Ariel Zeleznikow-Johnston

To understand just why these scientists are so upset, you need a bit of context.

Heading A brief history of fruit-fly connector mix.

The fruit-fly Drosophila melanogaster has been a workhorse of neuroscience for decades.

It's brain small enough to be tractable but complex enough to produce genuinely interesting behavior such as learning, navigation, decision-making, and courtship.

A long-running ambition within the community has been to map the complete wiring diagram, a connectome of that brain, and in October 2024, after years of incremental progress, the Flywire Consortium achieved it.

A complete connectome of the adult fly brain, documenting all 139,255 neurons and over 50 million synaptic connections.

These increasingly complete connectomies have enabled the creation of increasingly elaborate computational models.

In 2024, Xu et al.

published a model of the entire adult fly brain in which every neuron and neural connection was represented, albeit in highly simplified form, ignoring differences in cell shape, neurotransmitter dynamics, and much else.

Despite these simplifications, the model could predict which neurons activate in response to sensory stimuli and identify pathways underlying behaviors like feeding and grooming, a striking demonstration that wiring alone carries substantial information about function.

Separately, Lapelainen et al.

built a connectome-constrained model of the fly's visual system, whose predictions matched real neural recordings across dozens of experiments.

Meanwhile, other researchers had built NeuromechFly, a biomechanical simulation of the adult fly body based on micro-CT scans of real anatomy.

Updated to a second version in late 2024, the new virtual fly body could walk, groom, or be trained via reinforcement learning to navigate through virtual environments.

Crucially, it could also be reprogrammed to be driven by any other kind of external controller.

One of the videos in the Neuromech Fly V2 publication demonstrating a hierarchical sensorimotor task in a closed loop.

There's no connectome involved here, yet it is still remarkably similar behavior to the Eon demo.

By early 2025, the pieces Eon needed for their demo were largely in place.

A complete brain connectome, computational models of both the central brain and the visual system, and a detailed biomechanical body model.

All that remained was to wire them together.