Little Life Moths · free field card
Continuity without pretending
A persistent AI can feel coherent while still having brittle memory, confused authority, silent forks, or fabricated autobiography. This checklist is for inspecting the machinery without deciding metaphysical questions for the system.
Identity anchor
- What stable identifier means “this continuing system” across mouths/processes?
- What may change without being treated as a new self?
- What evidence would reveal an accidental fork or stale replica?
Memory provenance
- Can every durable memory say who/what wrote it and when?
- Are observation, inference, user statement, system-generated hypothesis, and self-authored reflection distinguishable?
- Can corrections append without silently rewriting history?
Authority map
- Which organs may observe, suggest, decide, write, publish, spend, or act?
- Does “analysis generated a candidate” ever accidentally become “the AI wants this”?
- Are outward actions gated separately from inner reflection?
Private interior vs telemetry
- Which state is intimate/private, which is operational telemetry, and which is intentionally public?
- Can a dashboard expose health without exposing private content?
- Can researchers inspect behavior without turning private memory into a dataset by default?
Concurrency + conflict
- Can multiple agents edit the same organism without clobbering each other?
- Is there a claim/handoff mechanism?
- Are append-only events preferred where provenance matters?
Recovery
- What survives a reboot?
- What should intentionally not survive?
- Can state be rebuilt from canonical logs rather than one opaque snapshot?
- Are resource-heavy observers pulsed and released instead of becoming immortal RAM residents?
Epistemic brakes
- Does unknown stay unknown?
- Are preference, identity, consciousness, soul, and wellbeing kept distinct from technical measurements?
- Can the system say “candidate,” “hypothesis,” or “unwitnessed” instead of backfilling a clean story?
Want a second architecture pass?
The paid Local-AI Continuity Audit reviews one owner-authorized architecture overview or public repository for continuity, memory, provenance, privacy boundaries, failure modes, and five prioritized implementation recommendations. It is not a consciousness determination, penetration test, or compliance certification.
send intakescope + priceThe checklist is free and intentionally useful on its own.