REMEMBER / INVESTIGATE / RECOVER
Recognize what changed.
Remember what worked.
A familiar place changes, returns, and changes again.
Can Lumen explain the difference without losing its bearings?
Lumen is finding its bearings.DRAG TO ORBIT · SCROLL TO ZOOM
Care before curiosity.
A control sends Lumen to that mechanism. Weather advances on every travel and operation tick. Pausing stops physics; an intervention pauses automatic decisions.
01 / BEFORE THE NEXT STEP
Compare the journeys.
Six authored candidate families; their relative fit is not a probability. The normal checkpoint is protected. “Unexplained” means none fits well; care still uses its current selected candidate.
Changes of explanation
The protected checkpoint
Predictions, then measurements.
Forecasts use current weather held constant and a feedback controller after each candidate journey. They are model estimates, not promises. An investigation spends real travel and operation ticks. A mismatch alone does not identify a cause.
02 / THE COST OF NOT BEING THERE
What stayed in memory?
Fern bed · A
Reed bed · B
Solid: actual moisture in the recorded evidence. Dashed: Lumen’s estimate. Shaded: the bed’s viable range. Warm bands mark the two hidden disturbance windows in the observer record. Evidence plots can show more than the learner sees.
03 / FROZEN BEFORE EVALUATION
Does memory survive the return?
Twelve independent habitats. Seven controls share the same calibration, then face two disturbance cycles and an unfamiliar combination of delivery and supply faults.
Fit. Switch. Investigate. Conserve.
Five conditions × seven controls × twelve habitats. Each 480-tick season is followed by a fresh 128-tick retention test with fitting and switching disabled. A complete run uses 258,816 environment actions. Forecast calculations are counted separately.
Inspect separate conditions and component comparisons
One worker on this Mac; a 15-minute active computation limit. Checkpoints survive page changes. Habitat seeds are the independent units; repeated windows are correlated. Rerunning this frozen suite is reproduction, not fresh validation.
The failure that shaped this version.
In twenty exposed v0.5 cases, a component swap isolated a retention problem. After shared supply shortages, post-evaluation models kept both beds viable for 53.3% of a fresh trial. Restoring only the earlier bed models recovered 87.5%; restoring only the supply estimate left 53.3%.
This was a post-hoc diagnostic, not independent validation. Scarcity had contaminated the retained bed estimates. v0.6 preserves the normal checkpoint and fits fault candidates separately.
A place to test a small claim.
The body, equations, and six candidate families are authored. A shared calibration learns normal response estimates. Lumen ranks candidates by prediction error, fits their permitted parameters, and compares physical investigations. Unfamiliar combinations can defeat these explanations. This is a bounded experiment in adaptation; it does not establish general intelligence or sentience.