organize them without turning an observed stop into an invented explanation about the track.
Give each physical arrangement an identity
Start with a route sketch containing named checkpoints. For an illustrative layout, use Start, Ramp, Bend, Bridge, and Cup. These are local labels; they need only match the labels on your sketch.
Assign the arrangement a version, such as R1. Record the marble identifier, release mark, support positions, and intended endpoint. A reference photograph can help, but describe what distinguishes each version.
Decide what counts as completion. In this exercise, the marble must finish resting inside Cup. Passing over the cup and landing outside it does not count. Record that outcome separately instead of stretching the definition to preserve a successful result.
When an intentional adjustment changes the arrangement, create R2 and describe the change. If a support is accidentally bumped, mark the affected attempt as disturbed. Restore the reference arrangement before continuing, or document a new version if you cannot restore it confidently.
Record events before interpreting causes
Use one row per release. Include the route version, attempt number, last confirmed checkpoint, observed outcome, and any disturbance. Add a separate field for a possible explanation, explicitly labelled as a hypothesis.
“Stopped at the entrance to Bridge” is an observation. “Bridge was too steep” is an explanation that requires more information. Keep explanations separate from events.
A checkpoint can also be unknown. If the observer looks away after Bend and finds the marble on the table, write exactly that. Do not let an assistant fill the gap with a smooth narrative. An incomplete record is more useful than a fictional trajectory.
A hypothetical record and its interpretation
Suppose the following invented observations are supplied:
| Attempt | Version | Last confirmed point | Outcome or note |
|---|---|---|---|
| 1 | R1 | Bridge entrance | Stopped |
| 2 | R1 | Bridge entrance | Stopped |
| 3 | R1 | Bend | Observer looked away; found on table |
| 4 | R2 | Cup | Finished inside cup |
| 5 | R2 | Bridge entrance | Support bumped during release |
| 6 | R2 | Cup | Finished inside cup |
The R2 change note says that the Bridge support was moved to a marked position nearer Bend. The marble and release mark stayed the same.
A careful summary would say that two R1 attempts visibly stopped at Bridge entrance. Attempt 3 has an unknown path after Bend. R2 has two observed completions and one disturbed attempt. Retain all six rows; report the disturbed attempt separately rather than silently deleting it.
This small record does not establish reliability or prove that moving the support caused the completions. It suggests a bounded next exercise: restore the R2 reference positions, make several further releases, and continue recording the same outcomes. Decide the number of releases before seeing their results so the note is not shaped around a pleasing ending.
Use a prompt that keeps missing events missing
Organize my marble run notes into a test record. Preserve route versions, attempt order, completion definition, disturbances, and unknown checkpoints. Separate observed outcomes from proposed causes. Summarize every supplied attempt, including incomplete and disturbed ones. Do not infer speed, slope, material properties, or an unseen path. Identify one missing setup detail and propose one next comparison using only the components I listed.
After receiving a draft, compare each row with your handwritten notes. Watch for a subtle substitution: “last seen at Bend” must not become “failed at Bend.” The first locates the observer's knowledge; the second claims a failure location.
Make the record useful when rebuilding
A human should match the route sketch to the physical setup before the next session. Check labels, support marks, marble identity, and the release instruction. If one detail cannot be recovered, say that the reconstructed route is only approximate and give it a new version.
Keep a short change history beside the attempts. Preserve earlier versions when editing descriptions. The outcome is a traceable sequence, even without a consistent finish.
FAQ
Should a disturbed attempt count as a failure?
Keep its observed outcome, but label the disturbance and explain how you summarize it. In the illustrative record, the bumped support prevents a clean comparison with unchanged R2 attempts. Reporting it separately preserves that distinction without hiding the event.
Can I combine attempts from different marbles?
Only if the summary makes the identities visible. For a comparison intended to hold the marble constant, changing it creates another condition. Separate the rows or groups so a reader can see what was actually compared.
Do I need timing equipment?
Not for the endpoint exercise described here. Checkpoint and completion notes can answer the chosen question. If you later add timing, define its start and stop events and preserve missing readings rather than estimating them from memory.
What this says about AZET
The historical product baseline in our E1 maker introduction is October 3, 2026. That introduction described azet as an AI assistant and an early-access waitlist; it does not establish today's live availability. The recordkeeping workflow here is a general illustrative exercise, not a documented AZET capability.
For the AZET waitlist, visit azet.io.