Select a task and a learned run, choose a nonnegative input, then press run.
The plot shows the EM43 inference trajectory as a space-time diagram.
The input is encoded by placing a red marker at position x relative to the encoding start. The automaton then evolves by synchronous radius-1 updates on a one-dimensional tape. During inference, the tape extends to the right when activity approaches the boundary.
The computation halts when blue cells occupy at least half of the nonblank cells. The output is decoded as the position of the rightmost red cell relative to the encoding start.
Each run is a different learned machine, defined by its own program and rule table. Different runs for the same task may solve the task through different space-time dynamics.