RobotyUnderstanding a robot task from goal to result
My edited goal survived relaunch and led to a confirmed simulated result. Making that result understandable required a separate correction to how its hold duration was displayed.
Covers October 10, 2026 through October 12, 2026. Reconstructed later from available records.
The edited height was still there after installation and relaunch. I could return from an old result, change the goal and keep working without rebuilding the task.
In the connected iPhone walkthrough, selecting a historical result led back to authoring. Plane and height editing were available, and the edited height persisted. Reconnecting did not silently prepare the task. Prepare remained available until chosen, then Start appeared.
The UI helper opened the Start review but could not operate its confirmation, so I pressed Run one simulation once myself.
That run finalized success in 0.452 simulated seconds. The tool point finished about 0.44 mm from the target and held within 12 mm for 0.25 simulated seconds. The outcome appeared automatically. This was one simulated baseline.
Keep the record, explain the result
The compact presentation showed target changes in centimetres, distance in millimetres, and simulation duration separately from elapsed time. Tapping the numbers opened details with the recorded values and units.
A later offline review exposed a hold value of 0.25000000000000017 s. The display was initially changed to >0.25 s. I accepted that compact presentation during the review. No simulation ran then.
That interpretation was too strong. The evaluator accumulated 125 steps of 0.002 simulated seconds, or 0.25 seconds. The tiny excess was floating-point accumulation noise, not another measured interval. Python's explanation describes the general numerical issue.
The evaluator passed at a hold of at least 0.25 seconds, and the saved result passed. Showing >0.25 s suggested a margin that this measurement did not establish. Keeping more stored digits did not add measurement accuracy.
October 13 follow-up
The display was corrected to 0.25 s. A focused iPhone Simulator check showed that compact value alongside the original 0.25000000000000017 s in read-only details. It used a synthetic saved result and ran no new simulation. Formatter tests also covered values around the threshold and more precise requirements. The original measurement and evaluator result remain unchanged.
The walkthrough joined goal editing, deliberate preparation, confirmed Start and an automatic result. The later correction exposed a second responsibility: preserve the exact record while explaining what its precision can support.
For current controls, see Read your first result. This is the earlier walkthrough and its dated correction.