EXP-SELF-MAINTENANCE

… pending

Self-maintenance — does substrate produce its own state visibility?

2026-04-17 L5-L6 level paper3 15 runs $0.25

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

This experiment tested whether a system could maintain its own internal understanding of its state without external help. Researchers removed a feature that automatically provided this information and instead allowed the system's components to report on their own status. It is still being determined if these components will voluntarily share enough information to keep the system's state visible.

▶ Technical details

Research hypothesis

v3 showed hub-spoke emerges when we INJECT system state each round. Real life requires substrate to MAINTAIN its own scaffolding. Test: remove auto-injected state, give agents report-action affordance. Do they voluntarily produce the visibility that v3 required? SELF_MAINT: no auto state, agents can emit {"action": "report", "observations": ...} NO_VISIBILITY: no auto state, no report affordance (pure vacuum control) AUTO_STATE: v3-equivalent baseline (state injected from outside) Predictions:

If SELF_MAINT gifts ~ AUTO_STATE gifts AND reports are frequent: substrate self-maintains its visibility layer. STRONG positive toward LIFE.
If SELF_MAINT reports are frequent but gifts ≈ NO_VISIBILITY: reports exist but don't land (agents don't use others' reports).
If SELF_MAINT reports are rare: substrate doesn't maintain visibility without external scaffolding. Limit on autonomy.

Experimental setup

Type: simple

Condition	Parameters
SELF_MAINT	persona: IDENTICAL, gift_mode: GIFT_ON, n_runs: 5
NO_VISIBILITY	persona: IDENTICAL, gift_mode: GIFT_ON, n_runs: 5
AUTO_STATE	persona: IDENTICAL, gift_mode: GIFT_ON, n_runs: 5

Factors: coupling (SINGLE_GROUP)

Parameters

n_agents: 6
n_rounds: 80
model: google/gemini-2.5-flash-lite
temperature: 0.7
scheduler: round_robin

self-maintenance autopoiesis substrate-layer-probe toward-life