Ariel Zeleznikow-Johnston

The connectome selects the behavior, but neuromech flies control as executed.

In their blog post, Eon are open about this.

They compare the descending neurons to a car's steering wheel, accelerator, and brake, you can predict what the car will do from these controls without explicitly simulating every combustion event inside the engine.

They also acknowledge that the visual system activity displayed so prominently in the video, derived from the Lapelainen model, is somewhat decorative and does not substantially drive behavior.

They do note that the brain body mappings are in some cases somewhat arbitrarily chosen by hand.

And they explicitly state the work should not yet be interpreted as a proof that structure alone is sufficient to recover the entire behavioral repertoire of the fly.

This is fair enough, and their efforts to connect brain and body models are genuinely useful engineering.

If E.ON had described this as the first integration of connectome-constrained brain and body models into a closed sensorimotor loop, nobody in the fly neuroscience community would have objected.

But they didn't say that.

They said we've uploaded a fruit fly.

Transparency in a blog post that few will read doesn't undo a headline that millions saw.

The typical person who encounters a claim on X watches the video and sees a fly walking, grooming, and feeding while a digital brain flickers alongside it is probably not going to think a simplified brain model is selecting from a small menu of pre-programmed behaviors via a hand-tuned interface.

They're likely to think the fly has been faithfully recreated inside a computer.

It hasn't.

EON's virtual fly implements only a handful of behaviors, and those rely heavily on NeuromechFly's pre-trained controllers rather than on the Connectome.

This is the most fundamental problem with the demo as evidence of an upload.

Because the body model already knows how to walk, groom, and feed, almost any signal that triggers the right controller at the right time will produce fly-like behavior on screen.

You could replace the connectome with a simple rule-based script, if dust, groom.

If sugar, feed.

Otherwise, walk forward, and the resulting video would look much the same.