So far, it is only possible to create unconditioned reflexes (without training), programming only due to the architecture of the network.
Further plans are to add a scene with a nematode (ringed worm) and possibly a swimming tadpole. Then learning according to Hebb. The mechanism for creating large populations of neurons according to a given pattern. Then there are endless creative possibilities.
I would like to see the management of hard, but I myself am not friends with hard. The source contains examples of interaction between browser tabs via Local Storage.