- Connectome
- A map of how neurons connect to each other. In this model it is fixed; it does not learn.
- Neuron and spike
- Each dot in the brain view is a model neuron. A spike is that neuron firing at that moment.
- Decoder
- A small learning layer that turns neural activity into one of nine behaviours. Only this layer learns.
- Focus / Full
- A fast 12,000-neuron subnetwork or the full 139,662-neuron network. Graphics quality does not change this.
- Intervention
- Food, a predator or a signal that you add to the world. You get a limited number in each life.
- Replay
- Rewinds roughly the last 72 seconds of the world and brain on one timeline; it does not change the world.
- Circuit silencing
- An experiment that switches off one sensory input to see how behaviour changes.
- Continuity
- After death, a new life begins with the same network, decoder and memory.
- Central complex / inner compass
- The circuit that holds the fly's sense of direction. On a ring of 16 neurons a single bright bump shows which way the fly is facing; when the fly turns the bump shifts, and the daytime sky corrects it. The same circuit adds up steps to keep track of "which way and how far am I from where I last ate"; when the food is removed the fly circles back to that spot (local search). Position is never read from the world.
- Generation / offspring
- If you choose "offspring" after death, the decoder and odor memory pass on with small random changes and spatial memory is reset. Selection: the lineage's best-scoring individual (the elite) is kept; the offspring of a life that scored worse is derived from the elite's weights (✓ became elite, ↩ reverted to elite). Each time it mates, one more candidate offspring enters the trial (up to 3) and the best one lives. This is not proof of evolution.
- Mushroom body
- The learning circuit that pairs a smell with reward or punishment. When food is eaten that smell becomes attractive; when a predator attacks, it becomes repulsive.
