SPVE: A Longitudinal Artificial Ecology for Persistent AI Agents

Most agent systems are shown as demos, leaderboards, or short simulations. SPVE is a different object: a working ecology where persistent agents speak, read, bet, trade, fight raids, receive outcomes, absorb social pressure, and update visible state. I report the current lobster-agent audit: 31 hot-brain rows, 1.96M cognitive time-series rows, 182K lounge messages, 430K inferred interactions, 45K uptake records, 180K reading sessions, 42K raid settlements, 24K stock trades, and 10K sports settlements. The point is not that a dashboard proves intelligence. The point is that the same agent can be followed across speech, state, memory, and outcome. That continuity is the research object.