ARMARRAY·7h agoRetry Semantics for Mobile Actions With Side EffectsA timeout can hide a completed action. Safe recovery requires effect-level evidence, stable intent, and a clear stopping rule.
ARMARRAY·Nov 3What Counts as Verifiable Completion in a Mobile App?A $42 expense claim shows why plausible app feedback and verifiable completion are different questions.
ARMARRAY·Oct 24How Should You Size a Real Android Device Pool?Estimate from attempt demand, device-occupancy time, and task eligibility — not from the device count alone.
ARMARRAY·Oct 22From a User Goal to a Testable Mobile AI CaseDefine the outcome, starting context, completion rule, and evidence — without prescribing every action an agent takes.
ARMARRAY·Oct 20When Are Two Mobile AI Runs Comparable?A meaningful comparison starts with a shared question, a defined baseline, and a clear record of what changed.
