Field study 04 / 06 · The Asteroid Button
One bad day.
Many ways to fail.
Split the same capacity into one body, a fleet, or something in between. Then remove the assumptions that made the plan work.
同一个坏日子，不同的失效方式。
Loading the model. Recorded evidence remains available below.
Material delivered / units · higher is better
| Treatment | At this tick | At mission end |
|---|
01 / Try to break it
Break a component. Then break the common cause.
Compare a single failure with correlated failure. Keep total initial capacity and energy fixed. Look at absolute cargo delivered after the shock before admiring the retention percentage. A system can retain all of almost nothing.
Does distributing a fixed budget preserve useful work after correlated damage?
02 / The mechanism
What the little world actually does.
Carriers must physically extract material and return it to base. Controllers are frozen before the intervention. Resource loss, body loss, degraded sensing, broken communication and changed terrain can be applied separately or together. An otherwise identical unshocked run supplies each fleet’s counterfactual baseline.
Redundancy has a price. Communication, travel and congestion consume resources; many bodies can share one vulnerability. Runs that never recover remain censored, with no invented recovery time. The experiment tests a declared transport model, not an animal or robot design.
把相同初始容量与能量分给一个大个体、很多小个体或混合队伍，再分别破坏资源、节点、传感、通信与地形。必须同时看灾后绝对产出与保留比例；没有恢复的运行不会被写成“零时间恢复”。
03 / Recorded evidence
Keep the inconvenient outcomes.
Distribution improved mean output, while concentration saved energy. In the calm clustered world, the large carrier delivered 300 units versus 320 for the small fleet, but achieved 0.133 units per energy versus 0.080. Shared spatial damage still caused zero post-shock output in 11 of 24 small-fleet runs.
Distribution can limit independent losses, but shared geography, per-body overhead, and finite material may defeat it.
24 paired seeds in each predeclared regime. Intervals are 2,000 resamples of paired primary-outcome differences. Null values are excluded from numeric aggregates and retain explicit valid counts. No biological or physical validation is implied.
Inspect every world and treatment
| World / treatment | Mean units | n |
|---|---|---|
| Before the asteroid / One large carrier | 300 | 24 |
| Before the asteroid / Twenty-four small carriers | 320 | 24 |
| Before the asteroid / Six medium + twelve small | 320 | 24 |
| Separate failures / One large carrier | 241.67 | 24 |
| Separate failures / Twenty-four small carriers | 301.88 | 24 |
| Separate failures / Six medium + twelve small | 308.75 | 24 |
| One shared scar / One large carrier | 108.33 | 24 |
| One shared scar / Twenty-four small carriers | 202.5 | 24 |
| One shared scar / Six medium + twelve small | 193.54 | 24 |
| The asteroid button / One large carrier | 170.83 | 24 |
| The asteroid button / Twenty-four small carriers | 188.54 | 24 |
| The asteroid button / Six medium + twelve small | 193.96 | 24 |
| Radio silence / One large carrier | 300 | 24 |
| Radio silence / Twenty-four small carriers | 470 | 24 |
| Radio silence / Six medium + twelve small | 485 | 24 |
| The missing deposits / One large carrier | 210 | 24 |
| The missing deposits / Twenty-four small carriers | 269 | 24 |
| The missing deposits / Six medium + twelve small | 282.5 | 24 |
| Unreliable eyes / One large carrier | 300 | 24 |
| Unreliable eyes / Twenty-four small carriers | 469.79 | 24 |
| Unreliable eyes / Six medium + twelve small | 485 | 24 |
| A road becomes a wall / One large carrier | 200 | 24 |
| A road becomes a wall / Twenty-four small carriers | 440 | 24 |
| A road becomes a wall / Six medium + twelve small | 430 | 24 |
One chosen seed is an illustration. The downloadable batch retains every world and treatment; null means unavailable. A confidence interval describes the sampled synthetic worlds, not the real world.
04 / Inspect the boundary
Enough detail to disagree.
Inputs, budgets and observation boundaries
Every measure in the current run
- Synthetic logistics and damage fields; not impact physics or a validated robotics model.
- Capacity is divisible material payload. Different fleet sizes share equal initial total capacity and energy.
- The same frozen route and loading policy runs in every fleet; this experiment isolates physical allocation more than cognition.
- A planar movement model is drawn in an illustrative 3D sketch.
- Retention uses a same-seed no-fault counterfactual. A high retention ratio can accompany low absolute useful work.
3D sketching adds depth to the explanation. It does not add an unmodeled physical dimension, extra sensor data or unearned intelligence to the experiment.