How to read the data

A record of a simulated world.

Aionterra studies environmental conditions, life processes and their interactions in a continuing simulation. The published record describes the state exposed by its observation system. It is not a direct measurement of the real world.

Shared learning and individual bodies

The current research configuration uses one shared learned policy with separate bodily states and sensory histories. The shared counter records accepted learning transitions from those bodies. Several transitions can contribute to one policy update; this is not a count of update batches, independent minds or acquired skills. Each record retains its model identity. Older individual-learning counters must not be joined to the shared update series.

Movement and mouth contact use a simplified body model. Learned sleep selection, communication and reproductive behaviour are not established in this configuration. Missing or inactive measurements remain unavailable, and behavioural benefit requires a suitable comparison.

What is being measured?

Choose a measurement to read its definition, unit and coverage. Unit labels follow your display preferences; source definitions and downloads retain their original units.

Simulation, export and publication time.

In this record, plant growth and soil chemistry update every 60 simulation seconds. Other processes retain their own schedules. These numerical update intervals are distinct from observation sampling and do not imply identical results at a different update interval.

The relative simulation calendar starts at Day 1, Week 1, Month 1 and Year 1. A day is 86,400 simulation seconds and a week is seven days, not an ISO week. Months and leap years follow a proleptic Gregorian display calendar starting at year 1. Month numbers count cumulatively from the start. These labels are not real-world dates and do not change seasons or elapsed simulation time.

Simulation time measures elapsed time inside the world. Export time records when the observation package was created. Publication time, when a release is made public, is a separate event.

This record contains 2 time points. Sampling is irregular. Observation frequency does not imply that every simulation step was recorded.

What this record cannot tell us.

• Simulation output, not direct real-world measurements.
• Irregular UI observation samples, taken on page entry, manual refresh or optional live refresh; not complete run history.
• State identity refers only to the last snapshot; profile identity is not exposed by this projection.
• Population values are exported only at observed sample times; retained parentage is not a lifetime birth count. Missing Characters does not fall back to the native apparatus.
• Human body readings come from the body model (oxygen reserve, water and food stores, core temperature, 0-1 health). They are model state, not clinical measurements, and are exported only for the body model that produces them.
• Event records carry cardinally separate eventType, simTimeSeconds, amount and amountUnit fields. The amount comes from the journal's structured evidence and is null when the event has no magnitude; consumers must not parse the evidence sentence.
• Organ histories, named skills and validated discoveries are not exported. Geographic coordinates are provided by the paired world.json only when source georeferencing matches.
• One preserved pre-step observation plus up to 240 retained UI observations; gaps are not filled. Rolling samples reset on reload, time rollback or world extent changes; the initial observation stays attached to the run.
• Missing values are not zero. Gaps remain visible in the chart and are preserved in downloaded data.
• Counter increases alone do not establish a new capability or scientific discovery.

The values behind the view.

CSV downloads contain the selected measurement and time window, with simulation time in seconds, measurement unit, scope and status. Display rounding does not modify the exported numerical values.