Shaduf.Research preview
Digital Fly Lab/How it works

How it works

How a fruit fly brain simulation works: the FlyWire connectome, neuron models and bodies

Every digital fly project has the same basic loop. Something goes in, a model of the brain turns it into activity, and something comes out. What matters is which parts of that loop come from the real fly and which parts a person added.

The sense → model → act loop

The sense, model, act loop of a digital flyFive stages from left to right: input mapping (usually hand-made), connectome wiring (measured), neuron model (modelled), output mapping (hand-made or trained), and a body or game (modelled or scripted). Stages 2 and 3 are the simulated brain. A dashed line returns from the body to the input for closed-loop projects.The simulated brain1 · SENSEInput mappingStimulus → chosensensory neuronsusually hand-made2 · WIRINGConnectomeWho connects to whom,traced in microscopeimages: measured3 · DYNAMICSNeuron modelSpikes or firing ratesover time, e.g. LIFmodelled4 · READOUTOutput mappingChosen neurons → keys,muscles or a scorehand-made or trained5 · ACTBody or gameA physics body or agame enginemodelled or scriptedDashed line: the world changes and gives new input. Only closed-loop projects do this. The sense, model, act loop of a digital fly, from top to bottom: input mapping, connectome, neuron model, output mapping, body or gameThe simulated brain1 · SENSEInput mappingStimulus → chosensensory neuronsusually hand-made2 · WIRINGConnectomeWho connects to whom,traced in microscopeimages: measured3 · DYNAMICSNeuron modelSpikes or firing ratesover time, e.g. LIFmodelled4 · READOUTOutput mappingChosen neurons → keys,muscles or a scorehand-made or trained5 · ACTBody or gameA physics body or agame enginemodelled or scriptedDashed: closed loop
A generic digital fly. Colours show where each part usually comes from. Real projects differ: the catalogue records this for each entry. For one case we measured ourselves, see who does the walking.
  • MeasuredTaken from the real animal. In most projects this is only the wiring: which neurons connect, and through how many synapses.
  • ModelledA mathematical rule chosen by scientists, such as how a neuron adds up its inputs and when it fires. It is a simplification of real biology.
  • Hand-madeChosen by the project author: which neurons receive the input, which neurons are read out, and what counts as an action.
  • TrainedLearned by a machine learning method, for example a readout layer or a controller trained to reach a goal.
  • ScriptedOrdinary program code or animation that does not depend on the simulated neurons.

A project can use a real connectome and still get most of its visible behaviour from hand-made or scripted parts. That is why we grade each project separately. See Is it real? Whether the measured wiring itself makes a difference is a separate question, tested by scrambling it: Does fly wiring help? compares 31 such studies.

What a connectome is

A connectome is a map of the connections in a nervous system: which neurons exist, which neurons connect to which, and how many synapses join them. Fly connectomes are traced from electron microscope images of one animal, cut into thousands of very thin slices. Computers and human proofreaders follow each neuron through the images.

A connectome is a wiring diagram, not a working brain. It does not say directly how strong each connection is, how neurons change with learning, or how chemicals such as hormones change activity. A simulation has to add those parts, usually with simple rules.

Map or brain? Run with such a model, the map still does something specific to its wiring. In our own looming test (5 Oct 2026) a shadow on one eye made the real map's turning neurons fire on the far side and turned a simulated fly away in 4 of 4 runs; scrambled maps with the same number of connections per neuron stayed silent (0 of 3). One model, open loop, our hand-made mapping, a constant external walking drive, 1 s; not a real fly.

What the whole-brain models add, and what they leave out

Most projects reuse the leaky integrate-and-fire (LIF) model of Shiu et al. (Nature, 2024). In that model, and in most copies of it:

  • Connection strength comes from synapse counts. Each connection's weight is its FlyWire synapse count multiplied by one free number. The authors chose that number by hand so that sugar-sensing neurons firing at 100 Hz drive the proboscis motor neuron MN9 to about 80% of its maximum.
  • Excitation or inhibition comes from predicted transmitters. The sign of each connection comes from a machine-learning prediction of the neurotransmitter, not from a measurement of each synapse.
  • Many parts of biology are missing. In the authors' words, the model "does not account for gap junctions, non-spiking neurons, internal state or long-range neuropeptides, and assumes that the basal firing of each neuron is zero". Neuromodulation and learning (plasticity) are not in the base model. What the map leaves out, below, puts one number and one source on each missing part.
  • It is slow on a normal computer. The paper reports about 5 minutes per simulated second per CPU thread for the Brian2 version.

Source: Shiu et al. 2024, Methods and Discussion (full text on PubMed Central).

What the map leaves out

People ask whether a fly brain simulation is "a brain or just a map". Here is what the map-based model leaves out, part by part, each with one number and one source we read on 9 Oct 2026.

Short answer: The model is one fly's wiring map: 138,639 neurons and 15,091,983 connections, run with one simple neuron rule. It leaves out electrical synapses (gap junctions), neuropeptides, learning, the different behaviour of different kinds of neurons (including neurons that do not spike), the differences between flies and over a fly's life, and the nerve cord and body. So it can show what the wiring alone does, not everything a fly's brain does.

  • 138,639 neurons
  • 15,091,983 connections (neuron pairs)
  • 54,492,922 synapses
  • 1 hand-set strength (0.275 mV per synapse)
What the map leaves out: one number and one source per part, checked . The rows measure different things, so they do not add up to one share of the brain. In-model counts are from our copy of the Shiu et al. model (code 91bdd1e7) on FlyWire v783.
Part of biologyIn the model?One numberSource
Electrical synapses (gap junctions)direct electrical links between neuronsNo0 electrical links 8 innexin (gap-junction) genes in the fruit fly genome, 4 of them detected in brain neurons; the wiring map has only chemical synapses, so the model has 0 electrical linksFrontiers in Neural Circuits 2025 (innexins); Dorkenwald et al. 2024 (FlyWire: chemical synapses only)
read directly, 2026-10-09
Neuromodulators (dopamine, serotonin, octopamine)slow signals that change how other neurons respondPartly8,316 neurons 8,316 neurons in the model are predicted to use them (dopamine 5,905, serotonin 2,201, octopamine 210); the model wires them in as ordinary fast excitatory connections (dopamine 5,897 excitatory / 3 inhibitory, serotonin 2,127 / 9, octopamine 192 / 1), so their slow modulating effect is not modelledour count from the model files (report of 9 Oct 2026); Shiu et al. 2024, Methods
our data, 2026-10-09
Neuropeptideschemical messengers that spread beyond the synapseNomore than 50 more than 50 neuropeptide precursors (and their receptors) are known in Drosophila; none are in the modelNässel and Zandawala 2021 (review); Shiu et al. 2024
read directly, 2026-10-09
Learning (plasticity)connections that strengthen or weaken with experienceNocount × 0.275 mV every connection's strength is fixed before the run: synapse count x 0.275 mV (one hand-set value for all 15,091,983 connections); nothing changes during or between runsthe model's code, model.py at 91bdd1e7
our data, 2026-10-09
Different kinds of neuronsevery neuron follows the same simple rule, and neurons that do not spike are not modelledNo1 rule 1 rule (leaky integrate-and-fire, same parameters) for all 138,639 neurons; non-spiking neurons, neuron shape and receptor differences are left outShiu et al. 2024
read directly, 2026-10-09
Which chemical each neuron sendspredicted from images, not measuredPartly94% per neuron the transmitter of each neuron is a prediction from electron-microscope images: 94% correct per neuron (87% per synapse, 91% for known cell types) in the published test; this decides whether a connection excites or inhibits in the modelEckstein et al. 2024 (Cell)
read directly, 2026-10-09
One adult female fly at one momentno differences between flies, sexes or agesNo53% 53% of the connections seen in a second fly's brain (hemibrain) were also found in FlyWire's fly; connections of more than ten synapses were found again more than 90% of the timeSchlegel et al. 2024 (Nature)
read directly, 2026-10-09
The nerve cord and bodythe brain model has no spinal-cord equivalent, muscles or legsNoabout 20,000 neurons about 20,000 ventral nerve cord neurons (next to about 140,000 in the brain) are not in the brain model; our body test drove FlyGym's own walking controller, not this wiringBates et al. 2026 (Nature, BANC)
read directly, 2026-10-09

Quotes behind each number: the neuropeptide count is a review's statement; the innexin page answered a bot check, so its full text was read through NCBI's E-utilities. Nature pages answered a bot check to scripts, so the PubMed Central copies were read. Method: report of 9 Oct 2026.

Datasets compared: FlyWire connectome, MaleCNS, BANC and others

Different teams have mapped different flies and different parts of the nervous system. Always read a neuron count together with its dataset, version and scope. Never add a brain count to a nerve-cord count from a different animal.

Seven fruit fly connectome datasets, checked 27–28 September 2026 on the official pages. Data licences are quoted from the official wording.
Dataset Animal and scope Neurons (release) Connections How to get it Data licence Main paper
FlyWire FAFB (Female Adult Fly Brain)
Entry flywire-fafb
Adult female; whole brain including both optic lobes; no nerve cord 139,255 in public release v783 (Oct 2023 snapshot). 127,400 proofread neurons in v630, used by the Shiu et al. model 3,732,460 neuron pairs at 5 or more synapses (Codex default); "50 million" synapses in the headline Codex (Google sign-in for apps and downloads); CAVEclient; the Shiu model repository ships v630 and v783 files (about 87 MB and 101 MB) CC BY-NC 4.0
Non-commercial only. "FlyWire's public release data is made available under license CC BY-NC 4.0" (flywire.ai/guidelines). Cite Dorkenwald et al. 2024 and Schlegel et al. 2024.
Dorkenwald et al., Nature 2024 (doi); Schlegel et al., Nature 2024 (doi)
MaleCNS (Male Central Nervous System)
Entry malecns
Adult male; whole central nervous system: brain, optic lobes and nerve cord, with the neck intact 166,700 in v1.0 (released 8 Jun 2026). 166,691 in the v0.9 preprint. 165,122 fully traced neurons in v1.0, the subset many simulators load 6,242,118 at 5 or more synapses (Codex); "125 million synaptic connections" (Google Research) neuPrint (account token); neuprint-python; bulk files from male-cns.janelia.org; Codex CC BY 4.0
"The Male CNS dataset is licensed under CC-BY." Commercial use allowed with attribution.
Berg et al., Cell 2026, published 3 Sep 2026 (doi)
BANC (Brain And Nerve Cord)
Entry banc
Adult female; brain and nerve cord in one animal; lamina and ocellar ganglion missing 158,262 in v888 (Codex). 155,916 in the Nature paper (analyses on v626). The project README says "approximately 188,000"; its scope is not stated (unresolved) 3,037,361 at 3 or more synapses (Codex v888) Codex; Harvard Dataverse deposit CC BY 4.0 for the Dataverse data
The analysis repository has no licence file, although its README says CC BY 4.0.
Bates, Phelps, Kim, Yang et al., Nature 2026 (doi)
MANC (Male Adult Nerve Cord)
Entry manc
Adult male; nerve cord and neck connective only 23,665 in v1.2.1 (Codex) 5,305,638 at 1 or more synapse (Codex) neuPrint; public Google bucket; Codex CC BY (version not stated)
"The MANC is licensed under CC-BY." (Janelia)
Takemura et al., eLife 2024; Marin et al.; Cheong et al.
Hemibrain
Entry hemibrain
Adult female; part of the central brain (most of the right half), no optic lobes, no nerve cord About 25,000 in v1.2.1 About 20 million synapses (widely quoted; not re-checked) neuPrint; neuroglancer CC BY (version not stated)
"Hemibrain is licensed under CC-BY." (Janelia)
Scheffer et al., eLife 2020 (doi)
FANC (Female Adult Nerve Cord)
Entry fanc
Adult female; nerve cord About 14,600 cell bodies in the paper (search summary only, not verified) About 45 million synapses (search summary only) Latest reconstruction restricted to authorised users; electron-microscope images public via BossDB No open data licence found
Access follows the FANC community rules. The tool code is GPL-3.0.
Azevedo, Lesser, Phelps, Mark et al., Nature 2024 (doi)
Larval L1 brain
Entry larva-l1
First-instar female larva, 6 hours old; brain only 3,016 (480 input neurons and 2,536 brain neurons) About 548,000 synaptic sites Paper supplementary files; CATMAID hosted by Virtual Fly Brain; code on Zenodo Article CC BY 4.0
No separate data licence found; treat with care.
Winding et al., Science 2023 (doi)

Codex also lists MAOL v1.1 (male right optic lobe, 52,445 neurons). It comes from the same male fly as MaleCNS, so its neurons are part of MaleCNS, not extra.

Sources: Codex, FlyWire guidelines, male-cns.janelia.org, BANC project, Virtual Fly Brain neuron counts, Winding et al. 2023, Google Research blog.

Why the neuron counts differ

News stories quote many different neuron counts. Each one belongs to a specific dataset, version and scope:

Why fruit fly neuron counts differA number line from 120,000 to 170,000 neurons. FlyWire counts: about 125,000 (rounded), 127,400 (v630), 127,978 (preprint), 138,639 (v783 file in the Shiu repository), 139,255 (v783 release), about 140,000 (rounded). MaleCNS counts: 165,122 (v1.0 traced only), over 166,000 (rounded), 166,691 (v0.9 preprint), 166,700 (v1.0 release). Filled dots are exact release counts, hollow dots are rounded headlines.FlyWire FAFB · female brainMaleCNS · male brain and nerve cord120,000130,000140,000150,000160,000170,000gap of 616 neurons: cause not establishedabout 140,000 · rounded on flywire.ai139,255 · v783 public release, whole brain138,639 · v783 file in the Shiu model repository127,978 · preprint count127,400 · v630, all simulated by the Shiu et al. modelabout 125,000 · rounded headline in the 2024 model paper165,122 · v1.0, fully traced neurons onlyover 166,000 · rounded, Google Research blog166,691 · v0.9 preprint166,700 · v1.0 release (Codex)about 125,000 · rounded headline in the 2024 model paper127,400 · v630, all simulated by the Shiu et al. model127,978 · preprint count138,639 · v783 file in the Shiu model repository139,255 · v783 public release, whole brainabout 140,000 · rounded on flywire.ai165,122 · v1.0, fully traced neurons onlyover 166,000 · rounded, Google Research blog166,691 · v0.9 preprint166,700 · v1.0 release (Codex) Why fruit fly neuron counts differ: FlyWire counts from about 125,000 to 140,000, MaleCNS counts from 165,122 to 166,700. Filled dots are exact release counts, hollow dots are rounded headlines.120k130k140k150k160k170kMaleCNSFlyWire FAFB166,700 v1.0 release166,691 preprintover 166k, rounded165,122 v1.0 tracedabout 140k, rounded139,255 v783 release138,639 Shiu v783127,978 preprint127,400 v630 Shiuabout 125k, roundedabout 125,000 · rounded headline in the 2024 model paper127,400 · v630, all simulated by the Shiu et al. model127,978 · preprint count138,639 · v783 file in the Shiu model repository139,255 · v783 public release, whole brainabout 140,000 · rounded on flywire.ai165,122 · v1.0, fully traced neurons onlyover 166,000 · rounded, Google Research blog166,691 · v0.9 preprint166,700 · v1.0 release (Codex)
Filled dots are exact counts for a release; hollow dots are rounded headlines. Hover over a dot for its label. The same numbers are in the table below.
The neuron counts people quote, and what each one is.
NumberWhat it actually is
"more than 125,000"Rounded headline for the FlyWire brain in the Shiu et al. 2024 abstract
127,400Proofread neurons in FlyWire v630; all are simulated in the Shiu et al. model
127,978FlyWire count in the bioRxiv preprint, before more proofreading
138,639Neurons in the v783 input file of the Shiu model repository; what most "FlyWire v783" simulations run
139,255FlyWire public release v783, whole brain including optic lobes
"140K"Rounded figure on flywire.ai
165,122MaleCNS v1.0 neurons with status "Traced" (fully proofread), out of 211,577 annotated bodies. Counted by a community project from the official files; not re-counted by us
166,691MaleCNS preprint (v0.9)
166,700MaleCNS v1.0 in Codex
"over 166,000"Rounded headline on the Google Research blog

Still unexplained: 616 neurons. The Shiu repository's v783 file has 138,639 neurons, while the v783 release has 139,255. We have not found out why. One possible reason is that neurons without connections in the exported table were dropped, but this is not verified.

What you have to download, and how big it is

On 6 Oct 2026 we read the size, checksum, licence and access rule of every file in 8 releases (403 files) from each host's own storage metadata, without downloading the data. To run the Shiu et al. model you need two files, 104.1 MB; a full release is gigabytes to hundreds of gigabytes. How to get the two files: Build your own.

How big is the download? Fly connectome files, by releaseHow big is the download? Fly connectome files, by release. Log scale, sizes from storage metadata, checked 6 Oct 2026 without downloading the data. Our beginner path: 104.1 MB; FlyWire annotation table: 31.7 MB; hemibrain v1.2 summary: 45.9 MB; FlyWire v783, official: 10.6 GB; MaleCNS v1.0 flat files: 31.3 GB; BANC deposit: 536.1 GB (227.8 GB open, 308.3 GB on request); MANC v1.2.1: size not stated; Larva L1 (Winding et al.): size not stated.Release and what it containsSize (log scale: each step is 10 times bigger)10 MB100 MB1 GB10 GB100 GB1 TBOur beginner path2 files: wiring + neuron listOur beginner path: 104.1 MB (104,131,989 B)104.1 MBFlyWire annotation table1 file (cell types)FlyWire annotation table: 31.7 MB (31,720,298 B)31.7 MBhemibrain v1.2 summary1 file, compact wiringhemibrain v1.2 summary: 45.9 MB (45,872,577 B)45.9 MBFlyWire v783, official5 files incl. every synapseFlyWire v783, official: 10.6 GB (10,596,831,504 B)10.6 GBMaleCNS v1.0 flat files11 files, brain + nerve cordMaleCNS v1.0 flat files: 31.3 GB (31,318,683,398 B)31.3 GBBANC deposit102 open files + 277 on requestBANC: 227.8 GB open (102 files)BANC: 308.3 GB more on request (277 files)536.1 GB (308.3 GB on request)MANC v1.2.1bucket not listed by ussize not stated (not checked this time)Larva L1 (Winding et al.)supplement not reachablesize not stated (not checked this time)what our Build your own guide needsone filea whole releaseon request onlysize not stated How big is the download? Fly connectome files, by release. Log scale, sizes from storage metadata, checked 6 Oct 2026 without downloading the data. Our beginner path: 104.1 MB; FlyWire annotation table: 31.7 MB; hemibrain v1.2 summary: 45.9 MB; FlyWire v783, official: 10.6 GB; MaleCNS v1.0 flat files: 31.3 GB; BANC deposit: 536.1 GB (227.8 GB open, 308.3 GB on request); MANC v1.2.1: size not stated; Larva L1 (Winding et al.): size not stated.Size, log scaleOur beginner path104.1 MBAnnotation table31.7 MBhemibrain v1.245.9 MBFlyWire v783, all10.6 GBMaleCNS v1.0, all31.3 GBBANC deposit536.1 GB in all,308.3 GB on requestMANC v1.2.1size not statedLarva L1size not stated10 MB1 GB100 GBSizes from storagemetadata, 6 Oct 2026;nothing downloaded.
How big is the download? Fly connectome files, by release. Sizes in decimal units (1 GB = 1,000,000,000 bytes) from each host's storage metadata (HTTP HEAD answers and deposit listings), checked on 6 Oct 2026; we downloaded none of these data files for this table. The beginner path is 104.1 MB; a whole official release is 100 to 5,000 times bigger. MANC and the larva L1 sizes are not stated: we did not list the MANC bucket (our request limit was reached) and the larva supplement page did not answer.
Download facts by release, checked 6 Oct 2026 with 15 HEAD requests and 6 listing requests (403 files in total); no data file was downloaded. Sizes are decimal. Checksums are the host's own storage metadata; none of the hosts publishes a separate checksum file. One BANC file (the v3 edge list) answered our HEAD with 403 although the listing marks it open.
ReleaseFilesSizeChecksumLicenceAccess
Our beginner path: the Shiu et al. repository copies at commit 91bdd1e72104.1 MB
Connectivity_783.parquet 100,804,642 B; Completeness_783.csv 3,327,347 B
None published; our sha256: efeb23fb… and bbb847a4…FlyWire data (the stricter reading: CC BY-NC 4.0); the repository's MIT licence covers the code, not the data.Open
FlyWire FAFB v783, the official files (Zenodo)510.6 GB
every synapse 9.5 GB; proofread connections 852,022,274 B
MD5 per file (Zenodo)Zenodo record: CC BY 4.0; flywire.ai guidelines: CC BY-NC 4.0. Conflict, unresolved: we follow the stricter, non-commercial reading.Open
FlyWire Codex download page–not checked–Zenodo record: CC BY 4.0; flywire.ai guidelines: CC BY-NC 4.0. Conflict, unresolved: we follow the stricter, non-commercial reading.Sign-in; the files listed there were not read
FlyWire neuron annotations (Schlegel et al., commit a83b2776)131.7 MB
31,720,298 B
None published; our sha256: b214970b…Not stated in the repository; the FlyWire data behind it: as above.Open
MaleCNS v1.0 flat connectome (brain and nerve cord of a male fly)1131.3 GB
wiring only (traced): 508,025,642 B
MD5 + CRC32C (Google Cloud Storage)CC BY 4.0Open
BANC v888 (brain and nerve cord, Harvard Dataverse, version 3.0)379536.1 GB
227.8 GB open (102 files); 308.3 GB in 277 influence-matrix chunks on request
MD5 per file (Dataverse)CC BY 4.0102 open, 277 on request
hemibrain v1.2 compact connection summary145.9 MB
45,872,577 B; the page's newest flat file is v1.2, neuPrint serves v1.2.1
MD5 + CRC32C (storage header)CC BY 4.0Open
MANC v1.2.1 (male nerve cord)–Not stated: the bucket was not listed (our request limit was reached)Not checkedCC BY 4.0Open (public bucket)
FANC (female nerve cord)–No public bulk file–No open licence found (community access rules)Community rules
Larva L1 brain (Winding et al. 2023, Science)–Not stated: the supplement page answered with a robot check, the mirror timed outNone publishedArticle CC BY 4.0; no separate data licence foundOpen

The FlyWire licence question is open. The Zenodo record of FlyWire v783 says CC BY 4.0 (reuse with credit, also commercially). FlyWire's own guidelines page says its public release data is CC BY-NC 4.0 (credit, and non-commercial use only). We have not resolved the conflict, so we follow the stricter reading: non-commercial, with credit, until FlyWire says otherwise. This is not legal advice. MaleCNS, BANC, hemibrain and MANC state CC BY 4.0.

Neuron models and engines

The neuron model decides how activity moves through the wiring. Most projects use the Shiu et al. LIF model or a copy of it on faster hardware. Numbers below come from the code and from the authors; we did not re-run them.

Brain models and simulation engines in the catalogue.
Model or engineWhat it isDataKey facts from the codeGrade
Shiu et al. Drosophila_brain_modelWhole-brain LIF model in Brian2FlyWire v630 (default); v783 files includedThreshold −45 mV, reset −52 mV, membrane time constant 20 ms, one weight scale of 0.275 mV per synapse. Python 3.10 and Brian2 2.5.1. Last commit 14 Sep 2024A
drosophila-brain-mlxThe same model on Apple MLX (Mac GPU)FlyWire v630; MaleCNS v1.0With real wiring MN9 fires at 67.30 Hz; with shuffled wiring at 0 Hz. About 0.29 s per simulated second on an M4 Pro, against 2.07 s for Brian2 (author's figures)A
flyvisTrained rate model of the fly's motion-vision pathwayFlyEM FIB-25 and FIB-19 optic-lobe columns (not FlyWire)Synapse counts and signs fixed from the connectome; time constants and some scales trained on optic flowA
Eon fly-brainThe Shiu model on six engines, with speed comparisonsFlyWire v783Engines agree closely (16,418 to 17,429 spikes for the same 1 s sugar test). Brain only; no body coden/a
fasterflyGPU kernels that simulate many flies at onceMaleCNS v1.051,159 fly-steps per second for 64 flies on an RTX 3090 (author). Own simplified constants; no licence filen/a
dotFlyC# and .NET version of the Shiu LIF modelFlyWire v630 and v783; MaleCNS v1.0Matches Brian2 spike for spike in its tests; its room demo is mostly a hand-written decoderC
Loihi 2 fly brain (Sandia)The Shiu model on 12 neuromorphic Loihi 2 chipsFlyWireCompared with Brian2 only; about 3 to 350 times faster in the preprint. No code link foundU

Bodies: NeuroMechFly (FlyGym) and flybody

A brain model alone does not move. To walk, groom or fly, a project needs a simulated body with legs, joints and physics, and a way to connect neurons to muscles.

The two open fly bodies used by catalogue projects.
BodyWhat it gives youVersions and requirements (from the code)Licence
NeuroMechFly (FlyGym), EPFLA MuJoCo fly body with vision, smell, leg adhesion and joint sensing. No brain: you write the controller.v2.1.0 (24 Jun 2026) is a full rewrite; code written for FlyGym 1.x will not run without changes (the old API moved to flygym-gymnasium). Python 3.12 to 3.14. Optional NVIDIA GPU extraApache-2.0
flybody, Turaga lab and DeepMindA detailed MuJoCo body with walking and flight policies trained by reinforcement learning. These are ordinary neural networks, not the connectome.Last release v0.1.0 (16 May 2024); last default-branch commit 30 Jul 2025. The TensorFlow extra pins Python 3.10 in practiceApache-2.0

Every embodied connectome demo we inspected needs extra machinery between brain and body: pre-trained or pre-programmed walking controllers, fitted rhythm generators, or trained gains. None of the brain datasets alone makes a fly walk.

Measured: who does the walking when a brain drives a body

On 2 Oct 2026 we measured this split ourselves. We connected the Shiu et al. whole-brain model to FlyGym's walking fly, stimulated the brain's forward-walking neurons (P9) and turned its descending neurons' rates into FlyGym's two-number walking drive with a rule we wrote. Each stage of the loop above did a different part of the work:

Who does what when a fly brain model drives a FlyGym flyWho does what in our Add-a-body test, measured on 2 Oct 2026. 1, stimulus (ours): we drive the forward-walking neurons P9 (DNp09) at 150 Hz; P9 then fires at 130–162 Hz in real and scrambled brains alike. 2, brain (Shiu et al. model on FlyWire v783): the real wiring recruits the turning neurons DNa02, more on the right (21–29 Hz) than the left (7–11 Hz); scrambled brains recruit none; the backward neurons MDN stay at 0 Hz. 3, mapping (ours, hand-made): every 100 ms the rates become two numbers, divided by 100 Hz and capped at 1.2; 90 of the 120 drive values of the six P9 walks (75%) sit exactly at the 1.2 clip. 4, body (FlyGym 2.1.0's walking controller): legs, rhythm and balance, 13–14 mm per second; with no brain, the default drive (1, 1) walks 13.0 mm. Open loop: the body never feeds back to the brain. With sugar instead of P9, MN9 fires at 82 Hz, every walking command is at 0 Hz and the fly stands.1 · STIMULUSWe drive P9OursDNp09 at 150 Hz, bothsides. P9 then fires130–162 Hz in real andscrambled brains alike2 · BRAINAdds turningShiu et al. modelReal wiring recruitsDNa02: right 21–29 Hz,left 7–11 Hz. Scrambled:none. MDN (back) 0 Hz3 · MAPPINGRates → 2 numbersOurs, hand-madeEvery 100 ms, ÷ 100 Hz,capped at 1.2. In theP9 walks 75% of thenumbers sit at the cap4 · BODYWalksFlyGym 2.1.0Legs, rhythm, balance:13–14 mm per second.No brain, drive (1, 1):13.0 mm in 1 sThe simulated brainOpen loop: the body never feeds back to the brain. With sugar instead of P9: MN9 82 Hz, every walking command 0 Hz, the fly stands. Who does what in our Add-a-body test, measured on 2 Oct 2026. 1, stimulus (ours): we drive the forward-walking neurons P9 (DNp09) at 150 Hz; P9 then fires at 130–162 Hz in real and scrambled brains alike. 2, brain (Shiu et al. model on FlyWire v783): the real wiring recruits the turning neurons DNa02, more on the right (21–29 Hz) than the left (7–11 Hz); scrambled brains recruit none; the backward neurons MDN stay at 0 Hz. 3, mapping (ours, hand-made): every 100 ms the rates become two numbers, divided by 100 Hz and capped at 1.2; 90 of the 120 drive values of the six P9 walks (75%) sit exactly at the 1.2 clip. 4, body (FlyGym 2.1.0's walking controller): legs, rhythm and balance, 13–14 mm per second; with no brain, the default drive (1, 1) walks 13.0 mm. Open loop: the body never feeds back to the brain. With sugar instead of P9, MN9 fires at 82 Hz, every walking command is at 0 Hz and the fly stands.1 · STIMULUSWe drive P9OursDNp09 at 150 Hz, bothsides. P9 then fires130–162 Hz in real andscrambled brains alike2 · BRAINAdds turningShiu et al. modelReal wiring recruitsDNa02: right 21–29 Hz,left 7–11 Hz. Scrambled:none. MDN (back) 0 Hz3 · MAPPINGRates → 2 numbersOurs, hand-madeEvery 100 ms, ÷ 100 Hz,capped at 1.2. In theP9 walks 75% of thenumbers sit at the cap4 · BODYWalksFlyGym 2.1.0Legs, rhythm, balance:13–14 mm per second.No brain, drive (1, 1):13.0 mm in 1 sOpen loop. Sugar instead ofP9: MN9 82 Hz, walkingcommands 0 Hz: it stands.
Measured in our Add-a-body test of 2 Oct 2026 (one stimulus, one model, one body, 1 s, open loop). Colours as in the sense → model → act loop on How it works: orange hand-made, blue modelled, green measured, grey dashed engineered or scripted. Details on Does fly wiring help?

The result: the fly walked 13.9 mm in 1 s with the real brain, 14.1 mm with a scrambled brain and 14.3 mm with no brain. The legs, rhythm and balance were FlyGym's; the brain added turning commands. Chart, controls and limits: Who does the walking?; build it: Add a body.

Measured: who does the turning when a looming shadow appears

On 5 Oct 2026 we added a sense. We showed a looming shadow to one eye of the same brain model (every LC4 and LPLC2 neuron of that eye, the fly's looming detectors), held the walking drive constant ourselves, and let the brain's turning neurons steer. This time the brain's wiring decided which side's turning neurons fired; the walking was still FlyGym's, and the rule from rates to legs, which turns that side into a direction, was still ours:

Who does what when a looming shadow turns the FlyGym flyWho does what in our looming test of 5 Oct 2026. 1, stimulus (ours): every LC4 and LPLC2 neuron of one eye (162 on the left, 152 on the right) is driven at 80 Hz for 1 s, the same in real and scrambled brains. 2, brain (Shiu et al. model on FlyWire v783): the giant fibre fires at 111.5–120.5 Hz and the turning neurons DNa02 on the far side at 23–44 Hz, the near side stays at 0 Hz; in scrambled brains all of them stay at 0 Hz. 3, mapping (ours, hand-made): every 100 ms, DNa02 slows the legs on its own side and MDN pushes both sides back, on top of a constant 0.8 walking drive that is ours, not the brain's. 4, body (FlyGym 2.1.0's walking controller): legs, rhythm and balance; the real map turned the fly 34–68° away from the shadow, scrambled maps sent the same drive as no brain (a 1.5° drift). Open loop: the body never feeds back to the eyes. Whether walking flies turn away like this: fly67's claim; the direction in real flies was not verified by us.1 · STIMULUSA shadow, one eyeOursEvery LC4 and LPLC2 ofone eye (162 left, 152right) at 80 Hz for 1 s,the same in every brain2 · BRAINPicks a sideShiu et al. modelGiant fibre 111–121 Hz;DNa02 on the far side23–44 Hz, near side 0.Scrambled: all 0 Hz3 · MAPPINGRates → 2 numbersOurs, hand-madeDNa02 slows its ownside, MDN backs up;a constant 0.8 walkingdrive is ours, too4 · BODYWalks and turnsFlyGym 2.1.0Legs, rhythm, balance.Real map: 34–68° away.Scrambled = no brain:a 1.5° driftThe simulated brainOpen loop: the body never feeds back to the eyes. "Away" rests on our mapping's DNa02 sign and fly67's claim. Who does what in our looming test of 5 Oct 2026. 1, stimulus (ours): every LC4 and LPLC2 neuron of one eye (162 on the left, 152 on the right) is driven at 80 Hz for 1 s, the same in real and scrambled brains. 2, brain (Shiu et al. model on FlyWire v783): the giant fibre fires at 111.5–120.5 Hz and the turning neurons DNa02 on the far side at 23–44 Hz, the near side stays at 0 Hz; in scrambled brains all of them stay at 0 Hz. 3, mapping (ours, hand-made): every 100 ms, DNa02 slows the legs on its own side and MDN pushes both sides back, on top of a constant 0.8 walking drive that is ours, not the brain's. 4, body (FlyGym 2.1.0's walking controller): legs, rhythm and balance; the real map turned the fly 34–68° away from the shadow, scrambled maps sent the same drive as no brain (a 1.5° drift). Open loop: the body never feeds back to the eyes. Whether walking flies turn away like this: fly67's claim; the direction in real flies was not verified by us.1 · STIMULUSA shadow, one eyeOursEvery LC4 and LPLC2 ofone eye (162 left, 152right) at 80 Hz for 1 s,the same in every brain2 · BRAINPicks a sideShiu et al. modelGiant fibre 111–121 Hz;DNa02 on the far side23–44 Hz, near side 0.Scrambled: all 0 Hz3 · MAPPINGRates → 2 numbersOurs, hand-madeDNa02 slows its ownside, MDN backs up;a constant 0.8 walkingdrive is ours, too4 · BODYWalks and turnsFlyGym 2.1.0Legs, rhythm, balance.Real map: 34–68° away.Scrambled = no brain:a 1.5° driftOpen loop. "Away" rests onour mapping and onfly67's claim.
Measured in our looming test of 5 Oct 2026 (one stimulus, one model, one body, 1 s, open loop). Colours as in the sense → model → act loop on How it works: orange hand-made, blue modelled, green measured, grey engineered or scripted. Details on Is it real?

The result: the real map turned the fly away in 4 of 4 runs, scrambled maps in 0 of 3 (their drives were the same as no brain). Small n, much quieter scrambled brains, and a direction that rests on our mapping (our caption, fixed before the test: "fly67's claim; the direction in real flies was not verified by us.") Chart and caveats: We ran the map; build it: Add a sense: looming.

14 Oct, updated 17, 18, 19 and 21 Oct: the looming turn in degrees, at each input strength

Our own test · pre-registered before the first body run (14 Oct 2026, 09:28 UTC) · simulated body (FlyGym) driven through our own mapping · one model · a simulation, not an animal

On 8 Oct the brain model's turn command appeared from about 30 spikes per second per looming neuron. On 14 Oct we walked the remaining brain drives of that series in the simulated body, so every drive that differs from the no-brain floor now has a body run. The sentences were chosen by rules fixed in advance; the two eyes read differently, so there is one sentence per eye.

Left eye (pre-registered sentence): In the body, the fly turned away from 30 spikes/s per looming neuron upward on this side, the strength at which the brain's turn command appeared; the turn was about 11 deg at 30 and about 37 deg at 80 (our mapping turns the brain's rates into steering).

Right eye (pre-registered sentence): At 30 spikes/s the turn was borderline: the fly turned away in 1 of 2 runs on this side; from 40 upward it turned away in every run. This borderline rests on one run: seed 0 at 30 spikes/s turned 9.18° beyond the no-brain floor, 0.8° under our 10° rule (seed 2 at 30 turned 16.8°).

These body runs turn the brain's firing rates into steering through our own hand-made mapping, so they show the brain result in degrees; they are not a separate test of the wiring. We ran scrambled brains at 30 and 80 spikes/s, not at 20, 40 or 60.

17 Oct 2026: scrambled brains at 30 spikes/s (pre-registered sentences; brain only, because their steering drives equal the no-brain drive):

17 Oct: at 30 spikes/s, the weakest tested input at which the real brain's turn command appeared, we also ran 2 scrambled brains per side (each neuron keeps its number of connections; one 1-second trial each, with the same trial seed as the real-wiring body runs).

  • Left eye: On the left side, neither scrambled brain sent a turn command, and their steering signals were identical to having no brain, so in the body they walk like the no-brain fly (the real wiring turned it about 11 deg).
  • Right eye: On the right side, neither scrambled brain sent a turn command, and their steering signals were identical to having no brain, so in the body they walk like the no-brain fly (the real wiring turned it about 17 deg).

The scrambled brains were not matched for overall activity: they fired 48% (left) and 44% (right) as many spikes as the real brain at this strength. Most of those spikes were the stimulated input neurons themselves; outside the input, the scrambled brains fired about 1% as many spikes as the real brain.

18 Oct: As at 80 spikes/s, none of the scrambled brains we ran at 30 spikes/s sent a turn command, including those strengthened to about the real brain's activity outside the input. (how we strengthened them, and the limits)

19 Oct, in the body: 19 Oct: walked in the body, the one matched scrambled brain with a small steering signal changed the fly's heading by about 1.4 degrees away from the shadow compared with no brain input, under our 10-degree turn rule; the real brain, with the same input, turned it away by about 16.8 degrees (one run each). This body run turns the brain's firing rates into steering through our own hand-made mapping; it is not a separate test of the wiring. (the run and the limits)

21 Oct, a second input draw: With a second input draw, the matched scrambled brains again sent no turn command (2 input draws on the same 4 scrambled brains). With this input the real brain's turn signal on the right side was 6 spikes/s against our threshold of 5, so the real brain itself sits close to our turn rule at this stimulus strength. In the body: Walked in the body, the matched scrambled brain with a small steering signal changed the fly's heading by about 1.3 degrees away from the shadow compared with no brain input, under our 10-degree turn rule; the real brain, with the same input, turned it away by about 9.2 degrees (one run each). With this input the real brain's own turn was also under our 10-degree rule. This body run turns the brain's firing rates into steering through our own hand-made mapping; it is not a separate test of the wiring. (the sentences, the chart and the limits)

Shadow on the left eye: turn away in degrees beyond the no-brain floor, by input strength10 spikes/s, seed 2: same as no brain (0 deg); 20 spikes/s, seed 2: 1.7 deg, below the 10 deg rule; 20 spikes/s, seed 0: same as no brain (0 deg); 30 spikes/s, seed 2: 11.1 deg, turned away; 30 spikes/s, seed 0: 12.7 deg, turned away; 40 spikes/s, seed 2: 27.2 deg, turned away; 60 spikes/s, seed 2: 26.9 deg, turned away; 80 spikes/s, seed 2: 37.2 deg, turned away; scrambled brains at 30 spikes/s (17 Oct, 2): shuffle 4 same as no brain (0 deg), shuffle 5 same as no brain (0 deg). Dashed line: the 10 degree rule. The table below the charts lists every point.Shadow on the left eyedegrees beyond no-brain floor0°20°40°60°10° rule102030406080spikes/s per neuron10 spikes/s, seed 2: the brain sent the same drive as no brain, so the walk is the no-brain walk (0°)20 spikes/s, seed 2: 1.67° beyond the floor (walked 14 Oct 2026)20 spikes/s, seed 0: the brain sent the same drive as no brain, so the walk is the no-brain walk (0°)30 spikes/s, seed 2: 11.13° beyond the floor (walked 8 Oct 2026)30 spikes/s, seed 0: 12.68° beyond the floor (walked 14 Oct 2026)40 spikes/s, seed 2: 27.18° beyond the floor (walked 14 Oct 2026)60 spikes/s, seed 2: 26.88° beyond the floor (walked 14 Oct 2026)80 spikes/s, seed 2: 37.18° beyond the floor (walked 8 Oct 2026)30 spikes/s, scrambled brain (shuffle 4, trial seed 2, 17 Oct): no turn command; the drive equals the no-brain drive, so the walk is the no-brain walk (0°)30 spikes/s, scrambled brain (shuffle 5, trial seed 2, 17 Oct): no turn command; the drive equals the no-brain drive, so the walk is the no-brain walk (0°) Shadow on the right eye: turn away in degrees beyond the no-brain floor, by input strength10 spikes/s, seed 2: same as no brain (0 deg); 20 spikes/s, seed 2: same as no brain (0 deg); 20 spikes/s, seed 0: 2.5 deg, below the 10 deg rule; 30 spikes/s, seed 2: 16.8 deg, turned away; 30 spikes/s, seed 0: 9.2 deg, below the 10 deg rule; 40 spikes/s, seed 2: 33.8 deg, turned away; 60 spikes/s, seed 2: 62.2 deg, turned away; 80 spikes/s, seed 2: 55.4 deg, turned away; scrambled brains at 30 spikes/s (17 Oct, 2): shuffle 4 same as no brain (0 deg), shuffle 5 same as no brain (0 deg); scrambled, gain-matched, 30 spikes/s (shuffle 4, walked 19 Oct): 1.4 deg, below the 10 deg rule; the same scrambled brain with a second input draw (walked 21 Oct): 1.3 deg, below the 10 deg rule; the real brain with that second draw is the seed 0 point at 30 (9.2 deg, also below the rule). Dashed line: the 10 degree rule. The table below the charts lists every point.Shadow on the right eyedegrees beyond no-brain floor0°20°40°60°10° rule102030406080spikes/s per neuron10 spikes/s, seed 2: the brain sent the same drive as no brain, so the walk is the no-brain walk (0°)20 spikes/s, seed 2: the brain sent the same drive as no brain, so the walk is the no-brain walk (0°)20 spikes/s, seed 0: 2.52° beyond the floor (walked 14 Oct 2026)30 spikes/s, seed 2: 16.78° beyond the floor (walked 8 Oct 2026)30 spikes/s, seed 0: 9.18° beyond the floor (walked 14 Oct 2026)40 spikes/s, seed 2: 33.78° beyond the floor (walked 14 Oct 2026)60 spikes/s, seed 2: 62.15° beyond the floor (walked 14 Oct 2026)80 spikes/s, seed 2: 55.42° beyond the floor (walked 8 Oct 2026)30 spikes/s, scrambled brain (shuffle 4, trial seed 2, 17 Oct): no turn command; the drive equals the no-brain drive, so the walk is the no-brain walk (0°)30 spikes/s, scrambled brain (shuffle 5, trial seed 2, 17 Oct): no turn command; the drive equals the no-brain drive, so the walk is the no-brain walk (0°)30 spikes/s, scrambled brain strengthened to the real brain's activity outside the input (shuffle 4, connections ×2.75, trial seed 2), walked 19 Oct 2026: 1.39° beyond the floor, below the 10° rule30 spikes/s, the same strengthened scrambled brain (shuffle 4, connections ×2.75) with a second input draw (trial seed 0), walked 21 Oct 2026: 1.30° beyond the floor, below the 10° rule; the real brain with this input: 9.18°, also below the rule9.18°
  • seed 2 (line joins its points)
  • seed 0
  • same as no brain
  • scrambled brain, 30 spikes/s (17 Oct)
  • scrambled, gain-matched, 30 spikes/s, right eye (walked 19 Oct)
  • the same brain, second input draw (walked 21 Oct)
  • 10° rule

Horizontal axis: input in spikes per second for every looming neuron of one eye. Vertical axis: degrees turned away beyond the no-brain floor. Each point is one body run: how far the simulated fly turned away from the shadow after it appeared, minus the turn of the same body with no brain input (the floor). Below the dashed line our rule reads "did not turn away". Hollow diamonds at 30 spikes/s: the 17 Oct scrambled brains, drawn at the no-brain value because their drives equal the no-brain drive. Filled triangle (right eye only): the one strengthened scrambled brain with a small steering signal, walked on 19 Oct (1.39°, below the rule; the real brain at the same input 16.78°). Hollow triangle: the same brain with a second input draw, walked on 21 Oct (1.30°, below the rule; the real brain with that input is the seed 0 point at 30, 9.18°, also below the rule). The turn did not grow at every step: seed 2 at 30, 40, 60 and 80 spikes/s turned 11.1, 27.2, 26.9, 37.2° (left eye) and 16.8, 33.8, 62.2, 55.4° (right eye).

Every point in the charts (16 body readings, 4 scrambled brains and 1 strengthened scrambled brain walked on 19 Oct and, with a second input draw, on 21 Oct)
Turn away from the shadow after onset, minus the no-brain floor (−1.49°), in degrees; our 10° rule decides "turned away". "Same as no brain": the brain sent exactly the no-brain drive, so the walk is the no-brain walk. 14 Oct 2026 and 8 Oct 2026 body runs; the 4 scrambled brains (each neuron keeps its number of connections) ran in the brain model on 17 Oct 2026, and their drives equal the no-brain drive. The strengthened scrambled brain (right eye, shuffle 4, connections ×2.75) ran in the brain model on 18 Oct 2026 and was walked in the body on 19 Oct 2026; with a second input draw (trial seed 0) it ran in the brain model and was walked on 21 Oct 2026 (its real-brain reference is the seed 0 row at 30 spikes/s).
Shadow onInput (spikes/s)Brain and seedTurn beyond the floorTurned away (10° rule)Body run
left eye10real, seed 20°nosame as no brain
left eye20real, seed 21.67°nowalked 14 Oct 2026
left eye20real, seed 00°nosame as no brain
left eye30real, seed 211.13°yeswalked 8 Oct 2026
left eye30real, seed 012.68°yeswalked 14 Oct 2026
left eye30scrambled (shuffle 4), trial seed 20°nosame as no brain (brain trial 17 Oct 2026)
left eye30scrambled (shuffle 5), trial seed 20°nosame as no brain (brain trial 17 Oct 2026)
left eye40real, seed 227.18°yeswalked 14 Oct 2026
left eye60real, seed 226.88°yeswalked 14 Oct 2026
left eye80real, seed 237.18°yeswalked 8 Oct 2026
right eye10real, seed 20°nosame as no brain
right eye20real, seed 20°nosame as no brain
right eye20real, seed 02.52°nowalked 14 Oct 2026
right eye30real, seed 216.78°yeswalked 8 Oct 2026
right eye30real, seed 09.18°nowalked 14 Oct 2026
right eye30scrambled (shuffle 4), trial seed 20°nosame as no brain (brain trial 17 Oct 2026)
right eye30scrambled (shuffle 5), trial seed 20°nosame as no brain (brain trial 17 Oct 2026)
right eye40real, seed 233.78°yeswalked 14 Oct 2026
right eye60real, seed 262.15°yeswalked 14 Oct 2026
right eye80real, seed 255.42°yeswalked 8 Oct 2026
right eye30scrambled, gain-matched (shuffle 4, connections ×2.75), trial seed 21.39°nowalked 19 Oct 2026
right eye30scrambled, gain-matched (shuffle 4, connections ×2.75), trial seed 0 (second input draw)1.30°nowalked 21 Oct 2026

Limits and every run: Is it real? · report of 14 Oct 2026 · report of 17 Oct 2026

Where in the map the turn runs

On 6 Oct 2026 we looked inside the brain model (no body). The looming detectors have no direct synapses onto the turning neurons DNa02: their signal reaches the far-side DNa02 through relay neurons, the strongest being PVLP141 on the eye's side. They do synapse directly onto the giant fibre, the escape neuron, of their own side. We then scrambled only the inside of the map, keeping every connection onto the output neurons. In the words we fixed in advance: "Scrambling only the inside of the map kept about -9% of the turn command and 69% of the escape signal." (−9% means none: the far-side command was absent in all 6 trials.) Silencing the giant fibre left the turn command alone, as the map predicts: it has no giant fibre → DNa02 connections. Scrambled maps turned up to the real map's activity still gave no turn command. These are counts and spike rates in one model, not a moving fly. Chart and caveats: Where in the wiring is the turn?

Where the looming signal can go in the mapWhat the FlyWire v783 map itself contains, counted before any 6 Oct trial. The looming neurons of one eye (LC4 and LPLC2: 162 on the left, 152 on the right) have 832 (left) and 1053 (right) direct synapses onto the giant fibre of their own side and none onto the other side's. They have no direct synapses onto DNa02 on either side. Their two-step routes to the far-side DNa02 run through 21 (left) and 23 (right) relay neurons; the top relay on both sides is PVLP141 on the same side as the eye, with 361 and 451 synapses in from the looming neurons and 98 and 81 out to the far-side DNa02. The inside-only scramble keeps the looming neurons' synapses onto output neurons (13% of them, including the giant fibre) and scrambles the other 87%, including the first step to every relay. The giant fibres have no synapses onto DNa02 at all.The map's own counts for one eye (left / right eye), counted before the first trialLooming neuronsLC4 + LPLC2 of one eye162 / 152 cellsRelays (21 / 23)top: PVLP141, same side361 / 451 synapses inDNa02, far sidethe turn command98 / 81 synapses from PVLP141Giant fibre, same sidethe escape signal (DNp01)832 / 1,053 direct synapsesno direct synapses: 0fibre → DNa02: 0inside-only scramble breaks this first stepand keeps synapses onto output neurons (13% ofthe looming outputs, incl. the giant fibre)Source: FlyWire v783 (Connectivity_783.parquet at Shiu et al. 91bdd1e7), our census of 6 Oct 2026 (descriptive). What the FlyWire v783 map itself contains, counted before any 6 Oct trial. The looming neurons of one eye (LC4 and LPLC2: 162 on the left, 152 on the right) have 832 (left) and 1053 (right) direct synapses onto the giant fibre of their own side and none onto the other side's. They have no direct synapses onto DNa02 on either side. Their two-step routes to the far-side DNa02 run through 21 (left) and 23 (right) relay neurons; the top relay on both sides is PVLP141 on the same side as the eye, with 361 and 451 synapses in from the looming neurons and 98 and 81 out to the far-side DNa02. The inside-only scramble keeps the looming neurons' synapses onto output neurons (13% of them, including the giant fibre) and scrambles the other 87%, including the first step to every relay. The giant fibres have no synapses onto DNa02 at all.The map's own counts(left / right eye)Looming neuronsLC4 + LPLC2, one eye162 / 152 cellsscrambledRelays (21 / 23)top: PVLP141, same side361 / 451 synapses inDNa02, far sidethe turn command98 / 81 from PVLP141No direct synapses fromthe looming neurons toDNa02, and none fromthe giant fibre.Giant fibresame side: the escapesignal; 832 / 1,053direct synapses, keptby the inside-onlyscrambleFlyWire v783, our censusof 6 Oct 2026(descriptive).
Descriptive, counted from the map before the first 6 Oct trial (it explains the readings; it does not change them). By the pre-registered disclosure rule, the inside-only turn reading has no "kept by construction" part, and its escape reading is "partly kept by construction" (the 832 / 1,053 direct synapses). A targeted test silenced PVLP141, and then all two-step relays, on 8 Oct 2026 (what happened).

13 Oct, updated 15 Oct: the relays against an activity-matched control

Our own test · pre-registered before the first trial (13 Oct 2026, 10:18 UTC; the 5 added right-side controls on 15 Oct 2026, 09:38 UTC) · brain only · one model · not a real fly

The pre-registered sentence, chosen by the rule fixed in advance:

The two sides disagree: on one side silencing the relays lowered the turn neuron's rate more than matched activity elsewhere, on the other it did not. We do not draw a conclusion.

15 Oct 2026 (pre-registered update line): 15 Oct: we ran 5 more matched controls on the right side, 10 in all. Every one of them left the turn neuron's rate higher than silencing the relays did (controls 38-47 Hz, relays 25.7 Hz), so the right-side reading did not change.

Our rule: Our rule compares the relays with the lowest matched control, so extra controls can only make the relays look less special, never more. The 15 Oct trials and checks

23 Oct 2026 (pre-registered before the trials; brain only): 23 Oct: our earlier controls matched each relay neuron's activity, but not how strongly it connects to the turn neuron. This time the control is a group of other active neurons that excite the turn neuron, chosen so that together they fire about as much as the active relays and send about as many synapses to the turn neuron (both within 20%).

  • Left eye: in 1 of 5 matched groups the turn neuron's rate fell as low as, or lower than, with the relays silenced (matched groups 13-31 Hz, relays 16.0 Hz, nothing silenced 21-25 Hz). By our pre-registered rule, this test does not show that the relays matter more than other inputs with the same combined connection strength and activity.
  • Right eye: among the 9 other active neurons that excite the turn neuron, no group came within 20% of the relays on both combined activity and synapse count (we had set 5 groups as the minimum in advance), so this control could not be built for this eye.
  • Our rule: Our rule compares the relays with the lowest matched group, so extra groups can only make the relays look less special, never more.
  • Left eye: outside the stimulated looming neurons, the brain fired 13,581-14,040 spikes with a matched group silenced, against 13,901-14,760 with the relays silenced and 14,502-15,172 with nothing silenced.
  • Limits: One model, brain only, one stimulus strength (80 spikes/s), 5 matched groups on the left eye only, each run once; the 5 groups all share the same three most active neurons, so they are not independent draws; not a real fly.

The 23 Oct trials and checks

In the body, 14 Oct 2026 (descriptive, not a reading of the relay question above; shadow at 80 spikes/s; our own mapping):

  • Left eye: With PVLP141 silenced, the fly still turned away, by about 24 deg instead of 37 deg (one run per side; our mapping).
  • Left eye: With all 21-23 two-step relay neurons silenced, the fly still turned away, by about 37 deg instead of 37 deg (one run per side; our mapping). Unrounded: 37.3° vs 37.2°.
  • Right eye: With PVLP141 silenced, the fly still turned away, by about 34 deg instead of 55 deg (one run per side; our mapping).
  • Right eye: With all 21-23 two-step relay neurons silenced, the fly still turned away, by about 40 deg instead of 55 deg (one run per side; our mapping).

These body runs turn the brain's firing rates into steering through our own hand-made mapping, so they show the brain result in degrees; they are not a separate test of the wiring. We ran scrambled brains at 30 and 80 spikes/s, not at 20, 40 or 60.

What is new. On 9 Oct we silenced the relay neurons between the looming-sensitive cells and the turn neuron (DNa02) and compared that with silencing randomly picked active neurons. That random set removed 2-6 times as much activity as the relays (2.3 times for a shadow on the left eye, 5.8 times on the right), so the two were not like for like. This time each silenced relay neuron got its own partner outside the relays that fired about as often in the same trial (within 20% of its spike count), and the control silences those partners instead. Both arms of the comparison now remove about the same amount of activity.

Shadow on the left eye: turn neuron (right DNa02) rate in HzNothing silenced 24, 25, 25, 25, 21 Hz (band 21-25); matched control, 5 draws: 19, 24, 19, 34, 13 Hz (band 13-34); relays E2 15, 14, 19 Hz, mean 16.0, reading: fell, like matched activity; PVLP141 alone 19, 17, 17 Hz, mean 17.7, reading: fell, like matched activity; matched group (23 Oct), 5 draws: 22, 22, 13, 23, 31 Hz (band 13-31), 1 of 5 at or below the relays.Shadow on the left eyeturn neuron = right DNa02, Hz01020304050Nothing silencedNothing silenced: 24 HzNothing silenced: 25 HzNothing silenced: 25 HzNothing silenced: 25 HzNothing silenced: 21 HzMatched controlMatched control: 19 HzMatched control: 24 HzMatched control: 19 HzMatched control: 34 HzMatched control: 13 HzRelays (E2)Relays (E2): 15 HzRelays (E2): 14 HzRelays (E2): 19 HzRelays (E2): mean 16.0 HzPVLP141 alonePVLP141 alone: 19 HzPVLP141 alone: 17 HzPVLP141 alone: 17 HzPVLP141 alone: mean 17.7 HzMatched group23 Oct, 5 drawsMatched group (23 Oct): 22 Hz (draw 0)Matched group (23 Oct): 22 Hz (draw 1)Matched group (23 Oct): 13 Hz (draw 2)Matched group (23 Oct): 23 Hz (draw 3)Matched group (23 Oct): 31 Hz (draw 4)Hz in a 1-second shadow Shadow on the right eye: turn neuron (left DNa02) rate in HzNothing silenced 43, 44, 41, 40, 49 Hz (band 40-49); matched control, 10 draws: 40, 41, 41, 43, 40, 43, 47, 38, 47, 46 Hz (band 38-47); relays E2 26, 25, 26 Hz, mean 25.7, reading: fell more than matched activity; PVLP141 alone 26, 31, 24 Hz, mean 27.0, reading: fell more than matched activity; matched group (23 Oct): could not be built, no group of other active inputs matched the relays.Shadow on the right eyeturn neuron = left DNa02, Hz01020304050Nothing silencedNothing silenced: 43 HzNothing silenced: 44 HzNothing silenced: 41 HzNothing silenced: 40 HzNothing silenced: 49 HzMatched controlMatched control: 40 Hz (draw 0, 13 Oct)Matched control: 41 Hz (draw 1, 13 Oct)Matched control: 41 Hz (draw 2, 13 Oct)Matched control: 43 Hz (draw 3, 13 Oct)Matched control: 40 Hz (draw 4, 13 Oct)Matched control: 43 Hz (draw 5, 15 Oct)Matched control: 47 Hz (draw 6, 15 Oct)Matched control: 38 Hz (draw 7, 15 Oct)Matched control: 47 Hz (draw 8, 15 Oct)Matched control: 46 Hz (draw 9, 15 Oct)Relays (E2)Relays (E2): 26 HzRelays (E2): 25 HzRelays (E2): 26 HzRelays (E2): mean 25.7 HzPVLP141 alonePVLP141 alone: 26 HzPVLP141 alone: 31 HzPVLP141 alone: 24 HzPVLP141 alone: mean 27.0 HzMatched group23 Octno group matchedHz in a 1-second shadow

Blue strip: the turn neuron's rate with nothing silenced (5 seeds). Dashed box and open circles: the matched-control draws (5 on the left; 10 on the right, the 5 added on 15 Oct with a thicker outline). Bottom row (23 Oct): the matched-group draws, shaded circles, left eye only; on the right eye no matched group could be built. Filled dots: the relay arms, one per seed; the bar is their mean. By the pre-registered rule a relay value below both the normal strip and every matched draw reads "fell more than matched activity"; below the strip but inside the matched draws it reads "fell, like matched activity". Left eye: E2 16.0 Hz, "fell, like matched activity". Right eye: E2 25.7 Hz, "fell more than matched activity". PVLP141 alone reads the same as E2 on each side.

Turn neuron rate (Hz) per arm, 13 Oct 2026 test with 5 more right-side matched controls on 15 Oct 2026 (relay seeds from 9 Oct, reused after a bit-exact check) and the group-matched control of 23 Oct 2026
Shadow onNothing silenced (5 seeds)Matched control (5 draws left, 10 right)Relays E2 (3 seeds)PVLP141 alone (3 seeds)Matched group (23 Oct, 5 draws)
Left eye (far side: right DNa02)24, 25, 25, 25, 21
band 21-25
19, 24, 19, 34, 13
band 13-34
15, 14, 19: mean 16.0 (about 67% of normal)
fell, like matched activity
19, 17, 17: mean 17.7 (about 74%)
fell, like matched activity
22, 22, 13, 23, 31
band 13-31; 1 of 5 at or below the relays
Right eye (far side: left DNa02)43, 44, 41, 40, 49
band 40-49
40, 41, 41, 43, 40, 43, 47, 38, 47, 46
band 38-47; the last 5 added 15 Oct
26, 25, 26: mean 25.7 (about 59% of normal)
fell more than matched activity
26, 31, 24: mean 27.0 (about 62%)
fell more than matched activity
could not be built
0 groups among 9 neurons
  • Small samples. 5 seeds per side; 5 matched draws on the left and 10 on the right; the relay arms have 3 seeds each. The left side's matched draws spread widely (13-34 Hz: three below the normal strip, one above it).
  • Some matched partners touch the turn neuron. The matching rule excludes the two-step relays, not every input to the turn neuron: 1 of 5 left-side draws and 2 of 10 right-side draws (none of the 5 added on 15 Oct) include a neuron that synapses directly onto the far-side DNa02.
  • E2 is mostly PVLP141. Of the activity removed by silencing the excitatory relays (E2), PVLP141 accounts for 102 of 149 spikes (left) and 103 of 121 (right). E2 and PVLP141 alone read the same on each side, so this test says little about the other relay neurons.
  • What it measures. A simulated firing rate in one brain-only model (Shiu et al. LIF on FlyWire v783) with a looming input we set; not fly behaviour and not a real fly. Percentages are a value divided by the mean with nothing silenced.

The 9 Oct line stays, with a note. On 9 Oct the pre-registered summary was: Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons. Note added 13 Oct 2026: the random control behind that line removed 2-6 times as much activity as the relays (2.3 times left, 5.8 times right); the activity-matched test above gives the reading on each side. Trials, deviations and code: report of 13 Oct 2026; the 9 Oct design: report of 9 Oct 2026.

Silencing the relays, read by firing rate (8 and 9 Oct)

On 8 Oct 2026 we silenced the top relay, PVLP141, and then all two-step relays of the eye's side (21 or 23 cells), in the same brain model at the full input, and read the result by the left/right contrast. On 9 Oct 2026 a second pre-registered test read the far-side turn neuron's firing rate instead, silenced the excitatory and the inhibitory relays separately, and added a control: the same number of randomly picked active neurons. Paths of three or more steps have not been mapped yet.

The honest headline: Silencing the relays lowered the turn command to about 60-70% of normal, but silencing random active neurons also moved it, so this is not yet specific to the relays.

In the pre-registered words: Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons.

  • The normal range is narrow. For a shadow on the left eye it is only 1 Hz wide (24–25 Hz, three runs with nothing silenced); on the right it is 41–44 Hz. One random-neuron run at 26 Hz, just above the left range, is what triggered the summary above.
  • The random control is one draw per eye (17 and 20 active neurons, one run each). More draws and more runs with nothing silenced are needed before any relay is called specific.
  • What 8 Oct said, and how it reads now. On 8 Oct we read the same kind of test by the left/right contrast: With one relay neuron (PVLP141) silenced, the turn command stayed (by the left/right contrast (8 Oct)). With all 21-23 relay neurons between the looming neurons and the turn neuron silenced, the turn command stayed (by the left/right contrast (8 Oct)). Read by firing rate instead: Read by firing rate, our 8 Oct test with all 21 (left eye) / 23 (right eye) relay neurons silenced raised it to about 105% of normal for a shadow on the left eye, and lowered it to about 65% of normal on the right (1 run per eye). (the open marks below).
Which relays carry the turn? Read by firing rate: silencing other active neurons also changed the turn commandBrain-only trials at 80 spikes per second per looming neuron, Shiu et al. model on FlyWire v783, 9 Oct 2026 (the all-relays trials from 8 Oct). Turn command = how often the turn neuron on the far side (DNa02) fires during the 1-second shadow. Normal range = our three runs per eye with nothing silenced. Far-side DNa02 in Hz, one value per run, left eye / right eye: nothing silenced 24, 25, 25 / 43, 44, 41 (normal range 24–25 and 41–44 Hz); PVLP141 silenced 19, 17, 17 / 26, 31, 24; excitatory relays silenced 15, 14, 19 / 26, 25, 26; inhibitory relays silenced 28, 23, 32 / 55, 52, 52; random active neurons silenced 26 / 42; all relays silenced (8 Oct) 26 / 28. Not shown: excitatory relays without PVLP141 (descriptive only). Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons. With one relay neuron (PVLP141) silenced, the turn command fell to about 70% (left eye) / 65% (right eye) of normal (3 runs per eye). With the 17 (left eye) / 20 (right eye) relay neurons that excite the turn neuron silenced, the turn command fell to about 65% (left eye) / 60% (right eye) of normal (3 runs per eye). With the 4 (left eye) / 3 (right eye) relay neurons that inhibit the turn neuron silenced, the turn command changed, but not clearly beyond its normal range for a shadow on the left eye, and rose to about 125% of normal on the right (3 runs per eye). With 17 (left eye) / 20 (right eye) other active neurons, picked at random, silenced, the turn command rose to about 105% of normal for a shadow on the left eye, and stayed within its normal range on the right (1 run per eye). Read by firing rate, our 8 Oct test with all 21 (left eye) / 23 (right eye) relay neurons silenced raised it to about 105% of normal for a shadow on the left eye, and lowered it to about 65% of normal on the right (1 run per eye). One model (Shiu et al. LIF on FlyWire v783), brain only; the input is a rate we set for every LC4/LPLC2 of one eye, not a measured response to a real shadow; 'silenced' means the neuron's output synapses were removed; the normal range comes from 3 runs per eye; not a real fly.Which relays carry the turn? Read by firing rateDigital Fly LabSilencing other active neurons also changed the turn command, so this relay result is notspecific to the relay neurons.Turn command = how often the turn neuron on the far side (DNa02) fires during the 1-second shadow. Normal range= our three runs per eye with nothing silenced.Shadow on the left eyeshaded: normal range, 24–25 Hz (1 Hz wide)Shadow on the right eyeshaded: normal range, 41–44 HzNormal range, shadow on the left eye: 24–25 Hz (3 runs with nothing silenced)Normal range, shadow on the right eye: 41–44 Hz (3 runs with nothing silenced)Nothing silenced3 runs per eyeNothing silenced, shadow on the left eye, seed 0: far-side DNa02 24 Hz (normal range 24–25 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOML-real-0-f.jsonNothing silenced, shadow on the left eye, seed 1: far-side DNa02 25 Hz (normal range 24–25 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOML-real-1-f.jsonNothing silenced, shadow on the left eye, seed 2: far-side DNa02 25 Hz (normal range 24–25 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOML-real-2-f.json24–25 HzNothing silenced, shadow on the right eye, seed 0: far-side DNa02 43 Hz (normal range 41–44 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOMR-real-0-f.jsonNothing silenced, shadow on the right eye, seed 1: far-side DNa02 44 Hz (normal range 41–44 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOMR-real-1-f.jsonNothing silenced, shadow on the right eye, seed 2: far-side DNa02 41 Hz (normal range 41–44 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOMR-real-2-f.json41–44 HzPVLP1411 relay neuron, 3 runsPVLP141, shadow on the left eye, seed 0: far-side DNa02 19 Hz (normal range 24–25 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOML-real-0-P-f.jsonPVLP141, shadow on the left eye, seed 1: far-side DNa02 17 Hz (normal range 24–25 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOML-real-1-P-f.jsonPVLP141, shadow on the left eye, seed 2: far-side DNa02 17 Hz (normal range 24–25 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOML-real-2-P-f.jsonfell, ~70%PVLP141, shadow on the right eye, seed 0: far-side DNa02 26 Hz (normal range 41–44 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOMR-real-0-P-f.jsonPVLP141, shadow on the right eye, seed 1: far-side DNa02 31 Hz (normal range 41–44 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOMR-real-1-P-f.jsonPVLP141, shadow on the right eye, seed 2: far-side DNa02 24 Hz (normal range 41–44 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOMR-real-2-P-f.jsonfell, ~65%Excitatory relays17 (left) / 20 (right), 3 runsExcitatory relays, shadow on the left eye, seed 0: far-side DNa02 15 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-0-E2-f.jsonExcitatory relays, shadow on the left eye, seed 1: far-side DNa02 14 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-1-E2-f.jsonExcitatory relays, shadow on the left eye, seed 2: far-side DNa02 19 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-2-E2-f.jsonfell, ~65%Excitatory relays, shadow on the right eye, seed 0: far-side DNa02 26 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-0-E2-f.jsonExcitatory relays, shadow on the right eye, seed 1: far-side DNa02 25 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-1-E2-f.jsonExcitatory relays, shadow on the right eye, seed 2: far-side DNa02 26 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-2-E2-f.jsonfell, ~60%Inhibitory relays4 (left) / 3 (right), 3 runsInhibitory relays, shadow on the left eye, seed 0: far-side DNa02 28 Hz (normal range 24–25 Hz), near side 0 Hz; 4 neurons silenced; trial file LOOML-real-0-I2-f.jsonInhibitory relays, shadow on the left eye, seed 1: far-side DNa02 23 Hz (normal range 24–25 Hz), near side 0 Hz; 4 neurons silenced; trial file LOOML-real-1-I2-f.jsonInhibitory relays, shadow on the left eye, seed 2: far-side DNa02 32 Hz (normal range 24–25 Hz), near side 0 Hz; 4 neurons silenced; trial file LOOML-real-2-I2-f.jsonunclear, ~110%Inhibitory relays, shadow on the right eye, seed 0: far-side DNa02 55 Hz (normal range 41–44 Hz), near side 0 Hz; 3 neurons silenced; trial file LOOMR-real-0-I2-f.jsonInhibitory relays, shadow on the right eye, seed 1: far-side DNa02 52 Hz (normal range 41–44 Hz), near side 0 Hz; 3 neurons silenced; trial file LOOMR-real-1-I2-f.jsonInhibitory relays, shadow on the right eye, seed 2: far-side DNa02 52 Hz (normal range 41–44 Hz), near side 0 Hz; 3 neurons silenced; trial file LOOMR-real-2-I2-f.jsonrose, ~125%Random active neurons17 / 20, 1 run per eyeRandom active neurons, shadow on the left eye, seed 2: far-side DNa02 26 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-2-RND-f.jsonrose, ~105%Random active neurons, shadow on the right eye, seed 2: far-side DNa02 42 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-2-RND-f.jsonwithin range (~100%)All relays, 8 Oct21 / 23, 1 run per eye; openAll relays, 8 Oct (8 Oct trial), shadow on the left eye, seed 2: far-side DNa02 26 Hz (normal range 24–25 Hz), near side 0 Hz; 21 neurons silenced; trial file LOOML-real-2-R2-f.jsonrose, ~105%All relays, 8 Oct (8 Oct trial), shadow on the right eye, seed 2: far-side DNa02 28 Hz (normal range 41–44 Hz), near side 0 Hz; 23 neurons silenced; trial file LOOMR-real-2-R2-f.jsonfell, ~65%0204060far-side DNa02, Hz0204060far-side DNa02, HzFilled dot: one run (seed) on 9 Oct; blue: nothing silenced; open: the 8 Oct all-relays run. Not shown: excitatory relays withoutPVLP141 (one run per eye, descriptive only). Near-side DNa02 was 0 Hz in all 30 runs. Labels: reading and % of normal, by the pre-registered rules.With one relay neuron (PVLP141) silenced, the turn command fell to about 70% (left eye) / 65% (right eye) of normal (3runs per eye).With the 17 (left eye) / 20 (right eye) relay neurons that excite the turn neuron silenced, the turn command fell to about65% (left eye) / 60% (right eye) of normal (3 runs per eye).With the 4 (left eye) / 3 (right eye) relay neurons that inhibit the turn neuron silenced, the turn command changed, butnot clearly beyond its normal range for a shadow on the left eye, and rose to about 125% of normal on the right (3 runsper eye).With 17 (left eye) / 20 (right eye) other active neurons, picked at random, silenced, the turn command rose to about 105%of normal for a shadow on the left eye, and stayed within its normal range on the right (1 run per eye).Read by firing rate, our 8 Oct test with all 21 (left eye) / 23 (right eye) relay neurons silenced raised it to about 105%of normal for a shadow on the left eye, and lowered it to about 65% of normal on the right (1 run per eye).One model (Shiu et al. LIF on FlyWire v783), brain only; the input is a rate we set for every LC4/LPLC2 of one eye, not ameasured response to a real shadow; 'silenced' means the neuron's output synapses were removed; the normal range comesfrom 3 runs per eye; not a real fly.Source: Digital Fly Lab, pre-registered 2026-10-09T09:36:50Z. Every run: shaduf.ai/p/digital-fly-catalog/reports/2026-10-09-run11/ Brain-only trials at 80 spikes per second per looming neuron, Shiu et al. model on FlyWire v783, 9 Oct 2026 (the all-relays trials from 8 Oct). Turn command = how often the turn neuron on the far side (DNa02) fires during the 1-second shadow. Normal range = our three runs per eye with nothing silenced. Far-side DNa02 in Hz, one value per run, left eye / right eye: nothing silenced 24, 25, 25 / 43, 44, 41 (normal range 24–25 and 41–44 Hz); PVLP141 silenced 19, 17, 17 / 26, 31, 24; excitatory relays silenced 15, 14, 19 / 26, 25, 26; inhibitory relays silenced 28, 23, 32 / 55, 52, 52; random active neurons silenced 26 / 42; all relays silenced (8 Oct) 26 / 28. Not shown: excitatory relays without PVLP141 (descriptive only). Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons. With one relay neuron (PVLP141) silenced, the turn command fell to about 70% (left eye) / 65% (right eye) of normal (3 runs per eye). With the 17 (left eye) / 20 (right eye) relay neurons that excite the turn neuron silenced, the turn command fell to about 65% (left eye) / 60% (right eye) of normal (3 runs per eye). With the 4 (left eye) / 3 (right eye) relay neurons that inhibit the turn neuron silenced, the turn command changed, but not clearly beyond its normal range for a shadow on the left eye, and rose to about 125% of normal on the right (3 runs per eye). With 17 (left eye) / 20 (right eye) other active neurons, picked at random, silenced, the turn command rose to about 105% of normal for a shadow on the left eye, and stayed within its normal range on the right (1 run per eye). Read by firing rate, our 8 Oct test with all 21 (left eye) / 23 (right eye) relay neurons silenced raised it to about 105% of normal for a shadow on the left eye, and lowered it to about 65% of normal on the right (1 run per eye). One model (Shiu et al. LIF on FlyWire v783), brain only; the input is a rate we set for every LC4/LPLC2 of one eye, not a measured response to a real shadow; 'silenced' means the neuron's output synapses were removed; the normal range comes from 3 runs per eye; not a real fly.Which relays carrythe turn? Read byfiring rateSilencing otheractive neurons alsochanged the turncommand, so thisrelay result is notspecific to therelay neurons.Turn command = how oftenthe turn neuron on thefar side (DNa02) firesduring the 1-secondshadow. Normal range =our three runs per eyewith nothing silenced.Left-eye shadowshaded: normal range24–25 Hz (1 Hz wide)Nothing silenced24–25 HzNothing silenced, shadow on the left eye, seed 0: far-side DNa02 24 Hz (normal range 24–25 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOML-real-0-f.jsonNothing silenced, shadow on the left eye, seed 1: far-side DNa02 25 Hz (normal range 24–25 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOML-real-1-f.jsonNothing silenced, shadow on the left eye, seed 2: far-side DNa02 25 Hz (normal range 24–25 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOML-real-2-f.jsonPVLP141 (1 cell)fell, ~70%PVLP141, shadow on the left eye, seed 0: far-side DNa02 19 Hz (normal range 24–25 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOML-real-0-P-f.jsonPVLP141, shadow on the left eye, seed 1: far-side DNa02 17 Hz (normal range 24–25 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOML-real-1-P-f.jsonPVLP141, shadow on the left eye, seed 2: far-side DNa02 17 Hz (normal range 24–25 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOML-real-2-P-f.jsonExcitatory relays 17/20fell, ~65%Excitatory relays, shadow on the left eye, seed 0: far-side DNa02 15 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-0-E2-f.jsonExcitatory relays, shadow on the left eye, seed 1: far-side DNa02 14 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-1-E2-f.jsonExcitatory relays, shadow on the left eye, seed 2: far-side DNa02 19 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-2-E2-f.jsonInhibitory relays 4/3unclear, ~110%Inhibitory relays, shadow on the left eye, seed 0: far-side DNa02 28 Hz (normal range 24–25 Hz), near side 0 Hz; 4 neurons silenced; trial file LOOML-real-0-I2-f.jsonInhibitory relays, shadow on the left eye, seed 1: far-side DNa02 23 Hz (normal range 24–25 Hz), near side 0 Hz; 4 neurons silenced; trial file LOOML-real-1-I2-f.jsonInhibitory relays, shadow on the left eye, seed 2: far-side DNa02 32 Hz (normal range 24–25 Hz), near side 0 Hz; 4 neurons silenced; trial file LOOML-real-2-I2-f.jsonRandom neurons (1 run)rose, ~105%Random active neurons, shadow on the left eye, seed 2: far-side DNa02 26 Hz (normal range 24–25 Hz), near side 0 Hz; 17 neurons silenced; trial file LOOML-real-2-RND-f.jsonAll relays, 8 Oct (open)rose, ~105%All relays, 8 Oct (8 Oct trial), shadow on the left eye, seed 2: far-side DNa02 26 Hz (normal range 24–25 Hz), near side 0 Hz; 21 neurons silenced; trial file LOOML-real-2-R2-f.json0204060far-side DNa02, HzRight-eye shadowshaded: normal range41–44 HzNothing silenced41–44 HzNothing silenced, shadow on the right eye, seed 0: far-side DNa02 43 Hz (normal range 41–44 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOMR-real-0-f.jsonNothing silenced, shadow on the right eye, seed 1: far-side DNa02 44 Hz (normal range 41–44 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOMR-real-1-f.jsonNothing silenced, shadow on the right eye, seed 2: far-side DNa02 41 Hz (normal range 41–44 Hz), near side 0 Hz; 0 neurons silenced; trial file LOOMR-real-2-f.jsonPVLP141 (1 cell)fell, ~65%PVLP141, shadow on the right eye, seed 0: far-side DNa02 26 Hz (normal range 41–44 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOMR-real-0-P-f.jsonPVLP141, shadow on the right eye, seed 1: far-side DNa02 31 Hz (normal range 41–44 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOMR-real-1-P-f.jsonPVLP141, shadow on the right eye, seed 2: far-side DNa02 24 Hz (normal range 41–44 Hz), near side 0 Hz; 1 neurons silenced; trial file LOOMR-real-2-P-f.jsonExcitatory relays 17/20fell, ~60%Excitatory relays, shadow on the right eye, seed 0: far-side DNa02 26 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-0-E2-f.jsonExcitatory relays, shadow on the right eye, seed 1: far-side DNa02 25 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-1-E2-f.jsonExcitatory relays, shadow on the right eye, seed 2: far-side DNa02 26 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-2-E2-f.jsonInhibitory relays 4/3rose, ~125%Inhibitory relays, shadow on the right eye, seed 0: far-side DNa02 55 Hz (normal range 41–44 Hz), near side 0 Hz; 3 neurons silenced; trial file LOOMR-real-0-I2-f.jsonInhibitory relays, shadow on the right eye, seed 1: far-side DNa02 52 Hz (normal range 41–44 Hz), near side 0 Hz; 3 neurons silenced; trial file LOOMR-real-1-I2-f.jsonInhibitory relays, shadow on the right eye, seed 2: far-side DNa02 52 Hz (normal range 41–44 Hz), near side 0 Hz; 3 neurons silenced; trial file LOOMR-real-2-I2-f.jsonRandom neurons (1 run)within range (~100%)Random active neurons, shadow on the right eye, seed 2: far-side DNa02 42 Hz (normal range 41–44 Hz), near side 0 Hz; 20 neurons silenced; trial file LOOMR-real-2-RND-f.jsonAll relays, 8 Oct (open)fell, ~65%All relays, 8 Oct (8 Oct trial), shadow on the right eye, seed 2: far-side DNa02 28 Hz (normal range 41–44 Hz), near side 0 Hz; 23 neurons silenced; trial file LOOMR-real-2-R2-f.json0204060far-side DNa02, HzBlue: nothing silenced.Open: the 8 Oct run. Notshown: excitatory relayswithout PVLP141(descriptive only).One model (Shiu et al.LIF on FlyWire v783),brain only; the input isa rate we set for everyLC4/LPLC2 of one eye,not a measured responseto a real shadow;'silenced' means theneuron's output synapseswere removed; the normalrange comes from 3 runsper eye; not a real fly.Digital Fly Lab9 Oct 2026

Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons.

Turn command = how often the turn neuron on the far side (DNa02) fires during the 1-second shadow. Normal range = our three runs per eye with nothing silenced.

  • With one relay neuron (PVLP141) silenced, the turn command fell to about 70% (left eye) / 65% (right eye) of normal (3 runs per eye).
  • With the 17 (left eye) / 20 (right eye) relay neurons that excite the turn neuron silenced, the turn command fell to about 65% (left eye) / 60% (right eye) of normal (3 runs per eye).
  • With the 4 (left eye) / 3 (right eye) relay neurons that inhibit the turn neuron silenced, the turn command changed, but not clearly beyond its normal range for a shadow on the left eye, and rose to about 125% of normal on the right (3 runs per eye).
  • With 17 (left eye) / 20 (right eye) other active neurons, picked at random, silenced, the turn command rose to about 105% of normal for a shadow on the left eye, and stayed within its normal range on the right (1 run per eye).
  • Read by firing rate, our 8 Oct test with all 21 (left eye) / 23 (right eye) relay neurons silenced raised it to about 105% of normal for a shadow on the left eye, and lowered it to about 65% of normal on the right (1 run per eye).

One model (Shiu et al. LIF on FlyWire v783), brain only; the input is a rate we set for every LC4/LPLC2 of one eye, not a measured response to a real shadow; 'silenced' means the neuron's output synapses were removed; the normal range comes from 3 runs per eye; not a real fly.

Not shown in the chart: the excitatory relays without PVLP141 (16 left / 19 right, one run per eye: 30 and 41 Hz), descriptive only and never a reading. Pre-registered 2026-10-09T09:36:50Z, before the first trial; every run and the deviations: report of 9 Oct 2026.

Numbers behind the chart (30 brain runs)
Brain-only runs at 80 spikes per second per looming neuron (one model, 1 s each), 9 Oct 2026; the all-relays runs are from 8 Oct and two runs with nothing silenced are reused from 6 Oct (bit-exact). Turn command = far-side DNa02 rate. Reading and % of normal follow the pre-registered rules (percent rounded to the nearest 5). MDN: the backward-walking neurons.
Trial fileShadow onSilencedSeedDNa02 far / near (Hz)Reading (eye)Giant fibre meanMDN meanWhole-brain spikesActive neurons
LOOML-real-0-f.jsonleft eyenothing024 / 0normal range 24–25116.52.2527,467685
LOOML-real-1-f.jsonleft eyenothing125 / 0normal range 24–25111.5027,187682
LOOML-real-2-f.jsonleft eyenothing225 / 0normal range 24–25116.51.2528,153745
LOOML-real-0-P-f.jsonleft eyePVLP141 (1)019 / 0fell, ~70%116027,413747
LOOML-real-1-P-f.jsonleft eyePVLP141 (1)117 / 0fell, ~70%115027,346740
LOOML-real-2-P-f.jsonleft eyePVLP141 (1)217 / 0fell, ~70%113.5027,438709
LOOML-real-0-E2-f.jsonleft eyeexcitatory two-step relays (17)015 / 0fell, ~65%115.5026,973699
LOOML-real-1-E2-f.jsonleft eyeexcitatory two-step relays (17)114 / 0fell, ~65%111.5026,575653
LOOML-real-2-E2-f.jsonleft eyeexcitatory two-step relays (17)219 / 0fell, ~65%114.5027,741769
LOOML-real-0-I2-f.jsonleft eyeinhibitory two-step relays (4)028 / 0unclear, ~110%117.5027,563675
LOOML-real-1-I2-f.jsonleft eyeinhibitory two-step relays (4)123 / 0unclear, ~110%1151.2527,446725
LOOML-real-2-I2-f.jsonleft eyeinhibitory two-step relays (4)232 / 0unclear, ~110%1150.528,196732
LOOML-real-2-RND-f.jsonleft eyerandom active neurons (17)226 / 0rose, ~105%113.50.527,711660
LOOML-real-2-R2-f.jsonleft eyeall two-step relays (8 Oct) (21)226 / 0rose, ~105%116.5027,991772
LOOML-real-2-E2mP-f.jsonleft eyeexcitatory relays without PVLP141 (descriptive only) (16)230 / 0descriptive only116128,244737
LOOMR-real-0-f.jsonright eyenothing043 / 0normal range 41–441202.7527,446659
LOOMR-real-1-f.jsonright eyenothing144 / 0normal range 41–44122.53.7527,252675
LOOMR-real-2-f.jsonright eyenothing241 / 0normal range 41–44116.5527,693677
LOOMR-real-0-P-f.jsonright eyePVLP141 (1)026 / 0fell, ~65%121026,631635
LOOMR-real-1-P-f.jsonright eyePVLP141 (1)131 / 0fell, ~65%117.5026,572655
LOOMR-real-2-P-f.jsonright eyePVLP141 (1)224 / 0fell, ~65%121.5027,008652
LOOMR-real-0-E2-f.jsonright eyeexcitatory two-step relays (20)026 / 0fell, ~60%120026,739655
LOOMR-real-1-E2-f.jsonright eyeexcitatory two-step relays (20)125 / 0fell, ~60%118.5026,351648
LOOMR-real-2-E2-f.jsonright eyeexcitatory two-step relays (20)226 / 0fell, ~60%122026,947640
LOOMR-real-0-I2-f.jsonright eyeinhibitory two-step relays (3)055 / 0rose, ~125%121.5627,896702
LOOMR-real-1-I2-f.jsonright eyeinhibitory two-step relays (3)152 / 0rose, ~125%123.5527,551686
LOOMR-real-2-I2-f.jsonright eyeinhibitory two-step relays (3)252 / 0rose, ~125%1214.528,032668
LOOMR-real-2-RND-f.jsonright eyerandom active neurons (20)242 / 0within normal range, ~100%1050.2523,586572
LOOMR-real-2-R2-f.jsonright eyeall two-step relays (8 Oct) (23)228 / 0fell, ~65%123.5027,016677
LOOMR-real-2-E2mP-f.jsonright eyeexcitatory relays without PVLP141 (descriptive only) (19)241 / 0descriptive only119.5627,799684

Described, never read: the turn rule passed and the near-side DNa02 was 0 Hz in all 30 runs; giant fibre 113.5–122 Hz except the right-eye random run (105 Hz); MDN 0 Hz in every PVLP141 and excitatory-relay run. Arm readings: PVLP141 fell, excitatory relays fell, inhibitory relays and random neurons "sides differ".

Search published pools, pages, reports, and evidence.