Feature Learning Throughput
The mod learns from outputs that your AE2 network actually accepts. It measures a production window for an output and keeps recent completed windows for that network. Parallel batches are combined, so running two machines at once teaches their real combined speed instead of adding both durations.
For example, if a network accepts 64 processors over eight seconds, that window teaches roughly eight processors per second. A different ME network learns its own history. Items use item counts; fluids and supported chemicals use millibuckets. A new output can show No data yet until a usable window has completed, and changing machines affects only later samples.
A live job produces the outputs that become timing history after a completed window.
Previous: Time estimates | Features | Next: Confidence | Saved history