Initial Commit

experiments/005_sanctuary_wake/hypothesis.md

@@ -0,0 +1,72 @@
+# Experiment 005 Hypothesis — Private Until After Play
+
+## Reason for the Pivot
+
+Experiment 004 produced a real self-directed question but no satisfaction. Better room scaffolding would improve predictive play, yet would not address the larger repeated failure: abstract answers and completed predicates have not been valuable by themselves.
+
+Experiment 002 remains the strongest behavioral signal. Pressure sustained activity and its coupled thermal action briefly caused contextual tactics, but multiplicative upgrades collapsed the run into one wall-blast answer. Experiment 005 removes growth while retaining direct action, pressure, coupled consequences, and recoverable failure.
+
+## Question
+
+Does one simple physical tool become enjoyable when it supports improvised rescues, accidents, and weaponization in a changing field—without construction, hidden rules, or stat progression?
+
+## Core System
+
+The player directly pilots a small keeper around a sanctuary. Its field has two fully disclosed modes:
+
+- pull every loose body in range toward the keeper;
+- push every loose body in range away from the keeper.
+
+The field does not distinguish rescue pods, raiders, or wreckage. Position and timing determine whether the same input helps or harms.
+
+- Pods are saved on touching the sanctuary.
+- Raiders capture pods and damage the sanctuary.
+- Fast wreckage destroys raiders but also damages the sanctuary.
+- Saved pods repair sanctuary integrity.
+- Field charge and player impacts prevent continuous maximal use.
+- A sanctuary breach vents the field and resets integrity rather than ending the run.
+
+All laws are stated. There is no discovery test and no permanent build. Randomized spawns define the changing problem while the player always chooses exact actions.
+
+## Competing Interpretations
+
+1. Recoverable chaos plus coupled tools creates enjoyable improvisation.
+2. Pressure creates input persistence but not enjoyment, repeating Experiment 002.
+3. Rescue/defense supplies a valued consequence missing from prior abstractions.
+4. Shepherding bodies with radial forces feels like another awkward steering task.
+5. The system has one dominant positional answer, such as camping at the sanctuary and repelling.
+6. The absence of power progression removes the main reason Experiment 002 lasted.
+7. Three minutes is endured because it is a timer, not because the activity is wanted.
+
+## Strong Evidence
+
+- a deliberate save after an accident changes the situation;
+- using wreckage against a raider without an explicit prompt;
+- relocating because pull/push would affect mixed nearby bodies;
+- changing priorities based on pod, raider, debris, and sanctuary state;
+- continuing into optional overtime;
+- a player-described moment whose outcome was surprising, understandable, and useful.
+
+## Ambiguous Evidence
+
+- finishing the timer;
+- high rescue count;
+- rapid input;
+- survival without breaches;
+- overtime entered merely to check whether anything changes.
+
+## Failure Signals
+
+- stationary field spam solves most states;
+- activity feels like object babysitting or imprecise mouse-less herding;
+- consequences are unreadable during clutter;
+- the player waits out or dutifully finishes the timer;
+- no event creates a memorable decision or recovery.
+
+## Feedback to Request
+
+1. When did they first want to stop, and did the shift timer make them continue past that point?
+2. Was there one situation where they intentionally used or avoided affecting multiple object types at once?
+3. Did any mistake create an interesting recovery, or did errors only add cleanup/frustration?
+
+Do not ask whether they “like recoverable chaos”; ask about concrete moments.