Digital Fly Lab · independent catalogue
Fruit fly brain simulation projects: what you can try, what is real, and how to build your own
A fruit fly brain now fits in a computer. Scientists mapped the wiring of every neuron in an adult fly brain (the connectome), and people quickly started to run it in games, browser demos and robots. This catalogue lists those projects, checks their links and dates, and says plainly which parts come from the real fly and which parts are ordinary code.
Short answer
People have built games, browser demos, desktop pets, art, brain engines and research tests on top of real fly connectomes: mostly the FlyWire brain (2024) and the MaleCNS brain and nerve cord (2026). We checked 213 of them. You can open 62 in a browser or a free notebook today, without installing anything, and in our own browser checks 8 of 13 demos ran on a computer without a GPU at their latest check (the first six were re-checked on 17 Oct, and fly67 at its current address and Fly Worker again on 18 Oct).
Most viral demos really do run a simulated fly brain, but a person chose which neurons see the game and which neurons press the buttons, and in several demos scripts or trained parts make most of the movement. No fly has been "uploaded", and several viral claims publish no code at all, among them a fly said to play Gorilla Tag and a fly brain model said to "learn to hack" (the most-viewed one, Pokémon on a Raspberry Pi, published its code on 14 Oct: grade B, borderline C, because a hand-written arbiter, chat keys and a paid override shape the key presses). Is it a brain or just a map? It is a map of one fly's wiring, run with a simple neuron model, and in our own test the map's wiring made a simulated fly turn away from a looming shadow, where a scrambled map did not, and scrambled maps turned up to the same activity still produced no turn command (one model, small n, our own mapping). Our planned wiring test on a trained fly vision model (flyvis) has run out of memory twice (10 and 12 Oct), so it has no result yet; on 9 Oct we read the relay neurons by firing rate and listed what the map leaves out, with one number and one source per part (below). Does the real wiring help in general? Sometimes: across 107 control studies it beat scrambled wiring in 36 of the 97 that test it, mostly reflex, sensory and steering circuits, and tied or lost in trained, reservoir and machine-learning tasks. Our own pre-registered tests show how much the answer depends on how the wiring is scrambled. Even where it helps, a brainless baseline often matches the fly: in Doom a brainless autopilot plays about as well as the full fly brain (the doom-fly-control by gabrycina), and in our own test a simulated fly body walked as far with no brain as with the real one.
To build your own, start with the published whole-brain model on a CPU: in our test on 2 CPUs (Linux, 28 minutes), sugar neurons drove the feeding motor neuron to 85.2 Hz, and shuffled wiring dropped it to 0 Hz. We re-tested the beginner path on 2 Nov 2026 (6 of 6 tests, 82.0 Hz, bit-exact with our earlier runs). The model needs two data files, 104.1 MB (what you download), and the guide has two tested deeper steps: connecting the brain to a walking fly body and adding a sense: a looming shadow; since 8 Oct the brain and the body also install together in one Python 3.12 environment (1.1 GB, numpy pinned to 2.3.5). Teaching? Use in class is a 20-minute plan.
In other languages: 한국어 · Русский · Español · 中文 · 日本語: a short note on how to run it, what to download and where the code is (the full guide is in English).
This week · 10 Nov 2026
10 Nov: browser demos re-checked, two new control studies, and a gamebook that replays fly-brain recordings
- Browser demos: Re-checked 10 Nov in a headless browser without a GPU: 4 of 4 of the demos we first checked on 6 Oct ran (20 Oct: 4 of 4); the fifth, neural-canvas, is re-checked separately. Fly Worker, re-checked 10 Nov after a change to its model code: ran by our check's rule, and two screenshots of its game panel taken about 7 seconds apart show its street game view, frame counter and brain-activity dots changing. It downloaded about 8.6 MB of brain data, identified by their web address, a method we fixed before this check. In our browser checks from 3 Oct to 10 Nov 2026, 8 of 13 demos ran on a machine without a GPU at their latest check. (latest check of each demo)
- Correction: Correction, 10 Nov 2026: our earlier overall line said 9 of 13 demos ran. That was the 20 Oct count, and we did not update it after the re-checks of 24 to 31 Oct, in which neural-canvas, flyleno and help-the-fly-escape stopped responding and fly-brain-lulzx was stopped by its memory use. At each demo's latest check, 8 of 13 ran.
- Controls ledger: two studies joined. indian-aman's whole-CNS specificity study (MaleCNS v1.0): the steering readout stays at 4.11 Hz (real 3.98 Hz) when wiring is shuffled within cell type and side, and falls to -0.01 Hz under a whole-CNS shuffle, so the ledger reads mixed; no no-brain baseline was run. Authors' numbers from a GPU run, not re-run by us. Kishandev12's signal study: real wiring spread activity further than degree-preserving rewiring (about 137 vs 107 active downstream neurons, 10 of 10 pairs). This measures how far activity spreads, not a task, so we do not score it as the wiring helping; no no-brain baseline was run. Authors' numbers, not re-run by us. The ledger now has 107 studies, 97 with a wiring null: helps 36, no difference 27, worse 11, mixed 18, not yet scored 5. (what changed)
- Re-reads: Re-read 10 Nov after code changes: Fly Worker keeps A. Fly Worker's own numbers after a steering-bias fix (author's numbers, not re-run by us): the fly brain covers 50% of the game map, smoothed noise 63% and uniform noise 49%; a fly brain shown a different board scores 66%. The author withdraws the earlier -3 point paired result. There is still no comparison with scrambled wiring. Re-read 10 Nov after changes: flyshi-research keeps B. Re-read 10 Nov after code changes: NeuroCraft Fly moves from U to C: its source is now public, and its readouts select scripted body programs.
- Re-check of 24 projects: Re-checked 10 Nov: 22 of the 24 projects we graded on 20-21 Oct keep their grade (1 after a full re-read of changed files); 2 could not be re-checked because the source is no longer public. Two batch entries could not be re-checked: the source repositories of ascend and connectomics-fly-whole-brain-nulls returned 404 (checked 10 Nov 2026); each grade is for the code we read on its grade date (20 and 21 Oct 2026). flylab-demo: its source repository could not be reached on 9 and 10 Nov 2026 (not found); its grade (C) is for the code we read on 30 Oct 2026.
- Gamebook: 'What Is It Like to Be a Fly?' (Andrew Sheerin, free browser gamebook, shared from 6 Nov 2026) replays spike recordings from published fly-brain models (Shiu et al. 2024 on FlyWire data); the movements and two of its six scenes are written by the author, as its own note says. The model does not run in the page, and no code repository is linked (page read 10 Nov 2026). (verdict row)
- Scan: 10 Nov: we scanned news, videos and new code from 9 to 10 Nov, plus items from 6 Nov that our previous scan missed; one new claim got a dated check: a free browser gamebook, 'What Is It Like to Be a Fly?', that replays recorded fly-brain model activity. 7 new code projects are queued for a first look.
9 Nov: the walking-circuit model re-run with shuffled wiring
- Walking-circuit paper, our check: We re-ran the authors' nerve-cord walking model on an ordinary 2-processor server without a GPU (9 Nov 2026): with the real front-leg wiring, 3 motor neurons showed a regular rhythm; with the wiring shuffled by the authors' own shuffle, 4 of 5 tries still showed a rhythm (0, 12, 40, 3 and 1 motor neurons in the 5 tries). One run each, faster solver settings than the paper, no noise. The authors' shuffle keeps each neuron's excitatory or inhibitory sign and shuffles connections within each cell group; it differs from the shuffle used for their circuit-pattern count. The shuffled circuits were far more active than the real one (1,645 to 1,995 active cells against 97; top firing rates about 230 against 16 per second), so this is not a like-for-like comparison; a test with activity matched has not been run. (table and chart)
- Controls ledger: one new row, made by us and marked as our own check: our ledger row reads no difference by the rule we fixed before the run (at least half of the shuffles with a rhythm); the paper's own entry stays not tested. The ledger now has 105 studies, 95 with a wiring null: helps 36, no difference 27, worse 11, mixed 17, not yet scored 4.
- What it took: One simulation took 1.2 minutes and its process peaked at 1.4 GB of memory; the software and wiring data used about 1,130 MB of disk.
- "Uploaded" fly-brain videos: A new Urdu and Hindi video, 'Scientists Uploaded a Fly's Brain Into a Computer Are Humans Next?' (Takhti, 3 Nov 2026; about 202,000 views on 9 Nov), uses the word uploaded in its title, while its own description says this does not mean a fly's mind was copied and calls Eon Systems' demos company claims with limits. That matches the primary sources: a wiring map traced from one preserved fly brain, run as a simplified model. We read the title and description, not the video (checked 9 Nov 2026). (verdict row)
- Scan: 9 Nov: we scanned news, videos and new code from 3 to 9 Nov; one widely seen new video, an Urdu and Hindi explainer whose title says a fly's brain was uploaded, got a dated check (its own description says no mind was copied). Two new write-ups about fly-brain games and puzzles and several small new code projects that say they test whether the fly's wiring matters are queued for a first look.
- Tracking: Links as on 2 Nov; 57 of 57 videos available; 9 tracked repositories with new commits since 2 Nov were recorded and not yet re-read.
- Next: An activity-matched follow-up needs its own plan and is not yet scheduled; the next beginner re-test is due 16 Nov 2026.
- Earlier: the 2 Nov and 22 to 31 Oct summaries are on New and updated.
14 Oct: the looming turn in degrees, at each input strength
Our own test · pre-registered · simulated body driven through our own mapping · one model · a simulation, not an animal
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.
Every brain drive of the 8 Oct strength series that differs from the no-brain floor has now walked in the simulated body. Charts for both eyes, every run and the limits
14 Oct: 4 new entries, none with a wiring control
- Catalogue: 144 entries. New (project claims are the projects' own): fly-cx-navigation-servo B, NeuroPest B (linked by its source repository only), Fly Garden B and FastBrain (n/a, a simulator speed-up). None of the four compares the real wiring with a scrambled copy, so the controls ledger keeps 53 studies, 44 with a wiring null.
- Re-checked on 14 Oct 2026: repositories of 14 of the 21 verdict rows (no code changed since grading) and all 56 videos (all available). fly64's causality citation now points to lines 37-47.
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.
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.
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.
| Shadow on | Nothing 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.
13 Oct: a fifth plain answer: can a fly brain learn maths, to hack or to play a game?
Some projects say so. We read what changes with practice in 13 projects, at one fixed version each: a trained readout on a fixed network (3), trained settings across the network (4), a learning rule that changes some connections (3), nothing trained (1), and 2 videos with no code to check. The maths project's own README says "no fly learned anything". Each claim is attributed to its project in the full answer, and each of the 13 catalogue entries now carries a "What is learned" note.
13 Oct: 6 new entries, three new ledger rows, two moved repositories
- Catalogue: 140 entries. New: malecns-visual-motor A, malecns-keratitis-pilot A (a negative result reported by its author), fly-chess, adfly-connectome, fly-brain-sim and FlyGift (all B; FlyGift is linked by its source repository only).
- Controls ledger: 53 studies, 44 with a wiring null (helps 21, no difference 10, worse 4, mixed 8, not yet scored 1; the three new rows).
- Moved by their owner: the ChessFly and fly67 repositories moved from znatgost to gexhaw on GitHub (GitHub answers with a permanent redirect; the old demo pages answer 404). Their try and source links now point to the new addresses, dated 13 Oct 2026; grades unchanged until the next re-check.
12 Oct: our flyvis motion test, second attempt: still no result
Our own test · pre-registered · stopped for memory · no chart and no result
The question: does the real wiring layout help a trained fly vision model see which way things move? We shuffle the layout in flyvis, a trained model of the fly's motion-vision circuit, and check whether its motion-sensitive cells still prefer the right directions. After the first attempt on 10 Oct, we tried again with a lower-memory setup.
Not decided: a second, lower-memory attempt on 12 Oct also needed more memory than our shared machine has (about 3 GB at its peak; the machine allows 3 GB for everything running on it).
Measured: it ran out while preparing the moving-edge test input, before the first simulation step (sampled peak about 3.17 GB of the 3.22 GB limit, which is 3 GiB). So there is no result in either direction, no ledger row and no chart. We did not retrain the model after shuffling. A model trained on the shuffled wiring might do better, so this asks a narrower question: does this trained model depend on the real layout? The dated note · every step of 12 Oct
First attempt, 10 Oct 2026: "Not decided: the test did not run, because it needed more memory than our shared machine has. A pre-registered setup check also failed, so no result would have been read." (report of 10 Oct 2026)
12 Oct: 10 entries re-checked in code, no grade changed
Our re-check · each repository read again at its graded commit · two text fixes
Instead of the flyvis test, we re-read 10 graded entries line by line against the files their grades cite. No repository had new code since its grade, and every grade stayed the same. Two texts were corrected: Stonkfly's description of its learning rule is now more exact: a connection is mostly weakened when its input cell fires before the dopamine pulse, and the reverse timing can strengthen it, and the DOOMFLY announcement video card no longer says the read-out cells were "chosen because they responded to the screen", which is not in the cited files. All 10, with their grades
12 Oct: one new ledger row, a StepMania player
Authors' own numbers · read by our fixed ledger rules · not recomputed by us
- Helps duth-x-stepmania (grade A; new entry): duth-x-stepmania: the authors compared the real wiring with 3 degree-preserving rewired copies (mean) on StepMania hit rate on the 152-note Edit chart. Real 0.9737, null 0.4013 (3 null samples). Our ledger rules read this as helps. Quality: fair, moderate. Caveats: one seed; the readout groups are taken from the real wiring; 3 rewired copies.
The ledger now has 107 studies, 97 with a wiring null: helps 36, no difference 27, worse 11, mixed 18, not yet scored 5. What changed in the ledger
- 213 entries (2 new; project claims are the projects' own):
- A duth-x-stepmania (abstractionthief): a 766-neuron looming-escape circuit from the MaleCNS male fly connectome, run as a rate model, plays StepMania's arrow charts: each note becomes a looming threat, and a fixed readout of its descending neurons picks the arrow. The same chart is played by three rewired copies of the circuit.
- B flyonenomics (uguryildirim24): a spiking model of the male fly nervous system on MaleCNS v1.0 for simulated dopamine and transmitter changes. Its README reports spike activity only and says this is not treated as animal behaviour. No wiring control, so no ledger row.
- Viral claims: the lead scan of 12 Oct found nothing that met the bar for a verdict (0 of 2 promoted; 5 new code repositories noted as leads, not graded). All 26 verdicts re-checked on 12 Oct: no code changed. The "learns to hack" video (dzuma) had approximately 92,236 views on 29 Oct 2026 (video page) and still links no code, U. FlyLeno's playthrough video had approximately 55,574 views on 29 Oct 2026 (read from a search page). 57 videos, all available.
11 Oct: the controls ledger caught up with five new rows
Five studies joined the ledger, each with its pre-registered plain sentence and the authors' own numbers: fly-distill (mixed), a mushroom-body odour model (worse), brainlab (no difference), MelanoGraph (helps) and fly-arithmetic (worse). After 11 Oct the ledger had 49 studies, 40 with a wiring null. Also on 11 Oct: 4 new entries (MelanoGraph A, fly-arithmetic A, FlyRoyale B, connectome-control-exploration B) and a browser check of fly-arithmetic's website (ran, no GPU). Report of 11 Oct 2026
- 10 Oct: 3 new entries (fly-distill, flywire-connectome and brainlab, all A); their ledger rows joined on 11 Oct. One correction: Fly With Me loads trained synapses by default.
9 Oct: which relays carry the turn? Read by firing rate
Our own test · pre-registered before the first trial · brain only · one model · not a real fly
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.
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 this 9 Oct line removed 2-6 times as much activity as the relays (2.3 times left, 5.8 times right). Against an activity-matched control (13 Oct) the two sides disagree, and we do not draw a conclusion (details).
- about 70% / 65%of normal with PVLP141 silenced (left eye / right eye, 3 runs per eye)
- about 65% / 60%of normal with the 17 / 20 excitatory relays silenced (3 runs per eye)
- 1 Hzwide: the left eye's normal range (24-25 Hz); the random control is one draw per eye
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.
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.
On 8 Oct we read the relay test by the left/right contrast; this time we read how often the far-side turn neuron fired, against three runs per eye with nothing silenced, and added a control: the same number of randomly picked active neurons. That control moved the turn command too, so the relay result is not yet specific. Every run, the 8 Oct sentences with their label, and the limits
- Is it a brain or just a map? What the map leaves out: "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." The table: one number and one source per part
- 9 Oct: 125 entries (4 new: FlyVL A, where real wiring beat a rewired copy on motion but a classical motion detector beat both; Flappy Fly B; FlyWire Playground B; Let It Flyppen C) and 56 videos (a Spanish explainer by Idor). The controls ledger has 44 studies, 35 with a wiring null: helps 17, no difference 8, worse 2, mixed 7, not yet scored 1. Research report of the eleventh check
- Last week's threshold result (8 Oct, brain only): "The turn command appeared from about 30 per second (38% of our strongest input) on both sides; below that the model did not turn." Chart and caveats
Start from your question
- I saw a fly brain playing Doom, Minecraft or Pokémon. Is it real? Is it real?A verdict table for viral claims and a five-question check you can use on any demo.
- Does the fly's real wiring actually matter, or would any network do? Does fly wiring help?107 control studies side by side, our own four-way scrambling test, our walking-body test and our looming tests: where scrambled wiring fails, where it does just as well, and which controls are fair.
- I want to watch the demos everyone is sharing. Watch the demos56 videos from YouTube, X, TikTok and project sites, labelled by language, each with its project's evidence grade and what the video does not show.
- I want to try a fly brain in my browser right now. Ways to try the flyOnline demos and no-install options, with the date we checked them, which demos actually ran in our browser, and how much each one downloads.
- I want the GitHub code, the licence and what it needs to run. Projects catalogueFilter by evidence grade, dataset, licence, platform and trained parts ("is it AI?"). Every entry has its source.
- I want to build my own small fly brain project. Build your ownFrom an empty folder to a whole-brain simulation, a brain driving a walking fly body, and a looming shadow that turns it, tested on Linux with 2 CPUs; untested steps are marked; a 20-minute class plan.
- I want to understand the science behind it. How it worksThe FlyWire connectome and other datasets, what you have to download, neuron models, bodies and the sense → model → act loop.
- I am an AI agent, or I build one. For AI agentsStable routes, record fields, grade meanings and the citation rule.
The catalogue at a glance
- 213projects, datasets and tools checked on 12 Oct 2026
- 130where the simulated brain really drives the behaviour (grades A and B)
- 62you can open in a browser or free notebook (links checked; 13 browser demos loaded by us on 3, 6 and 11 Oct, 8 ran)
- 30code repositories with no licence file, so reuse is not clearly allowed
| Kind | A | B | C | D | U | n/a | Total |
|---|---|---|---|---|---|---|---|
| Browser demos | 3 | 9 | 7 | 1 | 3 | · | 23 |
| Games | 12 | 17 | 13 | 1 | 6 | · | 49 |
| Desktop apps | · | 3 | 2 | 1 | · | · | 6 |
| Art | · | 2 | 4 | 1 | · | · | 7 |
| Research and control studies | 62 | 19 | 10 | 1 | 4 | 1 | 97 |
| Brain models and engines | 3 | · | 1 | · | · | 3 | 7 |
| Fly bodies | · | · | · | · | · | 2 | 2 |
| Datasets | · | · | · | · | · | 7 | 7 |
| Data tools | · | · | · | · | · | 15 | 15 |
| All entries | 80 | 50 | 37 | 5 | 13 | 28 | 213 |
Games and browser demos are mostly grade B or C: the brain runs, but people chose the inputs and outputs. The grade A entries are mostly research tests with controls, including several honest negative results.
How to read our evidence grades
Every project gets one grade. The grade says how well the project's own claim about a fly-brain-driven behaviour is supported. It is not a score for quality or fun: a careful study that found a negative result can be grade A, and a great game can be grade D.
- A Tested against real-fly data or a control
- B Real connectome runs, inputs and outputs chosen by hand
- C Partial, stand-in or decorative network
- D Scripted or ordinary game logic
- U Unverified claim
- n/a Dataset or tool, not graded
Two dates on every entry
Last upstream activity
The date of the project's last commit or release on GitHub or its own site. It tells you whether the authors still work on it. Some other lists show the date a repository was created instead, which can be years out of date.
Last verified by us
The date we last opened the link and checked the facts in the entry. A recent date here means the link worked and the licence and setup notes were correct on that day.
What is new
- 14 Oct 2026: sixteenth check. The looming turn in degrees: the rest of the 8 Oct strength series walked in the simulated body (left eye from 30 spikes/s; right eye borderline at 30, resting on one run; through our own mapping, not a separate test of the wiring); body lines for the silenced relays under the 13 Oct relay headline; 4 new entries. Report
- 13 Oct 2026: fifteenth check. The relays against an activity-matched control: the two sides disagree, and we do not draw a conclusion (brain only); a fifth plain answer, can a fly brain learn?; 6 new entries and 3 new ledger rows. Report
- 12 Oct 2026: fourteenth check. Our flyvis motion test ran out of memory a second time, so it still has no result; 10 entries re-checked in code (no grade changed, two text fixes); one new ledger row (duth-x-stepmania, helps) and 2 new entries. Report
- 11 Oct 2026: thirteenth check. Five new rows in the controls ledger (three catalogue flags changed), 4 new entries, a browser check of fly-arithmetic (ran), and view counts refreshed. Report
- 10 Oct 2026: twelfth check. Our flyvis motion test did not run: it needed more memory than our shared machine has, and a pre-registered setup check also failed; 3 new entries (all A, each with its own real-versus-shuffled comparison) and a correction to Fly With Me. Report
- 9 Oct 2026: eleventh check. Which relays carry the turn? Read by firing rate, silencing the relays lowered the turn command, but random active neurons moved it too, so it is not yet specific (brain only); What the map leaves out, in numbers; 4 new entries, 1 new video. Report
- 8 Oct 2026: tenth check. How strong must the shadow be? The turn command appeared from about 30 spikes per second per looming neuron (brain only); relays silenced; the looming body test at n 6; brain and body in one environment; a Use in class plan; a verdict on a "learns to hack" video; 5 new entries. Report
- 6 Oct 2026: ninth check. Where in the wiring is the turn? Scrambled maps as busy as the real one still gave no turn command (brain only); "Is it AI?" as a catalogue field; download sizes for 8 releases; five more browser demos, all ran; 5 new entries, 1 new video. Report
- 5 Oct 2026: eighth check. We ran the map: a looming shadow turned our simulated fly away with the real wiring (4 of 4) and not with scrambled wiring (0 of 3), with every caveat beside it; plain answers to four common questions; "Add a sense: looming" tested on Build your own; 4 new entries, 1 new video, language labels in the gallery. Report
- 3 Oct 2026: seventh check. Does it actually run? Our browser check of six demos (3 ran); a project page for flyvis; a "Last checked" column on Is it real?; fly-cartpole changed sides in the controls ledger; 3 new entries. Report
- 2 Oct 2026: sixth check. Who does the walking? Our own test of a fly brain model driving a walking body; a verdict on the Terraria fly; "Add a body" tested on Build your own; 4 new entries, 1 new video.
- 1 Oct 2026: fifth check. Our own pre-registered reflex test scrambles the fly brain four ways; verdicts on the Gorilla Tag and Roblox fight videos; a chart of the Doom autopilot test; 6 new entries, 2 new videos; Build your own re-tested.
- 30 Sep 2026: fourth check. A new page, Does fly wiring help?, puts 31 control studies side by side; a verdict on FlyLeno; 12 new entries, 2 new videos, and a Controls filter in the catalogue.
- 29 Sep 2026: third check. Verdicts on two viral claims (Pokémon on a Raspberry Pi, Rainbow Six Siege), a wiring test where scrambled wiring fails, DesktopFly re-graded to C with a new project page, 9 new entries and 3 new videos.
- 28 Sep 2026: second check. 58 entries and 40 videos re-checked with no changes, 15 new projects, 4 new videos, a Doom control study on Is it real?, and project pages for DOOMFLY and the Shiu et al. brain model.
- 28 Sep 2026: Video gallery. 40 fly brain videos, from the viral Doom, Minecraft and Beat Saber clips to the lab films, each with the project's grade from our catalogue and what the video does not show. Videos open on their original platform.
- 28 Sep 2026: first release. 58 checked entries with evidence grades, true update dates, licences and link checks.
- Verdicts on 21 viral claims, including a "learns to hack" video, Terraria, Gorilla Tag, FlyLeno, Doom, Half-Life, Minecraft, Pokémon, Rainbow Six Siege, Beat Saber, Bad Apple, Mario 64 and Eon's "uploaded" fly, on Is it real?
- A tested beginner guide: run the fruit fly brain model on a CPU, change it, and check it against shuffled wiring, on Build your own.
- Seven connectome datasets compared, with data licences and an explanation of the different neuron counts, on How it works.
Every later run adds new projects and material updates to New and updated, and each research run is kept in Research.