Can a wiring map become a fly?Can a wiring map become a fly?Stage 1 of 8 · 8 stages · ~5 min
Inside the fly-connectome story

Can a wiring map become a fly?

Trace the tissue. Run the neurons. Find out what makes a virtual fly move.

8 stages~5 min
  1. Map the wiring
  2. Run a circuit
  3. Test the behavior
ToySims

Fly connectome explained: from brain map to simulated fly

Explore how a fly connectome becomes a neural model and a simulated body, and test what the wiring alone cannot tell us.

Lesson summary

A connectome constrains a model; experiments test what the model can explain.

  1. Tracing and proofreading turn microscope images into identified neurons and contacts.
  2. Synapse counts give a directed graph, not a complete physiological model.
  3. A map’s specimen and anatomical coverage matter when comparing results.
  4. Identical wiring can produce different activity under different neuron rules.
  5. An embodied model adds sensory mappings, motor control and a feedback loop.
  6. Matched interventions test predictions; plausible movement alone cannot validate an entire brain.
Why check a reconstruction after software traces the cells?

A false join merges separate cells, while a missed continuation splits one cell. Both mistakes can corrupt the resulting circuit.

If matrix row S, column R contains three, what does it establish?

The toy map contains three contacts from S to R. It does not establish the reverse connection or an exact voltage effect.

What does the male CNS map include beyond a brain-only map?

It includes the connected ventral nerve cord. It is also a different specimen, so its totals are not growth in the earlier female brain.

Why did one relay fire while the other stayed quiet?

They received the same input, but slower leakage preserved enough voltage for successive inputs to cross the firing threshold.

Why does replaying the first cue change the arena result?

The body moves, but the controller keeps receiving the old distance. Fresh feedback would tell it to stop at the food.

What turns the relay experiment into a biological test?

Make the corresponding controlled intervention in an animal and compare the measured result with the model’s prediction. A toy result alone is not biological evidence.

Stage 1 of 8

What did the scan recover?

A virtual fly walks toward food, pauses to groom, and carries on. The picture is easy to understand; the claim behind it is harder. If its brain came from a scan, how much of the moving animal was actually recovered?

To follow that question, we need three objects: a neuron is a cell that processes signals, a synapse is a contact that lets one neuron influence another, and a connectome is a map of those connections.

We will follow the evidence from tissue to wiring, from wiring to activity, and from activity to a body. The small circuits in this lesson are invented so every step can be inspected.

Stage 2 of 8

How do slices become a neuron?

The map begins with preserved tissue under an electron microscope. A single image shows pieces of cells, rather than a complete neuron conveniently outlined for us.

Follow the orange profile through neighboring images. Its position changes, but those profiles belong to one continuous branch. Reconstruction software groups the image pixels into candidate cells.

People check the reconstruction because a wrong join can make two cells appear to be one, while a missed continuation can break one cell apart. Here, separating the orange and pink branches repairs an illustrative mistaken join.

Once the cells have been traced and their synaptic contacts identified, we can ask who influences whom. Mere proximity in a picture is not enough to establish a synapse. This small drawing illustrates the reconstruction problem; it is not microscope data.

Stage 3 of 8

What is inside the download?

Here is the idea at a scale we can count. In our invented example, S makes three contacts onto R, and R makes six onto O. Each contact has a direction.

We can compress all contacts from the same source to the same target into one directed edge, keeping the count attached. Three contacts become one edge labeled three; the original contact locations remain visible.

A matrix stores the very same information. Choose the source row and target column to find the count. The three in row S, column R does not put a three in the reverse cell.

That table is already useful for finding routes through a circuit. But three contacts do not tell us the exact voltage change caused by an incoming spike. A simulation still needs assumptions about strength, sign and timing.

Stage 4 of 8

Which fly is in the story?

Before running a model, check which map it uses. FlyWire’s 2024 resource reconstructed an adult female brain, with 139,255 neurons and roughly fifty million chemical synapses.

The male CNS resource highlighted in September 2026 includes the brain and ventral nerve cord, with over 166,000 neurons and 125 million connections in Google’s announcement. The nerve cord contains circuitry linking brain commands to body control.

These are different specimens and different coverage, so the counts do not describe one brain growing larger. The male resource had earlier releases; September 3 marks its paper publication and the Google announcement.

Eon’s March 2026 embodied demonstration used earlier FlyWire-based models. The new map and that older demo are related research, but they are not one experiment. What must a model add before either map can produce activity?

Stage 5 of 8

What makes a neuron fire?

A connection tells us where a signal can go. This small invented circuit lets us inspect what happens over time: sensory neuron S feeds relay R, which feeds output O.

The first input adds 0.6 to the relay’s voltage. Its firing threshold is 1, so this pulse alone is too small. Some voltage remains for the next input.

Keep eighty percent of that voltage, then add another 0.6. The total reaches 1.08. R fires and resets to zero; the downward stroke in the trace records that reset.

One tick later, the relay’s spike reaches O with a weight of 1.2. That crosses the output threshold. A route through the graph has become an event at a particular time.

Now compare the pink trace, where only twenty percent of voltage survives each tick. The wiring and input are identical, but R never reaches threshold. The model needs a rule for time as well as a map of connections.

Identical wiring and input can produce different activity when neuron dynamics change.

Change Voltage leak and inspect the selected circuit’s output. Use the time-axis detail to compare what accumulated between the same inputs.

Slow leakage lets successive inputs combine; fast leakage drains them before they can trigger a spike. These are teaching values, not measured fly voltages. A biological model needs evidence for its chosen dynamics.

Stage 6 of 8

How does activity acquire a body?

An output spike still has no legs. A virtual animal needs an interface that turns neural activity into body commands, and a body simulation that turns those commands into motion.

Eon reads a small selection of descending signals through an added control layer. Its technical account describes hand-chosen mappings and existing trained body controllers. That layer matters when interpreting a convincing walking video.

Our simpler arena makes the feedback problem visible. A sensory adapter reports distance to food, a toy controller chooses whether to walk, and each move changes the next distance. Distance sensing here is an invented teaching convenience, not a claimed fly sense.

With fresh feedback, the controller stops at the food. If we disconnect that feedback and keep replaying the first cue, the command continues even after the position has changed. The two histories keep the changed distance beside the cue actually received.

Try both Feedback settings and follow the history. The return path carries new information, rather than simply making the diagram look like a loop.

Fresh feedback changes the next decision; a replayed cue leaves the controller responding to an old situation. This toy makes the dependency explicit, while a biological model needs sensory and motor mappings supported by experiments.

Stage 7 of 8

What would convince us?

This gives us something we can test. Return to the same sensory burst and slow-leak neuron rule, then compare the intact circuit with one in which relay R is silenced.

The input still arrives, but the output spike disappears when R is silenced. That is a specific prediction about an intervention, not proof about a real fly.

Use Relay R to restore the missing output without changing the sensory input. Compare the two spike records before deciding which part of the circuit caused the difference.

In a biological study, researchers manipulate identified neurons and compare observed activity or behavior with a model’s prediction. Shiu and colleagues tested predictions in feeding and grooming circuits. Those successes support the tested transformations, not every possible behavior.

A moving body alone cannot reveal whether each internal circuit is correct. Learning, internal state and detailed physiology remain further questions. Whether a model is conscious, or preserves the original animal’s identity, is not settled by matching a movement.

Stage 8 of 8

What the evidence lets us say

Connected profiles gave us candidate cells. Checking their continuity prevented an image-processing mistake from becoming a mistaken neuron.

Identified contacts gave us a directed graph. The matrix retained both their counts and the direction of influence.

Anatomical coverage told us which nervous system the graph described. The female brain map and newer male CNS map are distinct resources.

The same wiring produced different activity under different leak rules. A wiring map constrains the model without choosing all of its dynamics.

The added body interface converted activity into movement, and fresh feedback changed the next input.

A controlled intervention turned an appealing animation into a prediction that could be checked. The small circuits here were illustrative; the cited map counts and research claims came from the published sources.

A connectome is a powerful constraint on explanations of a brain. The work is to find which additional rules make those explanations survive experiments.

Research behind the lesson

Research checked 12 September 2026. Published map totals are attributed to their datasets. All small circuit weights, scan profiles and arena trajectories here are illustrative; no biological dataset is being simulated in this page.