Does fly wiring help? · controls ledger, 10 Nov 2026
Does real fly wiring help? What 107 control studies found when the connectome was scrambled
A fly brain simulation can look impressive for many reasons: the game helps, a script helps, training helps. The real check is to run the same thing with the fly's wiring scrambled. If the scrambled brain does just as well, the real wiring was not doing the work. We collected every study in our catalogue that runs such a control and put their numbers side by side, and we ran three tests of our own, all pre-registered: four kinds of scrambling on a reflex (1 Oct 2026), a brain driving a walking body (2 Oct 2026) and a looming shadow that turns it (5 Oct 2026, with brain-only follow-ups on 6, 8 and 9 Oct and two more body runs on 8 Oct). On 9 Nov 2026 we added a fourth, also with its rule fixed before the run: a published nerve-cord walking model re-run with shuffled wiring, which was not activity-matched.
Short answer
Sometimes: fly wiring beat scrambled wiring in 36 of 97 studies, mostly untrained reflex, sensory and steering circuits (including our looming test); in trained, reservoir, ML, game, robot-arm and our walking tests it tied or lost (38).
Of 107 control studies in our catalogue, 97 compare the real wiring with a scrambled or rewired copy (a "wiring null"). The real wiring helps in 36, makes no difference in 27, does worse in 11, gives mixed results in 18, and is not yet scored in 5. The clear wins are untrained reflex, sensory and steering circuits. Networks trained for machine-learning or game tasks usually do as well with scrambled wiring. No scoring rule changed on 9 Nov; the counts moved because our own check of a published nerve-cord walking model joined as a separate row, which reads no difference by the rule we fixed before the run: 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. What changed on 9 Nov. On 31 Oct three studies joined: Fly-NAF reads helps (7 of 8 game threats answered on the threat's side first with the real wiring against 0, 3 and 3 for three shuffles, on one replayed night, with constants tuned on the real wiring, so we rate it unfair); Fly Hearing Lab and DrosoCode read worse (rhythm tuning 0.70 Hz real against 2.66 Hz shuffled; code-token accuracy 12.6% real against 18.1% rewired, one run each). What changed on 31 Oct. On 30 Oct four studies joined: fly-connectome-forecast shows no difference (mean forecast skill 0.355 with the fly wiring, 0.353 rewired, 0.355 random); RiskFly reads mixed (MAE 0.00576 with the fly wiring against 0.00723 for degree-preserving rewired wiring, fly better in 5 of 5 seeds, author's p = 0.0625, called exploratory, while a dense network did best (0.00537)); fly-novelty-scada reads mixed (CARE score 0.30 against 0.43 for shuffled wiring, interval including zero, and 0.44 for random wiring); and the Pugliese et al. walking-circuit study is not yet scored, because its test counts circuit patterns in the wiring (21,544 real against about 307 in shuffles) rather than walking with shuffled wiring. What changed on 30 Oct. On 29 Oct three studies joined: Buzz helps (the escape neuron fired in 30 of 36 looming trials with intact wiring and in none with shuffled weights), but its gains were hand-tuned on the intact wiring and one shuffle seed was used, so we rate it unfair; Synaptera BANC helps (test R2 0.9773 against 0.9504 for degree-preserving shuffled wiring), but the difference is small, an unconstrained network fitted better (0.9984), and shuffled wiring scored higher at 90 deg/s; Oruk Labs shows no difference by the authors' write-up (16.84% against 16.88% mAP; we did not re-run it; no code). What changed on 29 Oct. On 24 Oct three studies joined, by the authors' committed numbers: somahouse helps against a degree-preserving rewired copy (cost 0.1589 against 0.1622; lower is better), but dense wiring (0.1538) and a conventional controller with no brain (0.1172) both cost less than the fly wiring; fly-trader shows no difference against a random-weight circuit (excess return over buy-and-hold +280.3 against +211.0 points; the interval spans zero); fly-chess is worse: the shuffled brain matched Stockfish's move more often (20.9% against 18.0% top-1). What changed on 24 Oct. On 23 Oct seven studies joined: drosophila-connectome-sim helps (MN9 67.45 Hz on the real weights against 0.0 Hz in all 10 weight-shuffled brains; no no-brain baseline reported, no chance level), fly-brain-resuscitation and owen-silva-fly-brain read worse (for fly-brain-resuscitation we read a higher critical fraction as a sturdier network; the author does not present it as a score), two flywire-connectome steps show no difference and two fly-v4 stages are mixed. What changed on 23 Oct. On 22 Oct two studies joined: Matcha Fly helps (241.7 giant-fibre spikes on the real wiring against at most 21.0 in 100 shuffled brains; no no-brain baseline is reported and a spike count has no chance level, and random sets of other visual neurons gave 2.0 to 18.7), and Maxwell's Demon shows no difference (good:bad ratio 0.645 real, 0.597 shuffled, 1.035 for random decisions; higher is better by the author's write-up, so random decisions scored above both wirings). What changed on 22 Oct. On 21 Oct eight studies joined: Drosophila Cadet (pinball) and the whole-brain nulls study help against scrambled wiring, but in the pinball study a no-brain hand-written reflex scores higher (11.480 vs the real wiring's 11.221; shuffles 10.770-10.985), and the whole-brain nulls study (0.879 vs null means 0.501-0.821) reports no no-brain baseline (0.5 is chance); FlyGateDemo (by the numbers reported in its README), Fruit Fly Plays Fruit Ninja, Fly Touhou, The Fly Hears and Boltzmann-fly show no difference; Fly School has no wiring null. What changed on 21 Oct. On 20 Oct nine studies joined: ASCEND, fruitfly-lab and flybrain-intransitive help against scrambled wiring, but in fruitfly-lab a no-brain oracle scores 11.5 kills against the fly's 10.2, and in flybrain-intransitive a greedy rule with no brain wins 55 games, as many as the fly; alpha-fly is worse (746 vs 958 Elo); FlyBrain Trader, Fly Tic-Tac-Toe and Flybrain (yyk2007) show no difference; therealfly is mixed (6 of 6 vs 5 of 6 seeds; the author's own criterion does not credit the wiring); ENSOphilia is not yet scored (lower RMSE is better; the author's interval includes no difference). What changed on 20 Oct. On 19 Oct three studies joined: MaleCNS Cube (no difference; a trainable network on a 256-node MaleCNS v1.0 subgraph estimates how far a Rubik's cube state is from solved; a degree-preserving rewired graph (test error 0.338) and a parameter-matched MLP (0.325) do as well as the real wiring (0.351), 3 seeds), QedSDL Endpoint (mixed; the FlyWire v783 wiring from projection neurons to Kenyon cells is the fixed input layer of a novelty detector for simulated Raman spectra (a materials-lab task); AUROC real 0.9505, degree-preserving shuffle 0.9505, random wiring 0.9391 (20 seeds)) and flybrain (worse; a MaleCNS v1.0 network runs flies in an evolving foraging economy; weight-permuted wiring ends with more energy than the real wiring on all 3 seeds (742.93 vs 360.27 mean energy), and zeroed weights starve). What changed on 19 Oct. On 18 Oct two studies joined, by the authors' own numbers: FlySnake (helps; a MaleCNS v1.0 network plays Snake through a trained readout; on 100 held-out games the real wiring scored 20.03 food per game, a rewired graph with its own trained readout 3.63, and a hand-written rule 21.52) and flywire-snn-stdp (mixed; a FlyWire v783 hearing circuit (1,239 neurons) trained with a local learning rule tells apart five courtship-song rhythms: accuracy 0.88, against 0.724 for a configuration-model graph and 0.428 for a random graph (5 seeds); without learning the real wiring is not ahead (0.566 vs 0.604), and a random input mapping with learning reaches 0.920). Eight rows now carry a caveat marker for reading only (what it means). What changed on 18 Oct; 17 Oct; 13 Oct; 11 Oct (five studies joined); on 10 and 12 Oct our flyvis test gave no result; 9 Oct.
The answer depends on how you scramble. In our own pre-registered test of the sugar reflex, degree-preserving and weight shuffles silence it (0 Hz against 85 Hz), keeping the sensory and motor boundary keeps about 46% of it, and a sign shuffle makes the whole brain fire. See the test
Our own tests (1 to 8 Oct 2026): a reflex scrambled four ways (the answer depends on the scramble), a walking body (the legs do the walking, not the wiring) and a looming shadow on one eye: with the real wiring our simulated fly turned away in 6 of 6 runs, with scrambled wiring in 0 of 3 body runs. On 5 Oct that row scored Helps with an unfair null, because the scrambled brains were much quieter. On 6 Oct we turned the scrambled brains up to the real activity, at brain level: "We turned the scrambled maps up until they were as busy as the real one. They still did not produce the turn: the wiring, not the amount of activity, makes it." By the same fixed rules the row now reads fair, with a strong method; the runs are still few and one model. On 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." The input is a rate we set, not a measured response to a real shadow. On 9 Oct, brain only, read by firing rate: "Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons." (the relay chart; brain-level results never enter the ledger's effect.) Brain only, at 30 spikes/s (17 and 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.
Matching the number of spikes does not match the pattern: 1,152 to 2,228 neurons fired in the matched scrambled brains, against 379 (left) and 432 (right) in the real brain. On 17 Oct, before strengthening: 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. 21 Oct, with 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.
See the looming test
Our walking test, 2 Oct: who does the walking? We let the same whole-brain model drive a simulated FlyGym fly. With its forward-walking neurons stimulated, 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 a brainless constant drive. The legs, rhythm and balance come from FlyGym's controller, and our hand-made mapping caps the drive, so the result says more about the stimulus and the body than about the wiring. See the test and its limits
And beating a scrambled brain is not the same as beating a simple controller. Where the wiring helps, a brain-free baseline often matches the fly anyway: in Doom the fly survives 53.3 s, shuffled wiring 6.8 s, and a brainless autopilot 51.9 s (the doom-fly-control by gabrycina). Across the 49 studies that test a no-brain baseline, the fly network beats it in only 7: our looming test (against random steering), two untrained game demos with one match or shuffle each, a hash function that loses to its own wiring nulls, and a foraging economy (against a coin-flip decision) whose weight-permuted wiring does better still.
Checked 10 Nov 2026 · two new studies · authors' numbers, not re-run by us · no rule changed
Changed on 10 Nov: two new studies, one not scored
- Mixed Whole-CNS specificity (steering readout): 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.
- Not scored FlyWire signal-spread study: 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. Small sampled subnetworks, not the whole brain.
- No wiring null Fly Worker: 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.
The ledger now has 107 studies, 97 with a wiring null: helps 36, no difference 27, worse 11, mixed 18, not yet scored 5. Fly wiring beats scrambled wiring in 36 of 97 studies; it tied or lost in 38. Our own 9 Nov walking-model row stays as it was. Method and sources: report of 10 Nov 2026.
Checked 9 Nov 2026 · one new row, made by us · rule fixed before the run · not activity-matched
Changed on 9 Nov: our own check of the walking-circuit model
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.
- No difference Our check: Pugliese walking model shuffles (our run): by the rule we fixed before the run (if at least half of the shuffles still show a rhythm, the row reads no difference), 4 of 5 shuffles did. We rate it partly fair (the shuffled circuits were far more active, and whether the shuffle keeps each neuron's number of partners was not checked) and moderate. Stat note, in plain words: one real network against 5 shuffled networks, one run each, no noise, and looser solver tolerances than the paper (1e-4 and 1e-7 against 2e-6 and 5e-9); the code's stimulus adjustment was off (its default); solver status and traces were not saved; no statistics.
This is a separate row marked as our own check. The Pugliese et al. walking-circuit row (the authors' circuit-pattern count) stays not scored, and no scoring rule changed. Fly wiring now beats scrambled wiring in 36 of 95 studies; it tied or lost in 38. Method, table and chart: report of 9 Nov 2026.
Checked 31 Oct 2026 · three new rows · the authors' numbers, not re-run by us · each row with its stat note
Changed on 31 Oct: three new rows
Three studies joined, scored only by our fixed rules: Fly-NAF, re-read on 31 Oct, now compares its real wiring with shuffled copies, and two new catalogue entries compare real and scrambled wiring. Fly wiring now beats scrambled wiring in 36 of 94 studies; it tied or lost in 37.
- Helps Fly-NAF (game night replay): On one replayed game night the real wiring answered 7 of 8 threats on the threat's side first; three shuffled brains answered 0, 3 and 3. The game constants were tuned on the real wiring, so we rate the comparison unfair (and weak: one night).
- Worse Fly Hearing Lab (song tuning): In the male hearing circuit, rhythm tuning was larger with shuffled wiring (2.66 vs 0.70 Hz; one shuffle seed, 5 simulation seeds); the author finds no robust effect. Partly fair, weak.
- Worse DrosoCode (code tokens): A rewired mushroom body of the same size learned to predict code tokens better than the real wiring (18.1% vs 12.6%, one run each). Partly fair, weak.
The Pugliese et al. walking-circuit row stays not scored. We checked whether the authors' walking model can be re-run with shuffled wiring on an ordinary 2-processor server without a GPU (31 Oct 2026): its software would need an estimated 620 MB once installed, plus 86 MB of wiring data, more than our 600 MB test space. The test was not run. 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. No scoring rule changed.
Checked 30 Oct 2026 · four new rows · the authors' numbers, not re-run by us · each row with its stat note
Changed on 30 Oct: four new rows
Four studies joined, scored only by our fixed rules, from four new catalogue entries. None of them adds a "helps": fly wiring still beats scrambled wiring in 35 of 91 studies.
- Not scored Pugliese et al. 2026 (Nature): The authors' code includes a test that shuffles the wiring while keeping the balance of excitatory and inhibitory connections; their saved notebook output shows the number of three-neuron circuits of the walking-rhythm type among front-leg nerve-cord neurons: 21,544 with the real wiring and 306.95 ± 29.10 (mean ± SD) across 100 shuffles. We did not re-run it (checked 30 Oct 2026). This is a count of circuit patterns in the wiring, not a walking test with shuffled wiring, so our ledger does not score it. A nerve-cord study, not a whole-brain or body simulation; peer-reviewed, MIT licence.
- No difference fly-connectome-forecast: The fly reservoir forecast Polish wind and solar power as well as random and rewired reservoirs, not better (mean skill over six tasks: 0.355 fly, 0.353 rewired, 0.355 random; 5 seeds).
- Mixed RiskFly: MAE 0.00576 with the fly wiring against 0.00723 for degree-preserving rewired wiring, fly better in 5 of 5 seeds, author's p = 0.0625, called exploratory, while a dense network did best (0.00537); by our fixed rules the rewired comparison reads as no difference and the row as mixed, because a random sparse reservoir did clearly worse. The target is a formula, not an investment strategy; no licence file.
- Mixed fly-novelty-scada: The mushroom-body detector found almost no wind-turbine faults, and the real wiring scored no better than shuffled wiring (CARE score 0.30 against 0.43; difference −0.13, 95% interval −0.29 to +0.03); random wiring scored 0.44, so by our fixed rules the wiring nulls disagree and the row reads mixed.
The Pugliese et al. row is in the ledger but not scored: its shuffle test counts circuit patterns in the wiring (a count of circuit patterns in the wiring, not a walking test with shuffled wiring), so it cannot say whether the real wiring walks better. RiskFly and fly-novelty-scada read "mixed" because their wiring nulls disagree; in both, the main rewired or shuffled comparison does not separate from the real wiring by our rules. No scoring rule changed.
Checked 29 Oct 2026 · three new rows · the authors' numbers, not re-run by us · each row with its stat note
Changed on 29 Oct: three new rows
Three studies joined, scored only by our fixed rules, from three new catalogue entries. Fly wiring now beats scrambled wiring in 35 of 87 studies.
- Helps Buzz (looming escape): from the author's committed trial files (counted by us; not re-run), the giant-fiber escape neuron fired in 30 of 36 looming trials with the intact FlyWire wiring and in 0 of 36 with shuffled weights, but the input gains were hand-tuned on the intact wiring and only one shuffle seed was used, so we rate the comparison unfair. Weak.
- Helps Synaptera BANC (turn command): in the author's pre-registered test (10 seeds; not re-run by us), the real nerve-cord wiring fitted a turn command better than degree-preserving shuffled wiring (test R2 0.9773 against 0.9504), but the difference is small, an unconstrained network fitted better (0.9984), and shuffled wiring scored higher at the held-out 90 deg/s. Partly fair, moderate. A separate project from the Synaptera flight test (grade B).
- No difference Oruk Labs (voice emotion): The authors report that a 499-neuron fly circuit with a trained readout scored 16.84% mAP on held-out voices, and the same circuit with scrambled wiring scored 16.88% (paired difference −0.04 points, 95% interval −0.16 to +0.07; 6 Sep 2026). We did not re-run it. Grade U: write-up only; partly fair, weak.
Both new "helps" rows have limits that weigh on the result: in Buzz the gains were tuned on the intact wiring and only one shuffle was run; in Synaptera BANC the gap is small and an unconstrained network does better. The Oruk row is the authors' write-up only; no code is published. No scoring rule changed.
Checked 24 Oct 2026 · three new rows · the authors' committed numbers, not re-run by us · each row with its stat note
Changed on 24 Oct: three new rows
Three studies joined, scored only by our fixed rules: two from new catalogue entries and fly-chess's newly committed shuffled-wiring control. Fly wiring now beats scrambled wiring in 33 of 84 studies.
- Helps SomaHouse (house control): from the author's committed held-out results (60 test scenarios, 3 training seeds; not re-run by us), the fly-derived circuit cost 0.1589 against 0.1622 for a degree-preserving rewired copy (lower is better), but dense wiring (0.1538) and a conventional controller with no brain (0.1172) both cost less than the fly wiring. Fair, strong; the no-brain baseline is not beaten.
- No difference FlyTrader (stock trading): from the author's committed matched comparison (3 seeds, 10 unseen tickers; not re-run by us), the excess return over buy-and-hold (the no-brain baseline, 0 by definition) was +280.3 points with the FlyWire circuit against +211.0 with a random-weight circuit of the same size (paired difference +69.3, 95% interval -186.6 to +325.3). Back-test only; partly fair (random weights, not a degree-preserving rewiring), strong.
- Worse fly-chess (readout probe): from the author's committed metrics (single runs, one shuffle seed; not re-run by us), the shuffled wiring matched Stockfish's best move more often than the real MaleCNS wiring (20.9% against 18.0% top-1; board-blind baseline 11.7%); cutting the central brain gave 17.9%. Partly fair, weak.
somahouse counts as "helps" only against its rewired copy: dense artificial wiring and a conventional controller with no brain both cost less than the fly wiring, and the difference rests on 3 training seeds. No scoring rule changed.
Checked 23 Oct 2026 · seven new rows · the authors' committed numbers, not re-run by us · each row with its stat note
Changed on 23 Oct: seven new rows
Seven studies joined, scored only by our fixed rules: three from new catalogue entries and four from projects already listed. Fly wiring now beats scrambled wiring in 32 of 81 studies.
- Helps drosophila-connectome-sim (sugar to MN9): from the author's committed control file (20 trials per arm; not re-run by us), sugar input drove the feeding motor neuron MN9 at 67.45 Hz on the real FlyWire weights and at 0.0 Hz in every trial of all 10 weight-shuffled brains; no no-brain baseline reported, no chance level. Stimulating 21 random central neurons instead gave 0.0 Hz. Fair, moderate.
- Worse fly-brain-resuscitation (brain structure): from the author's committed structure results (one rewired copy, not re-run by us), sensory-to-motor reachability in the real wiring collapsed after removing 3.5% of neurons in order of connectivity, against 26.6% in a degree-preserving rewired copy; we read a higher critical fraction as a sturdier network; the author does not present it as a score; no no-brain baseline applies to this structural measure and it has no chance level. The anoxia and reperfusion results behind the main claim have no wiring control. Partly fair, weak.
- Worse owen-silva-fly-brain (odour learning): from the author's pre-registered results (20 seeds; not re-run by us), the mushroom-body model on the real wiring reached 57.5% final accuracy over 110 odours, against 64.4% with shuffled and 66.8% with uniform wiring; no no-brain baseline is reported and chance is about 0.9%. Only the projection-neuron to Kenyon-cell layer of a rate model is tested. Fair, moderate.
- No difference flywire-connectome step 68 (decomposition): from the author's committed results (not re-run by us): real wiring 0.575 against 0.589 for random wiring that keeps each Kenyon cell's input strength (median of 10); chance 0.053.
- No difference flywire-connectome step 69 (wiring flatness): from the author's committed results (not re-run by us): real wiring 0.954 against 0.959 for random wiring (median of 10); no no-brain baseline reported.
- Mixed fly-connectome-study fly-v4 stage 7 (odour classes): from the author's committed results (not re-run by us): real 0.349, shuffled 0.332, random 0.329; raw responses without a Kenyon-cell layer 0.351; chance about 0.11.
- Mixed fly-connectome-study fly-v4 stage 3b (learning a teacher): from the author's committed results (not re-run by us): real 0.650, input-total control 0.648, random 0.628; no no-brain baseline reported.
In fly-v4 stage 7 the second arm is the author's "shuffled" wiring; we did not re-read how it was shuffled in this stage. The author's planted positive control checks the method, not fly wiring. connectomics-fly-whole-brain-nulls: The repository was no longer online on 22 and 23 Oct 2026; the grade and the numbers describe the code we read on 21 Oct. Its link status is now gone; grade A of 21 Oct and its ledger row stay. maxwells-demon-fly-brain: the test good:bad crossing ratio was 0.645 for the real wiring, 0.597 for the shuffled wiring and 1.035 for random decisions, the no-brain baseline, and higher is better by the author's write-up (the code computes the ratio; the README does not say), so random decisions scored above both wirings. No scoring rule changed.
Checked 22 Oct 2026 · two new rows · the authors' committed numbers, not re-run by us · each row with its stat note
Changed on 22 Oct: two new rows
Two studies joined, scored only by our fixed rules. Fly wiring now beats scrambled wiring in 31 of 74 studies at the time (later count, 23 Oct 2026: 32 of 81).
- Helps Matcha Fly (looming escape) (grade A; new entry): From the author's pre-registered looming-escape experiment (10 seeds; not re-run by us), driving the looming detectors gave 241.7 giant-fibre spikes on the real wiring, while the highest of 100 shuffled-wiring brains reached 21.0; no no-brain baseline is reported and a spike count has no chance level, and driving random sets of other visual neurons instead gave 2.0 to 18.7 spikes. Stat note: the shuffled value is the highest of 100 brains (the mean is not in the file); the author's two taste experiments failed their own first criterion, so their wiring control was not run. Fair, moderate.
- No difference Maxwell's Demon (door keeping) (grade A; new entry): In the author's benchmark (3 seeds, reported by the author; not re-run by us), the test good:bad crossing ratio was 0.645 for the real wiring, 0.597 for the shuffled wiring and 1.035 for random decisions, the no-brain baseline. The author's files do not state whether a higher ratio is better; we read higher as better, and on that reading random decisions scored above both wirings. Stat note: in a 2-seed run the ratios were 0.893, 0.391 and 1.091; the author reduced the sensory drive because real and shuffled wiring behave the same at full drive. Fair, moderate.
Re-checked on 22 Oct, rows unchanged in value: fly-connectome-study stays A at its current commit (the author's numbers, not re-run by us): in the author's planted positive control, real minus shuffled wiring was +0.064 [+0.058, +0.071] and real minus random wiring +0.066 [+0.059, +0.074], while the no-brain linear readout was level with the real wiring at −0.004 [−0.008, +0.001]. The ledger's rule on its unchanged arms still reads it as mixed. flyvis, flydoom-mutkuoz and fly-dino were re-read in full after code changes and their rows were rebuilt with the same values. No scoring rule changed.
Checked 21 Oct 2026 · eight new rows · the authors' committed numbers, not re-run by us (README numbers for FlyGateDemo) · each row with its stat note
Changed on 21 Oct: eight new rows
Eight studies joined, scored only by our fixed rules. Fly wiring now beat scrambled wiring in 30 of 72 studies at the time (later count, 22 Oct 2026: 31 of 74; later count, 23 Oct 2026: 32 of 81). Both new "helps" rows carry their baseline: in the pinball study a no-brain hand-written reflex scores higher than the real wiring, and the whole-brain nulls study reports no no-brain baseline.
- Helps Drosophila Cadet (pinball) (grade A; new entry): a 373-neuron MaleCNS v1.0 looming-escape circuit presses the flippers of 3D Pinball: in the author's committed log the real circuit scores 11.221 against 10.770-10.985 for six degree-preserving shuffles, while a no-brain hand-written reflex scores higher, 11.480 (score per life as log(1 + points), 96 paired lives). Stat note: not re-run by us; a separately trained 444-neuron brain shows no difference (real 11.186, shuffles 11.209-11.311); the author notes the comparison was framed after the fact. Fair, strong; the no-brain baseline is not beaten.
- Helps Whole-brain nulls (Shiu et al. model) (grade A; new entry): a reimplementation of the Shiu et al. whole-brain LIF model on FlyWire v783, scored on 157 of 164 published experimental tests: in the author's committed results the measured wiring scores 0.879 balanced accuracy against null means of 0.501-0.821 (four families, 100 draws each, 0 of 100 reaching it); no no-brain baseline is reported, and 0.5 is chance. Stat note: not re-run by us; two null families leave the network largely silent, and the margin over the cell-type block null is narrow and moves with analysis choices (0.039 to 0.064). Fair, moderate.
- No difference FlyGateDemo (grade A; new entry): a C# rate model of the MaleCNS v1.0 olfactory circuit blocks scanner probes on a blog after one punishment per probe family: by the numbers reported in the README, real wiring blocks 59% of later probes, shuffled wiring 54-59% and random wiring 64-68% (3 seeds each). Stat note: no result file is committed and we did not re-run it; a no-brain baseline, cosine similarity on the 53 receptor channels, catches 62-98% of probes per family. Fair, weak.
- No difference Fruit Fly Plays Fruit Ninja (grade A; new entry): a 4,386-neuron MaleCNS v1.0 circuit with a fitted readout: in the author's committed evaluation, mean reward 24.85 real, 24.80 for a target-permuted circuit with its own readout, and 24.95 for a no-brain pixel rule (20 episodes). Stat note: one rewired graph; not re-run by us. Partly fair, moderate; the no-brain baseline is not beaten.
- No difference Fly Touhou (grade A; new entry): a 600-neuron MaleCNS v1.0 rate circuit with an evolved readout dodges bullets: in the author's committed results it survives 1,386 frames, one degree-preserving rewired circuit trained the same way 1,398, and standing still 886 (20 test seeds). Stat note: the author's p = 0.9351; one rewired circuit; not re-run by us. Partly fair, moderate.
- No difference The Fly Hears (grade A; new entry): spoken digits through a 30,000-neuron MaleCNS v1.0 LIF reservoir with a trained reader: in the author's committed results 70.0% real, 69.7% with one scrambled graph and 66.9% for a reader on audio features alone (3,000 clips, 6 speakers). Stat note: with synthetic noise the scrambled graph is ahead (70.0% vs 68.2%); not re-run by us. Partly fair, weak.
- No difference Boltzmann-fly (grade A; new entry): a Deep Boltzmann Machine world model with couplings masked to the MaleCNS v1.0 mushroom body: in the author's results, visit AUC 0.712 with the fly mask, 0.712 with a degree-preserving rewired mask and 0.713 for a plain MLP. Stat note: one seed for data, mask and training; not re-run by us. Partly fair, weak.
- No wiring null Fly School (grade B; new entry): a MaleCNS v1.0 whole-brain LIF fly with a dopamine-gated mushroom-body rule: in the author's committed results a rewarded odour cuts the chosen output by 45-49% in 8 of 8 simulated flies, a never-paired odour by 1-4%. No scrambled wiring was compared, so the row has no wiring null and is not counted in the 72. Weak.
flybench (A) and NeuroPest (B) were re-graded after new commits and keep their grades; flybench's ledger row did not change, and NeuroPest still has no wiring comparison. Music in the Body C has a rewired listening mode but measures nothing, so it has no ledger row. No scoring rule changed.
Checked 20 Oct 2026 · nine new rows · authors' own numbers unless marked · each row with its stat note
Changed on 20 Oct: nine new rows
Nine studies joined, scored only by our fixed rules. Fly wiring then beat scrambled wiring in 28 of 65 studies. Later count (21 Oct 2026): fly wiring beat scrambled wiring in 30 of 72 studies at the time (later count, 22 Oct 2026: 31 of 74; later count, 23 Oct 2026: 32 of 81). Two of the new "helps" rows come with a simple rule that needs no brain and does as well as the fly, so beating scrambled wiring is not beating a simple controller.
- Helps ASCEND (grade A; new entry): a re-built FlyWire v783 whole-brain model: MN9 fires at 70.35 Hz under sugar with the real wiring and at 0 Hz in all 9 null networks (degree-kept rewiring, weight permutation and random graphs, 3 seeds each); a second readout gives the same pattern. Stat note: values read from the committed file, not re-run; the authors' 6-cell selectivity score (6 of 6 real, 4 of 6 every null) counts silent 'off' cells as passes. Fair, moderate.
- Helps fruitfly-lab (grade A; new entry): a 3,963-neuron FlyWire v783 steering circuit plays Doom (defend_the_center): 10.2 kills per episode with the real wiring against 0.7 with degree-preserving scrambled wiring, but a no-brain oracle that reads the enemy direction scores 11.5 (oracle 11.5 vs fly 10.2 kills). Stat note: the author's p = 0.0019 against scrambled wiring (test not named in the file); the oracle is not different from the fly (p = 0.35); 10 episodes. Fair, strong; the no-brain baseline is not beaten.
- Helps flybrain-intransitive (grade A; new entry): an untrained MaleCNS v1.0 network scores moves in rock-paper-scissors chess: 55 of 100 wins against a greedy bot with the real wiring, 0 with scrambled-target wiring, while a greedy rule with no brain also wins 55 (greedy rule 55 vs 55 wins). Stat note: the scrambled network gives no response to any move feature, so its moves are ties resolved at random; one scramble seed. Partly fair, weak; the no-brain baseline is not beaten.
- Worse alpha-fly (grade A; new entry): the full MaleCNS v1.0 wiring with trained synapse gains is trained at chess by self-play: 746 Elo with the real wiring against 958 with shuffled presynaptic partners at about 39,000 games. Stat note: rows read by us at matched game counts (difference -212 Elo, combined standard error about 35); the shuffled run also leads at 10,000, 20,000 and 30,000 games; the control stopped at about 12% of the main run; one shuffle. Partly fair, moderate.
- No difference FlyBrain Trader (grade A; new entry): a FlyWire v783 mushroom-body subgraph sizes positions in a simulated crypto portfolio: mean test-window return 10.15% real, 10.58% degree-preserving rewired, 11.48% random graph (6 walk-forward windows). Stat note: means and spreads computed by us from the author's table; no raw result files are committed; the author reads the gap as within noise. Fair, moderate.
- No difference Fly Tic-Tac-Toe (grade A; new entry): a 98-cell MaleCNS v1.0 circuit with a trained readout scores 0.854 against a random player, a degree-preserving rewired graph retrained the same way 0.850, and a plain board network without the circuit 0.905 (1,000 games). Stat note: against perfect play the real circuit loses 11 of 200 games and the rewired one 19; the number of rewiring draws is not stated; one checkpoint seed. Partly fair, weak; the no-graph baseline is not beaten.
- No difference Flybrain (yyk2007) (grade A; new entry): a MaleCNS v1.0 rate model with a trained flap decoder clears 15 gates per held-out course with the real wiring and 15 with presynaptic-ID shuffled wiring. Stat note: both arms hit the 15-gate cap on every course, so the task cannot separate them; the author states the result does not establish a wiring advantage. Partly fair, weak.
- Mixed therealfly (grade A; new entry): a MaleCNS v1.0 nerve-cord model driven through two descending neuron types: a 5-15 Hz leg rhythm in 6 of 6 seeds with the real wiring and 5 of 6 with a degree-preserving scrambled cord. Stat note: the author's own pre-set criterion needed the scrambled cord to fail on at least 5 of 6 seeds, so the author's criterion does not credit the rhythm to the wiring; the band peak is a harmonic of a fast runaway oscillation in both arms. Fair, moderate.
- Not yet scored ENSOphilia (grade A; new entry): an 80-neuron MaleCNS v1.0 circuit as a reservoir forecasts the El Nino index six months ahead: held-out RMSE 0.807 real against 0.742 for the median of 100 degree-preserving rewires, and 0.573 for a plain regression with no network. Stat note: lower RMSE is better; the author's paired 95% interval of real minus rewire median is -0.047 to +0.148, so it includes no difference; the real circuit is worse than 85 of 100 rewires. Fair, moderate; the no-brain baseline is not beaten.
flywire-connectome-scz was re-graded after new commits and stays A; its ledger numbers did not change. Flyshi Research B has a degree-preserving shuffle in its plan but no result yet, so it has no ledger row. No scoring rule changed.
Checked 19 Oct 2026 · three new rows · what each activity number counts · authors' own numbers unless marked
Changed on 19 Oct: three new rows, what each activity number counts, and a body walk
- No difference MaleCNS Cube (grade A; new entry): a trainable network on a 256-node MaleCNS v1.0 subgraph estimates how far a Rubik's cube state is from solved; a degree-preserving rewired graph (test error 0.338) and a parameter-matched MLP (0.325) do as well as the real wiring (0.351), 3 seeds. Fair, strong by our fixed rules; the MLP baseline is not beaten.
- Mixed QedSDL Endpoint (grade A; new entry): the FlyWire v783 wiring from projection neurons to Kenyon cells is the fixed input layer of a novelty detector for simulated Raman spectra (a materials-lab task); AUROC real 0.9505, degree-preserving shuffle 0.9505, random wiring 0.9391 (20 seeds); means recomputed by us from the committed per-seed files. Fair, strong.
- Worse flybrain (foraging economy) (grade A; new entry): a MaleCNS v1.0 network runs flies in an evolving foraging economy; weight-permuted wiring ends with more energy than the real wiring on all 3 seeds (742.93 vs 360.27 mean energy), and zeroed weights starve; the simulation kernel is not in the repository. Partly fair, moderate.
What each activity number counts. Where a study row shows brain activity against the real brain (display only, never scored), it now says what was counted: Activity counted over: the whole brain
, the neurons outside the stimulated input
, network units
or unclear
. Eight arms count the whole brain (our own tests) and two count network units (fly-connectome-reservoir). For the gain-matched arm of our looming test, counted over the whole brain (0.9864× the real brain), the count outside the stimulated input is shown beside it (0.98×), because that is what the arm was matched on.
Our own looming row is not re-scored. On 19 Oct we walked the one strengthened scrambled brain that sent a small steering signal in the simulated 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.
(below).
On 21 Oct we repeated the matched test with a second input draw: 21 Oct: we repeated the matched test with a second, independent draw of the same input on the same scrambled brains (one 1-second trial each).
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 (one run): 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.
(below).
Another public list, townie/awesome-fruit-fly (checked 19 Oct 2026), also notes null results; of the 10 projects in both lists with a wiring control, its notes state a direction for 8, and all 8 agree with our ledger.
Checked 18 Oct 2026 · two new rows · a caveat marker · authors' own numbers, not recomputed by us
Changed on 18 Oct: two new rows, a caveat marker, and the 30 spikes/s scrambled brains strengthened
- Helps FlySnake (grade A; new entry): a MaleCNS v1.0 network plays Snake through a trained readout; on 100 held-out games the real wiring scored 20.03 food per game, a rewired graph with its own trained readout 3.63, and a hand-written rule 21.52. The readout is trained and there is one rewired graph, so the no-brain baseline (the hand-written rule) is not beaten. Partly fair, moderate by our fixed rules.
- Mixed flywire-snn-stdp (grade A; new entry): a FlyWire v783 hearing circuit (1,239 neurons) trained with a local learning rule tells apart five courtship-song rhythms: accuracy 0.88, against 0.724 for a configuration-model graph and 0.428 for a random graph (5 seeds); without learning the real wiring is not ahead (0.566 vs 0.604), and a random input mapping with learning reaches 0.920. Fair, strong by our fixed rules.
The caveat marker. A "Caveat" marker on nine rows (eight on 18 Oct; Fly Touhou joined on 21 Oct, the author's own p = 0.9351) means that a control other than scrambled wiring came close to the real result (one without the fly wiring, or one that keeps the wiring but scrambles or replays the activity), or that the authors' own test did not separate the real wiring from a control. It is for reading only: it changes neither the row's wiring effect nor any count. The eight rows: Flycraft, fly-cartpole, fly-connectome-reservoir, Wired Different, flight-test-the-fly, flyarm, flyaim, Larva reservoir.
Our own looming row, byo-flygym-loom, is not re-scored. On 18 Oct we strengthened the 30 spikes/s scrambled brains, brain only; the sentences are in its note and below.
Checked 17 Oct 2026 · two new rows · authors' own numbers, not recomputed by us
Changed on 17 Oct: two new rows, and scrambled brains at 30 spikes/s
- Helps Flycraft (grade A; new entry): a frozen MaleCNS v1.0 nerve-cord graph drives an 18-joint simulated hexapod through a fixed hand-made motor map; in the author's test the real nerve-cord graph moved the hexapod 0.85 mm forward in 8 s, fully rewired graphs 3.5 mm backward, and a time-shuffled real signal 0.72 mm; the author's results file is marked pending the author's own review, and the paper's acceptance is the author's statement only. Every distance is under 1 mm, and the README says no alternating tripod gait forms. Fair, strong by our fixed rules; the grade describes what the code does, not the size of the effect.
- Helps fly-connectome-reservoir (grade B; new entry): a trained linear readout on a frozen MaleCNS-derived network; its separate reservoir test scores real wiring 100%, target-permuted wiring 83.9% and no connections 50% (3 seeds); a raw-stimulus readout also reaches 100%, so the no-brain baseline is not beaten. Partly fair (the permutation moves inputs and outputs to other neurons), strong. The fly-swatter game itself has no wiring control.
On 15 Oct five studies had joined: flyscore (helps), flybrain-prove-ai (mixed), malecns-training-gomoku (no difference), fly-connectome-study (mixed) and flylab-pharmacology (not scored: the metric has no better or worse direction). No scoring rule changed on 15 or 17 Oct.
Our own looming row, byo-flygym-loom, is not re-scored: its arms stay the 80 spikes/s test. The 17 Oct brain-only trials at 30 spikes/s are in its note (the sentences below).
Checked 13 Oct 2026 · three new rows · authors' own numbers, not recomputed by us
Changed on 13 Oct: three new rows, and the relays against an activity-matched control
- Helps malecns-visual-motor (grade A; new entry): the author compared the original MaleCNS network with 3 degree-preserving rewired networks on DNp15 direction contrast under identical input. Real +81.2 ± 2.3 Hz, rewired -3.4, +1.4 and -0.2 Hz. Quality: partly fair, strong (source data/common-input/analysis.md lines 7-14 at 9bd8336). Caveats: the optic-lobe front end is pretrained (FlyVis); single author, unpublished manuscript; not recomputed by us.
- No difference malecns-keratitis-pilot (grade A; new entry; a negative result reported by the author): real circuit AUROC 0.479 vs 0.464 for the mean of 10 rewired circuits; the author's paired 95% interval of the difference is -0.037 to +0.064. Quality: fair, strong (results/crop-summary.json lines 63-77 and 244-266 at bea3cb1). Both are near or below chance (0.5); the clinical data are private, so we could not recompute it.
- Helps fly-brain-sim (grade B; new entry): giant fibre 163-177 Hz with the real loom circuit, 47-73 Hz with shuffled synapse strengths and 0 Hz with scrambled targets (3 seeds). Quality: partly fair, weak: the numbers are only in a lab notebook (NOTEBOOK.md line 19 at 12f6c77), with no per-seed values; we scored the midpoints of the ranges.
The ledger now has 107 studies, 97 with a wiring null: helps 36, no difference 27, worse 11, mixed 18, not yet scored 5. Three other new entries of 13 Oct have no row: fly-chess has a shuffle option in its code but no committed results; adfly-connectome's "shuffled" arm moves where the simulated damage goes, not the wiring; FlyGift has only a zeroed-weights check.
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
A stricter control, counted (17 Oct 2026): We also looked for a stricter control: other active inputs that excite the turn neuron with a similar firing rate and a similar number of connections onto it. In this model there are none for any of the active relay neurons on either side (count of 15 Oct: the busiest relay, PVLP141, fires about 100 spikes/s; no other such input fires above 43), so that stricter test cannot be built here.
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.
Changed on 12 Oct: one new row, and our flyvis test again gave no result
- 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 (authors' numbers, not recomputed by us; source results/play_test.json at c6b829d). Caveats: one seed and no statistics; the readout groups (which cells count as up or down presses) are taken from the real wiring and reused on the 3 rewired copies; there is no no-brain baseline.
- No result Our flyvis motion test (made by us), second attempt: "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)." No result, so no ledger row and no chart.
On 12 Oct the ledger had 50 studies, 41 with a wiring null: helps 19, no difference 9, worse 4, mixed 8, not yet scored 1. A second new catalogue entry of 12 Oct, flyonenomics, reports spiking activity only and has no wiring control, so it has no row.
Changed on 11 Oct: five new rows, three of them catalogue flag changes
Each row was scored only by our fixed ledger rules, and each sentence was written down before the numbers were entered. Quality labels are rule-based; "weak" here means one rewired graph or one seed.
- Mixed fly-distill (grade A; catalogue flag was "no difference" until 10 Oct): fly-distill: the authors compared the real wiring with a degree-preserving shuffle on bits per byte (lower is better). Real 3.9572, null 3.9171 (4 null samples). Our ledger rules read this as mixed. Quality: fair, strong (authors' numbers, not recomputed by us; source STATUS.md lines 9-18 at ad5aef6).
- Worse flywire-connectome (mushroom-body model) (grade A; catalogue flag was "mixed" until 10 Oct; only the task metric is scored): flywire-connectome (the author's mushroom-body rate model): the authors compared the real wiring with strength-preserving null wirings (median of 100) on odour discrimination d'. Real 18.32, null 19.45 (100 null samples). Our ledger rules read this as worse. Quality: partly fair, strong (authors' numbers, not recomputed by us; source HALLEM_COMMUNITY_RESULTS.md lines 9 and 52-57 and PREREGISTRATION_16_18.md line 55 at 8aea8e8).
- No difference brainlab (grade A; catalogue flag was "not tested" until 10 Oct; a task metric was found in its committed results): brainlab: the authors compared the real wiring with a target-permuted copy on held-out decision accuracy (%). Real 92.91, null 92.20 (1 null sample). Our ledger rules read this as no-difference. Quality: partly fair, weak (authors' numbers, not recomputed by us; source reports/2026-10-05-scorer-polis-s2-screen.md lines 39-40 at 1fe1aa4).
- Helps MelanoGraph (grade A; new entry): MelanoGraph: the authors compared the real wiring with a degree-preserving rewired copy on the receiver's gain over a resampled message. Real 0.4941, null 0 (1 null sample). Our ledger rules read this as helps. Quality: partly fair, weak (authors' numbers, not recomputed by us; source docs/NEURAL_C2B_REPORT.md lines 35-42 at 45ad05d).
- Worse fly-arithmetic (grade A; new entry): fly-arithmetic: the authors compared the real wiring with a random network of the same size on held-out addition accuracy (%). Real 48, null 98 (10 null samples). Our ledger rules read this as worse. Quality: partly fair, strong (authors' numbers, not recomputed by us; source docs/RESULTS.md lines 4 and 11-15 at e36dd2b).
After 11 Oct the ledger had 49 studies, 40 with a wiring null: helps 18, no difference 9, worse 4, mixed 8, not yet scored 1. For the mushroom-body model, only the task metric (odour discrimination) is scored, not its structure statistic. brainlab's row uses the task metric found in its committed results; its spiking-activity comparison is not scored.
Changed on 10 Oct: no ledger change, and our flyvis test did not run
- Not run Our flyvis motion test (made by us): "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." No result, so no ledger row.
- Three new catalogue entries run their own real-versus-shuffled comparisons (fly-distill, a mushroom-body model on BANC and brainlab). Their rows joined this ledger on 11 Oct (above).
Changed on 9 Oct: a motion encoder where a classical detector wins
- Helps FlyVL (grade A): a frozen male-CNS connectome as a motion encoder, with a trained input patch and readout. In the author's committed result files, real wiring scores 90.7% against 59.8% for a degree-preserving rewired copy (+0.309, 95% CI 0.287-0.332, 3 seeds). But a classical Reichardt motion detector scores 99.8%, so the no-brain baseline is not beaten, and the author's stricter controls (which also keep input-layer locality and each neuron's excitatory/inhibitory make-up) remove the gap. Fair, strong; not recomputed by us.
- Helps Our looming test (made by us): unchanged. Its note now carries the 9 Oct brain-only relay test read by firing rate, which never enters the effect: "Silencing other active neurons also changed the turn command, so this relay result is not specific to the relay neurons."
Changed on 8 Oct: a looming escape that a hand-written controller ties, and an unscored landing test
- Helps Connectocopter (grade A): the Shiu et al. FlyWire model flies a simulated quadcopter-rover and escapes 15 of 15 looming balls with the real wiring against 0 of 15 with one degree- and sign-preserving rewiring (raw files recounted by us). But a hand-written controller also escapes 15 of 15, so the no-brain baseline is not beaten. One rewiring seed, thresholds tuned on the real wiring: unfair, weak.
- Not scored Synaptera (grade B): the model's landing neuron DNp07 responds to expanding flow (9.6 Hz more than to a static grating), but no more than in the author's shuffled connectome (up to 15.6 Hz). We did not inspect how the shuffle was made, so by the fixed rules its null type is "other" and the row counts as baselines only; the author's own reading is that the real wiring does not beat the shuffle. From the author's report: unclear, weak.
- Helps Our looming test (made by us): 6 of 6 real-brain body runs turned away (two more on 8 Oct), and a new inside-only arm: "With only the inside of the map scrambled, the fly turned away in 0 of 6 runs." Still fair, strong. Details below.
Changed on 6 Oct: two game wins on one shuffle, a mixed racer, a correction
- Helps Our looming test (made by us): now fair, strong, after the turned-up scrambles of 6 Oct; the effect is unchanged. Details below.
- Helps Fly tennis (grade A): two untrained MaleCNS flies rally 54 hits in one 40 s match against 10 with one shuffled wiring and 14 for a parked fly, from committed logs. One match per condition, and the logged closed-loop steering check is near zero: partly fair, weak.
- Helps FlyBrain Flappy (grade B): Flappy Bird through the looming-escape pathway scores 100.8 against 0 with shuffled or cut wiring, author-reported in the README. Unfair: the two input gains were tuned on the real wiring only. Weak.
- Mixed Fly-Racer (grade C): a PPO-trained MaleCNS steering core scores 903.2 against 883.7 for a degree shuffle (beyond 2 standard errors) but 896.9 for a sign shuffle (a tie), and a plain MLP reaches 894.0. README numbers, one seed each: partly fair, weak.
- Corrected: Fly Dino at commit e34c6614: a 243-parameter network with no connectome survives 180.00 s on 100 of 100 courses against 179.37 s for the connectome, so "no-brain baseline beaten" changes from yes to no. Still no wiring null.
Changed on 5 Oct: four new rows
- Helps Our looming test (made by us): a looming shadow on one eye turned the FlyGym fly away in 4 of 4 real-brain runs and 0 of 3 scrambled runs (6 of 6 real runs since 8 Oct). Unfair null and weak method by the fixed rules on 5 Oct; fair and strong since 6 Oct.
- Helps Flight-test the fly (grade A): a 20,556-neuron FlyWire motion-to-steering circuit damps an aircraft's yaw better than 10 degree-preserving shuffles (3.19 vs 3.92 deg/s), but a classical yaw damper does far better (1.69): the no-brain baseline is not beaten. Fair null, strong method.
- No difference FlyAim (grade A): the whole MaleCNS aims a crosshair no better than its shuffle (412.5 vs 430.9 px from the target), and a random walker does better (251.7 px). Pre-registered by its author.
- No difference FlyArm (grade B): a frozen MaleCNS between a trained encoder and decoder lifts objects with a robot arm in 72.2% of episodes against 55.9% for 9 shuffles, not significant over 6 seeds (p = 0.125). The author wrote "mixed" across several tasks; on the ledger's one metric the fixed rules say no difference.
Changed on 3 Oct: a flip and a clean negative
fly-cartpole: no difference → helps, but the baseline is not beaten
The author redesigned fly-cartpole around a MaleCNS flight-stabilisation circuit (5,459 linear units; only the sensor gains are tuned). With degree-preserving shuffled wiring the cart-pole score drops from 499.9 to 165.9 steps (20 seeds, one shuffle each, the author's p = 0.0001), so by our fixed rules the real wiring now "helps".
But a much simpler controller does as well. The circuit replaced by its own linear map, tuned the same way, scores 499.6, and LQR, a textbook controller designed from the cart-pole equations, scores 500. The fly circuit does not beat its no-brain baseline, and its real score sits at the 500-step ceiling. Grade A unchanged; the earlier mushroom-body result (392.5 vs 390.3, p = 0.48) is still in the repository. We did not recompute these numbers.
ChessFly: a clean negative result
ChessFly (new, grade A) feeds a chess board into the real FlyWire mushroom-body input layer (285 projection neurons, 5,177 Kenyon cells) and lets it learn by self-play. After 10,000 games each, the real wiring rates 1680 Elo, shuffled wiring 1675, and the author's original random design 1850.
The real wiring is no better than shuffled wiring and a little worse than a random design. The design is fair in spirit (same learning rule and schedule for every wiring), but there is one fly and one shuffle per arm and the exams have 40 games, so differences of tens of Elo are within noise: method quality weak.
New 1 Oct 2026 · made by us · pre-registered
We scrambled the fly brain four ways
Almost every number on this page is someone else's. This one is ours. We took the sugar reflex from our Build your own guide: 21 sugar-taste neurons firing at 150 Hz drive MN9, the motor neuron that moves the proboscis for feeding, in the Shiu et al. whole-brain model on FlyWire v783. We ran it on the real wiring and on four scrambled versions, five shuffles each. The design, the arms and the rules for reading them were written down at 10:51 UTC on 1 Oct 2026, before the first scrambled trial started (10:58 UTC).
- 85.2 Hzreal wiring, ± 3.56 over 5 trials; our re-runs gave the same spikes, bit for bit
- 0 Hzdegree-preserving and weight shuffles, 5 shuffles each: silent
- 39 Hzinterior-only (boundary-preserving) shuffle: keeps 46% of the response
- 214 Hzsign shuffle: above real, but only because the whole brain ignites
Numbers behind the chart
| Arm | What it keeps | MN9 (Hz), mean ± sd | Per shuffle (Hz) | Response kept | Activity vs real | Neurons active | Reading |
|---|---|---|---|---|---|---|---|
| Real wiring | The measured FlyWire v783 connectome | 85.2 ± 3.56 (5 trials) | 82, 89, 81, 86, 88 | 1 | 1× (13,372–14,284 spikes) | 366–385 | reference |
| Degree-preserving | Keeps partner counts | 0 ± 0 | 0, 0, 0, 0, 0 | 0 | 0.291× | 103 | silent |
| Weight shuffle | Keeps who connects to whom | 0 ± 0 | 0, 0, 0, 0, 0 | 0 | 0.388× | 109 | silent |
| Boundary-preserving | Keeps sensory and motor wiring | 39 ± 1.22 | 38, 39, 39, 41, 38 | 0.458 | 0.648× | 215 | partly |
| Sign shuffle | Swaps excite and inhibit | 214.2 ± 43.27 | 268, 188, 191, 253, 171 | 2.514 | 257.637× | 60,501 | runaway |
| Gain check on degree-preserving shuffles 3–5 (1 and 2 Oct) | As degree-preserving, with every synapse ×1.5 to ×3 | 0 at every gain | shuffle 3 ×1.5: 0, shuffle 3 ×2: 0, shuffle 3 ×2.5: 0, shuffle 3 ×3: 0, shuffle 4 ×2.5: 0, shuffle 5 ×2.5: 0 | 0 | 0.37×, 0.50×, 0.73×, 1.74×, 1.76×, 2.12× | 193, 349, 709, 4,257, 2,776, 4,104 | not matched on 3 shuffles: ×2.5 put shuffle 3 within ±30% of the real activity (10,138 spikes), shuffles 4 and 5 above it |
What each scrambled brain keeps, and what happened
-
Degree-preserving silent activity reduced 0.29×
Keeps every neuron's number of inputs and outputs, and its sign. Changes who its partners are.
MN9 0 ± 0 Hz in all 5 shuffles. The brain goes quiet: about 29% of the real spikes, about 100 active neurons and almost no motor or descending neurons, although the sugar neurons still fire at about 150 Hz.
-
Weight shuffle silent activity reduced 0.39×
Keeps who connects to whom. Changes which synapse strength sits on which connection.
MN9 0 ± 0 Hz in all 5 shuffles, with about 39% of the real spikes. By our pre-registered reading, both silent arms mean the reflex needs the specific wiring these nulls destroy.
-
Boundary-preserving partly activity comparable 0.65×
Keeps every output of the sensory and ascending neurons and every input to motor, descending and endocrine neurons. Changes only the wiring in between.
MN9 39 ± 1.22 Hz (retention 0.46) with about two thirds of the real spikes. Part of the reflex runs on boundary pathways that this null keeps; the central wiring matters for the rest.
-
Sign shuffle runaway activity runaway 258×
Keeps the whole graph and every synapse size; each neuron still has one sign (Dale's law). Changes which neurons excite and which inhibit.
MN9 214 ± 43 Hz, above the real rate, but only because the whole brain ignites: about 3.6 million spikes (258 times the real run), about 60,500 active neurons, and about 1,033 of the 1,299 descending neurons. The real assignment of excitation and inhibition is what keeps the response selective.
Why our own study row scores lower
By the fixed ledger rules, our study row is now Mixed, with an unfair null and weak method; on 30 Sep, with 2 shuffles, it was "helps", partly fair and moderate. Without a matched gain the rules cannot mark the scrambled brains as "tuned equally", so a null that loses counts as unfair (lower excitability could explain the loss), and "unfair" caps the method at weak. And on the ledger's single metric, the MN9 rate where higher is better, the sign-shuffle runaway counts as "the null does better", so the wiring effect comes out mixed.
We did not override the rules for our own study, and no scoring rule changed on 2 Oct: the gain check above ran as measured and did not match, so the row stays mixed, unfair and weak. What we added instead is display-only: each arm now shows how active the whole brain was against the real one, so readers can see that the sign shuffle is a runaway. These labels never enter any score: degree-preserving reduced 0.29×, weight shuffle reduced 0.39×, boundary-preserving comparable 0.65×, sign shuffle runaway 258×. Our row in the ledger
What it shows
- For this reflex, the specific wiring matters: two standard scrambles silence it, and a more excitable scrambled brain does not bring it back.
- Keeping the sensory and motor boundary keeps about half of the response, so the reflex runs partly through boundary pathways and partly through the central wiring.
- Flipping excitation and inhibition does not make a better reflex. It makes a brain that fires everywhere.
Limits
- One reflex (sugar to MN9) in one model (Shiu et al. leaky integrate-and-fire, upstream parameters), 1 s trials.
- One trial per shuffle at one trial seed; 5 shuffles per arm.
- MN9 rate is the only metric, and it cannot tell selective drive from global ignition. That is why the chart shows the brain's activity next to it.
- The gain check found no gain that matches the real activity on 3 shuffles: ×2.5 matched shuffle 3 only (2 Oct 2026).
- The boundary split comes from FlyWire's cell-class annotation (138,625 of 138,639 neurons matched). The 581 neurons of the combined "sensory_ascending" class were counted as input side, a class the pre-registration had not named. Two of the five degree-preserving shuffles are those of our 28 Sep run, re-used as planned.
The model is philshiu/Drosophila_brain_model at commit 91bdd1e7, with its default settings. Want to try a smaller version? See Try a fair control yourself. Full method and deviations: research report of 1 Oct 2026; the gain check of 2 Oct: report of 2 Oct 2026.
New 2 Oct 2026 · made by us · pre-registered
Who does the walking? We gave the brain a body
Viral videos say a fly brain "learned to walk". So we connected the same whole-brain model to the walking fly of FlyGym 2.1.0 (NeuroMechFly v2), the most-used simulated fly body, and compared a real brain, scrambled brains and no brain at all. FlyGym's walking controller takes two numbers, one per side, that set how strongly each side's legs step. We stimulated the brain's forward-walking command neurons P9 (DNp09) at 150 Hz, recorded what its descending neurons sent down every 100 ms, and turned those rates into the two numbers with a mapping we wrote by hand. The design was written down at 09:35 UTC on 2 Oct, before the first trial.
- 13.9 mmreal brain, ± 0.58 over 3 seeds, in 1 s of walking
- 14.1 mmscrambled brain, 3 shuffles with identical drives: in effect one walk
- 14.3 mmno brain: a constant drive matched to the brain's average
- 0 mmreal brain tasting sugar: MN9 82 Hz, every walking command 0 Hz
What makes the fly walk. (1) Legs, rhythm and balance come from FlyGym's walking controller. (2) The brain sends two numbers, left and right, every 100 ms. (3) How neuron rates become those two numbers is our hand-made mapping. Paths are drawn to scale from the thorax position every 10 ms; forward is to the right. The three scrambled-brain paths coincide because their drives were identical. Our test of 2 Oct 2026, pre-registered before the first trial.
Open the chart as an image (SVG) to share it; it carries its own title, numbers and source line.
Read this before sharing. The fly walks because we stimulate its forward-walking neurons (P9); P9 fires at the same rates whether the rest of the brain is real or scrambled. Our mapping caps each drive number at 1.2, and 90 of the 120 drive values of the six P9 walks (75%) sit exactly at the 1.2 clip, so forward speed could hardly differ between conditions by construction. Another hand-made mapping could let the wiring matter more. One stimulus, one model, one body, 1 s of walking, open loop (the body never feeds back to the brain).
The brain is not doing nothing. Only the real wiring recruited the turning neurons DNa02, more on the right than the left in all three seeds; the scrambled brains recruited no mapped neuron. That turned the real-brain fly by −0.4° to −26°, but random no-brain drives turned it by −7° to −27° in the same second, so 1 s of walking cannot tell the two apart. Sugar: the same brain tasting sugar fires its feeding neuron MN9 at 82 Hz while every walking command neuron stays at 0 Hz, so the fly stands.
Numbers behind the chart (all 18 runs)
| Run | Group | Forward (mm) | Heading (°) | Path (mm) | Drive L / R | Brain sends (Hz) | Files |
|---|---|---|---|---|---|---|---|
| Real brain, P9, trial seed 0 | Real brain | 13.26 | -26.3 | 18.89 | 1.20 / 1.00 | P9 130/162, DNa02 7/29, MDN 0/0 | body/P9-real-0.json drives/P9-real-0.json brain/P9-real-0.json |
| Real brain, P9, trial seed 1 | Real brain | 14.35 | -0.4 | 19.64 | 1.16 / 1.11 | P9 146/132, DNa02 11/21, MDN 0/0 | body/P9-real-1.json drives/P9-real-1.json brain/P9-real-1.json |
| Real brain, P9, trial seed 2 | Real brain | 14.17 | -14.4 | 19.50 | 1.19 / 1.06 | P9 141/136, DNa02 10/25, MDN 0/0 | body/P9-real-2.json drives/P9-real-2.json brain/P9-real-2.json |
| Scrambled brain, P9, shuffle 3 | Scrambled brain | 14.15 | -14.6 | 19.52 | 1.20 / 1.06 | P9 130/162, DNa02 0/0, MDN 0/0 | body/P9-D3.json drives/P9-D3.json brain/P9-D-3.json |
| Scrambled brain, P9, shuffle 4 | Scrambled brain | 14.15 | -14.6 | 19.52 | 1.20 / 1.06 | P9 130/162, DNa02 0/0, MDN 0/0 | body/P9-D4.json drives/P9-D4.json brain/P9-D-4.json |
| Scrambled brain, P9, shuffle 5 | Scrambled brain | 14.15 | -14.6 | 19.52 | 1.20 / 1.06 | P9 130/162, DNa02 0/0, MDN 0/0 | body/P9-D5.json drives/P9-D5.json brain/P9-D-5.json |
| No brain, constant drive (L 1.18, R 1.06) | No brain | 14.28 | -11.3 | 19.49 | 1.18 / 1.06 | no brain | body/matched.json drives/matched.json |
| No brain, random drive, seed 0 | No brain | 14.01 | -9.7 | 19.07 | 1.16 / 1.05 | no brain | body/random-0.json drives/random-0.json |
| No brain, random drive, seed 1 | No brain | 14.39 | -7.5 | 19.65 | 1.18 / 1.08 | no brain | body/random-1.json drives/random-1.json |
| No brain, random drive, seed 2 | No brain | 13.48 | -27.2 | 18.81 | 1.19 / 0.97 | no brain | body/random-2.json drives/random-2.json |
| Real brain tasting sugar | Sugar | 0.00 | -0.0 | 0.12 | 0.00 / 0.00 | P9 0/0, DNa02 0/0, MDN 0/0 | body/SUG-real-0.json drives/SUG-real-0.json brain/SUG-real-0.json |
| Zero drive | Floor | 0.00 | -0.0 | 0.12 | 0.00 / 0.00 | no brain | body/zero.json drives/zero.json |
| FlyGym default drive (1, 1), sanity check | No brain (sanity) | 13.01 | 6.7 | 17.48 | 1.00 / 1.00 | no brain | body/const-1.json drives/const-1.json |
| Scrambled brain tasting sugar, shuffle 3 | Sugar (fill) | 0.00 | -0.0 | 0.12 | 0.00 / 0.00 | P9 0/0, DNa02 0/0, MDN 0/0 | body/SUG-D3.json drives/SUG-D3.json brain/SUG-D-3.json |
| Scrambled brain tasting sugar, shuffle 4 | Sugar (fill) | 0.00 | -0.0 | 0.12 | 0.00 / 0.00 | P9 0/0, DNa02 0/0, MDN 0/0 | body/SUG-D4.json drives/SUG-D4.json brain/SUG-D-4.json |
| Scrambled brain tasting sugar, shuffle 5 | Sugar (fill) | 0.00 | -0.0 | 0.12 | 0.00 / 0.00 | P9 0/0, DNa02 0/0, MDN 0/0 | body/SUG-D5.json drives/SUG-D5.json brain/SUG-D-5.json |
| Real brain, left P9 only | Real brain (fill) | -4.77 | 254.2 | 12.75 | -0.07 / 1.11 | P9 131/0, DNa02 14/0, MDN 0/0 | body/P9L-real-0.json drives/P9L-real-0.json brain/P9L-real-0.json |
| Scrambled brain, left P9 only, shuffle 3 | Scrambled brain (fill) | -2.18 | 210.3 | 10.16 | 0.00 / 1.11 | P9 131/0, DNa02 0/0, MDN 0/0 | body/P9L-D3.json drives/P9L-D3.json brain/P9L-D-3.json |
Retention, (null − zero) / (real − zero) on the mean forward distance: scrambled 1.016, brainless constant 1.025, brainless random 1.002. Mapping (ours, hand-made, not fitted): drive_s = clip(pbar − 0.5·(p_s − p_o) − m_s − 0.5·(a_s − a_o), −1.2, 1.2) with p, m, a = DNp09, MDN and DNa02 rates ÷ 100 Hz.
Replay it, 100 ms at a time
Pick runs and step through the second: what the brain sends down, the two numbers our mapping makes of it, and where the fly goes.
What the brain sends down
Tasting sugar sends no walking command at all. P9 fires because we stimulate it, at the same rates in real and scrambled brains. What only the real wiring adds is turning: it recruits DNa02, more on the right than the left in every seed, and the scrambled brains recruit no mapped neuron. No trial sent a backward command (MDN).
| Trial | MN9 (feeding) | Whole-brain spikes | Active descending neurons (of 1,299) | P9 (forward) | DNa02 (turning) | MDN (backward) |
|---|---|---|---|---|---|---|
| Sugar, real wiring (our gate trial) | 82 Hz | 13,372 | 65 | 0 / 0 | 0 / 0 | 0 / 0 |
| Sugar, scrambled (shuffle 3) | 0 Hz | 3,937 | 3 | 0 / 0 | 0 / 0 | 0 / 0 |
| Sugar, scrambled (shuffle 4) | 0 Hz | 4,152 | 4 | 0 / 0 | 0 / 0 | 0 / 0 |
| Sugar, scrambled (shuffle 5) | 0 Hz | 4,043 | 1 | 0 / 0 | 0 / 0 | 0 / 0 |
| P9, real wiring, seed 0 | 0 Hz | 1,370 | 41 | 130 / 162 | 7 / 29 | 0 / 0 |
| P9, real wiring, seed 1 | 0 Hz | 1,037 | 25 | 146 / 132 | 11 / 21 | 0 / 0 |
| P9, real wiring, seed 2 | 0 Hz | 1,364 | 44 | 141 / 136 | 10 / 25 | 0 / 0 |
| P9, scrambled (shuffle 3) | 0 Hz | 1,055 | 4 | 130 / 162 | 0 / 0 | 0 / 0 |
| P9, scrambled (shuffle 4) | 0 Hz | 1,016 | 2 | 130 / 162 | 0 / 0 | 0 / 0 |
| P9, scrambled (shuffle 5) | 0 Hz | 1,013 | 5 | 130 / 162 | 0 / 0 | 0 / 0 |
| Left P9 only, real wiring | 0 Hz | 393 | 9 | 131 / 0 | 14 / 0 | 0 / 0 |
| Left P9 only, scrambled (shuffle 3) | 0 Hz | 393 | 2 | 131 / 0 | 0 / 0 | 0 / 0 |
Whose is each part of the walk?
What it shows
- In this model and body, the walk comes from the stimulated command neurons and FlyGym's controller: a scrambled brain and a brainless constant drive walk as far as the real brain.
- The real wiring is not idle: it adds turning commands (DNa02) that scrambled brains lack. In 1 s those turns are as large as the turns of random brainless drives, so they cannot be told apart yet.
- Tasting sugar drives the feeding neuron, not the legs: the fly stands.
Limits
- The mapping is ours and decides how much the wiring can matter. It caps each drive at 1.2, and 90 of the 120 drive values of the six P9 walks sit at the cap, so forward speed could hardly differ by construction.
- One stimulus (P9 at 150 Hz), one model, one body, 1 s of walking, open loop: the body never feeds back to the brain.
- 3 trial seeds; 3 shuffles that behaved as one.
- Changed before the first body run: the turning sign of P9, to match Bidaye et al. 2020 (FlyGym turns toward the slower side). Details in the report of 2 Oct 2026.
Build it yourself: Add a body on Build your own has the tested commands, pins and expected numbers.
New 5 Oct 2026, brain-only follow-ups 6 and 8 Oct 2026, two more body runs 8 Oct · made by us · each pre-registered before its first trial
We ran the map: a looming shadow on one eye
We ran the map. A looming shadow on one eye turned our simulated fly away in 6 of 6 runs. With the same map scrambled (same number of connections per neuron) it turned away in 0 of 3. Here the map's wiring, not the stimulus or our code, makes the turn.
The walking test above showed that FlyGym's legs walk with or without a brain. So this time we fixed the walking drive (a constant 0.8 on both sides, ours, not the brain's) and asked whether the brain can steer: every LC4 and LPLC2 neuron of one eye, the fly's looming detectors, fired at 80 Hz for 1 s, and the brain's turning neurons (DNa02) and backward neurons (MDN) set the difference between the two sides through our hand-made mapping. The chart with every path, the brain table with fly67's own numbers and the full caveats are on Is it real?
- 6 of 6real map: turned away, 34–68° (mean 49.9°; 4 on 5 Oct, 2 on 8 Oct); left shadows right, right shadows left
- 0 of 3scrambled body runs: every readout 0 Hz, the same drive as no brain, so in effect one walk (+3 drives identical to the floor)
- 3 of 6no brain, random steering: "turned away" by chance (all three were right-eye runs)
- 0.50activity of the scrambled brains relative to the real one on 5 Oct: much quieter
Whose is each part of the turn?
6 Oct: as busy as the real brain, still no turn
The 5 Oct scrambles were much quieter than the real brain, so we went back to the brain model alone (no body) and ran four arms with the same shadow. Every sentence below was fixed before the first trial; the rules picked which one applies. The chart, the census of the map and every caveat are on Is it real?
- Pre-registered Scrambled and turned up to the real activity.
We turned the scrambled maps up until they were as busy as the real one. They still did not produce the turn: the wiring, not the amount of activity, makes it.
Reading by the rule: "the turn needs the wiring even at the same activity": 3 shuffles matched the real activity (gain 2.25, outside-activity 0.98 × real, whole brain 27,166–27,515 spikes vs 27,686), 0 of them produced the turn command. - Pre-registered Scrambled inside only (every edge onto descending, motor and endocrine neurons and out of sensory and ascending neurons kept; the rest shuffled).
Scrambling only the inside of the map kept about -9% of the turn command and 69% of the escape signal.
In plain words, −9% means none (the far-side command was absent in all 6); it is slightly negative because in one trial the near-side DNa02 fired 1 Hz. Readings by the rules: turn command "the turn needs the wiring in between" (fly67's DNa02 rule passed in 0 of 6); escape signal "partly" (giant-fibre retention 0.69, below the 0.9 bar for "rides on direct connections"; above 50 Hz in 6 of 6).Descriptive, not pre-registered Per neuron, the kept part is one side: the near-side giant fibre, which gets 832 (left) and 1,053 (right) direct synapses from the looming neurons, kept firing at 151–169 Hz (real map: 129 and 149–156 Hz); the far-side giant fibre went to 0 Hz (real map: 84–104 Hz).
- Pre-registered Giant fibre silenced (every synapse out of both giant fibres, DNp01, set to zero; the fibres themselves still spike).
With the escape neuron silenced, the turn command stayed.
Reading: "steering survives silencing the giant fibre" (fly67's rule passed on both sides; far-side DNa02 changed by −2 Hz and 0 Hz against the paired real trials). This was expected from the wiring: the map has no giant fibre → DNa02 connections at all, so the test checks fly67's claim in our engine, not a hidden route. - Pre-registered Plain scramble (same number of connections per neuron): 6 trials over 3 shuffles, every readout 0 Hz, and the brain about 14 times quieter outside the stimulated neurons (0.07 × real).
8 Oct: six real runs, an inside-only arm, and how strong the input must be
Two more real-map body runs (one per eye) turned away, so the pre-registered rule applies as written: 6 of 6. "With the escape neuron silenced, the fly still turned away." (2 of 2) "With only the inside of the map scrambled, the fly turned away in 0 of 6 runs." In the brain alone: "The turn command appeared from about 30 per second (38% of our strongest input) on both sides; below that the model did not turn." Silencing the relay PVLP141, or all two-step relays, left the turn rule passing, while the far-side rate fell with PVLP141 silenced (left 25 → 17 Hz, right 41 → 24 Hz). One model; the input is a rate we set for every LC4/LPLC2 of one eye, not a measured response to a real shadow. Charts: Is it real?
The ledger row after 8 Oct: helps, fair, strong, now with 6 real runs. Our row byo-flygym-loom reads, by the same fixed rules: Our test: a looming shadow turned the fly away in 6 of 6 real-brain runs, 0 of 3 scrambled body runs (+3 floor-identical); our mapping.
- Real arm: 6 body runs, mean 49.9° away (± 13.12 sd); 4 walked on 5 Oct, 2 on 8 Oct. The no-brain floor run repeated the 5 Oct run exactly.
- Plain scramble arm: 6 drives over 3 shuffles (3 body runs; 3 drives identical to the no-brain floor, whose walk follows by determinism). Turned-up arm: 3 drives at activity 0.99 × real, all identical to the floor.
- New inside-only arm (edges onto output neurons kept): 6 drives over 3 shuffles, -0.2° beyond the floor.
With only the inside of the map scrambled, the fly turned away in 0 of 6 runs.
- In the note, not in the effect:
With the escape neuron silenced, the fly still turned away.
(2 of 2), and the brain-only results of 8 Oct (how strong the input must be; relays silenced), with their caveat. - The body does the walking with its own controller (FlyGym's), and the rate-to-drive mapping and the constant 0.8 walking drive are ours.
17 Oct, updated 18 and 19 Oct: scrambled brains at 30 spikes/s
Pre-registered, brain only. At 30 spikes/s the real brain's turn command is small (far-side DNa02 8 Hz for a shadow on the left eye, 12 Hz on the right).
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: we then strengthened the connections inside scrambled brains (not the input) until their neurons outside the stimulated input fired about as many spikes as in the real brain (within 30%; same input, one 1-second trial each).
- Left eye:
On the left side, 2 scrambled brains matched this way, and neither sent a turn command.
- Right eye:
On the right side, 2 scrambled brains matched this way, and neither sent a turn command.
Matching the number of spikes does not match the pattern: 1,152 to 2,228 neurons fired in the matched scrambled brains, against 379 (left) and 432 (right) in the real brain.
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.
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.
21 Oct, a second input draw (pre-registered): 21 Oct: we repeated the matched test with a second, independent draw of the same input on the same scrambled brains (one 1-second trial each).
- Left eye:
On the left side, both scrambled brains matched again, and neither sent a turn command.
- Right eye:
On the right side, both scrambled brains matched again, and neither sent a turn command.
Matching the number of spikes still did not match the pattern: 1,256 to 2,741 neurons fired in the matched scrambled brains, against 427 (left) and 446 (right) in the real brain with this input.
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.
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.
With a second input draw, the matched scrambled brains again sent no turn command (2 input draws on the same 4 scrambled brains).
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.
Not part of the row's effect. The chart and table: Is it real?; every trial and check: report of 17 Oct 2026; strengthened: report of 18 Oct 2026; the 19 Oct body walk: report of 19 Oct 2026; the second input draw: report of 21 Oct 2026.
Read with the result: one model (Shiu et al. LIF on FlyWire v783), open loop, our hand-made mapping, a constant external walking drive, 1 s; not a real fly. 6 real body runs; the scrambled drives that equal the floor's take its walk by determinism (the floor run repeated exactly on 8 Oct). The direction rests on our mapping's DNa02 sign and fly67's claim; as written before the test: "fly67's claim; the direction in real flies was not verified by us." Retention 0.00 by the ledger's floor method.
Build it yourself: Add a sense: looming on Build your own has the tested commands.
The 107 studies at a glance
- 107control studies in our catalogue, 3 of them our own tests
- 36 of 97with a wiring null: the real wiring beats scrambled wiring
- 38tie (27) or lose (11) against scrambled wiring: trained, reservoir, ML, game, odour and robot-arm tasks, and our walking test
- 7 of 49with a no-brain baseline: the fly network beats it
Where the wiring helps: by task family
The split follows the kind of task. Three of the four reflex-circuit studies find that the real wiring helps; the fourth, our own test, is "mixed" by rule because a sign shuffle makes the whole brain fire. No reservoir study does: both read worse, including fly-arithmetic, where a random network of the same size did better at addition. Machine-learning benchmarks are split. The one scored language-model study, fly-distill, is mixed (a random graph did best), and the forecasting study has no scored wiring null. Sensory models now include a mushroom-body odour model (worse, on its task metric) and MelanoGraph (helps, but weak: one rewired graph). In steering a body, our open-loop walking test is the first "no difference": the walk came from the stimulus and the body controller. Games are split: two untrained game demos added on 6 Oct beat one shuffle each (weak method), while trained game players mostly tie.
| Task family | Helps | No difference | Worse | Mixed | Not tested | Studies |
|---|---|---|---|---|---|---|
| Reflex circuits | 10 | · | · | 2 | 2 | 14 |
| Sensory models | 6 | 2 | 3 | 2 | 2 | 15 |
| Steering a body | 7 | 3 | · | 2 | 2 | 14 |
| Playing games | 10 | 10 | 2 | 3 | 4 | 29 |
| Machine-learning benchmarks | 1 | 7 | 1 | 4 | · | 13 |
| Reservoir computing | 1 | 2 | 2 | 2 | 1 | 8 |
| Language models | · | · | 1 | 2 | 1 | 4 |
| Forecasting | · | 3 | · | · | 1 | 4 |
| Graph analysis, no simulation | 1 | · | 1 | · | 2 | 4 |
| Evolved controllers | · | · | 1 | 1 | · | 2 |
| All studies | 36 | 27 | 11 | 18 | 15 | 107 |
Sensory models, our own test (tried on 10 and 12 Oct 2026): we tried to test whether the real wiring layout helps a trained motion-vision model (flyvis). Neither attempt gave a result, so it adds nothing to the Sensory models row.
How much survives when the wiring is scrambled?
To compare studies that measure very different things (hertz, seconds, accuracy), we use one number per study, null retention: how much of the real wiring's result the scrambled copy keeps. 0 means the scrambled brain does no better than no network at all. 1 means it does as well as the real wiring. Above 1, the scrambled copy does better.
The diamond adds the second question: how well does a no-brain baseline do, as a share of the fly network's result? Look at the Doom row. Scrambled wiring keeps almost nothing (0.03), so this brain needs its wiring to act at all. But a brainless autopilot reaches 0.97 of the fly's survival time, so the task does not need the brain.
The formula is (null − floor) / (real − floor) when a study has a "no network" floor, and null / real otherwise. The chart uses each study's degree-preserving null where it has one (76 of the 97 studies with a wiring null), because it answers the question most readers ask: does who connects to whom matter, beyond how many connections each neuron has?
Explore the studies
Filter by wiring effect, task family, method quality or null fairness, and sort by retention. Every study also has a row in the table below, with links to its catalogue entry and its upstream source.
All 107 studies
| Study | Task family | Result: real wiring vs null | Retention | Wiring effect | Null and method | Numbers from |
|---|---|---|---|---|---|---|
| Reflex circuits (14) | ||||||
| ASCEND: a preregistered re-engine of the Shiu et al. 2024 FlyWire v783 whole-brain LIF model, with readout selectivity tested against nine matched null networks A Real FlyWire wiring drives MN9 at 70 Hz under sugar; all nine matched nulls, including degree-kept rewiring, leave both readouts silent. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The authors compared the real FlyWire v783 whole-brain model with 9 null networks (degree-kept rewiring, weight permutation and random graphs, 3 seeds each) on MN9 firing under sugar. Real 70.35 Hz, every null 0 Hz; a second readout (Johnston's organ to DNg62/DNge078) gives the same pattern, 19.7 Hz vs 0 Hz. Our ledger rules read this as helps. Stat note: values read from the committed file, not re-run; the authors' 6-cell selectivity score is 6 of 6 for real and 4 of 6 for every null, because the four 'off' cells pass trivially when a readout is silent. Brain activity vs real (display only). Activity counted over: unclear. rewired 0.297× reduced; weights 0.272× reduced; random 0.274× reduced |
0.00 | Helps | Fair null Moderate method |
result files commit fbd22d2 checked 20 Oct 2026 |
| Buzz: an agent-operated lab bench on a whole-brain FlyWire LIF simulation, with weight-shuffled controls A The escape neuron fired in 30 of 36 looming trials with intact wiring and in none with shuffled weights, but the input gains were hand-tuned on the intact wiring and only one shuffle seed was used, so we rate the comparison unfair. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author drove the whole FlyWire v783 brain model with looming stimuli (3 seeds, 6 intensities, 2 durations) with the intact wiring and with the edge weights shuffled on the same graph. From the author's committed trial files, counted by us, the giant-fiber escape neuron fired in 30 of 36 looming trials with the intact wiring and in 0 of 36 with shuffled weights, but the input gains were hand-tuned on the intact wiring and this phase used one shuffle seed (in a later phase one of three shuffles kept a weak one-sided response), so we rate it unfair and weak. Not re-run by us. | 0.00 | Helps | Unfair null Weak method |
files, recomputed by us commit f412c26 checked 29 Oct 2026 |
| drosophila-brain-mlx A Reimplemented Shiu model: MN9 67.3 Hz on real wiring, silent on all 5 degree-preserving shuffles with identical input. Catalogue entry · Upstream source |
Reflex circuits open-loop |
MN9 67.3 Hz with real wiring; 0 Hz on all 5 degree-preserving shuffles with the same input. | 0.00 | Helps | Fair null Weak method |
README only commit e417b33 checked 30 Sep 2026 |
| Drosophila connectome brain simulation: a GPU re-run and extension of the Shiu et al. 2024 FlyWire LIF model, with a weight-shuffled control A Sugar input drives MN9 at 67.45 Hz on real FlyWire weights, 0 Hz in all 10 weight-shuffled brains; no no-brain baseline reported, no chance level. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author re-ran the Shiu et al. 2024 whole-brain model on FlyWire v630 weights and drove the sugar neurons at 100 Hz, against 10 brains with the weights shuffled over the existing connections (20 trials each). In the author's committed per-trial file the feeding motor neuron MN9 fired at 67.45 Hz on the real weights (mean computed by us) and at 0.0 Hz in every trial of all 10 shuffled brains; no no-brain baseline is reported, no chance level, and stimulating 21 random central neurons instead gave 0.0 Hz. Our ledger rules read this as helps. Stat note: the shuffle keeps every connection, so it tests the weight arrangement, not the connection graph. | 0.00 | Helps | Fair null Moderate method |
result files commit 144c6e1 checked 23 Oct 2026 |
| FlyScore: a wiring-specificity benchmark that runs seven published fly experiments on MaleCNS v1.0, on the real wiring and on ten scrambled copies A Sugar drives MN9 in 40/40 trials on real MaleCNS wiring and in none of 10 weight-shuffled or rewired copies; escape and courtship survive weight shuffles. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author ran a sugar-to-MN9 test on MaleCNS v1.0 with the real wiring and with 5 weight-shuffled and 5 degree-preserving rewired copies. Real 1.0 (40 trials), weight shuffle 0.0, rewired 0.0. Our ledger rules read this as helps. Not recomputed by us. | 0.00 | Helps | Fair null Moderate method |
result files commit 227340f checked 15 Oct 2026 |
| Drosophila_brain_model (Shiu et al. 2024) A Real FlyWire weights activate MN9 in every run; shuffled weights in 1 of 100: a large, clean wiring effect on one reflex. Project page · Catalogue entry · Upstream source |
Reflex circuits open-loop |
MN9 active in 100% of real-wiring runs vs 1 of 100 runs with shuffled weights (paper). | 0.01 | Helps | Partly fair null Weak method |
paper commit 91bdd1e checked 30 Sep 2026 |
| Matcha Fly: a tea-ceremony game on a MaleCNS v1.0 spiking model, with a pre-registered lab and shuffled-wiring controls A Driving the looming detectors gives 241.7 giant-fibre spikes on the real MaleCNS wiring; none of 100 shuffled wirings reaches half that (highest 21.0). Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author drove the LC4 and LPLC2 looming detectors at 150 Hz in a MaleCNS v1.0 whole-nervous-system LIF model and counted giant-fibre spikes, against 100 shuffled-wiring brains in which each neuron keeps its outputs and synapse counts, in a pre-registered experiment (10 seeds). In the author's committed verdict the real wiring gave a mean of 241.7 spikes and the highest of the 100 shuffled brains 21.0; no no-brain baseline is reported and a spike count has no chance level, and driving random sets of other visual neurons instead gave 2.0 to 18.7 spikes. Our ledger rules read this as helps. Stat note: not re-run by us; the shuffled value is the highest of 100 (the mean is not in the file); the author's two taste experiments failed their own first criterion, so their wiring control was not run. | 0.09 | Helps | Fair null Moderate method |
result files commit 69fffdf checked 22 Oct 2026 |
| Flybrain (fly-brain-sim): a Numba/JavaScript port of the Shiu FlyWire model with browser apps and scrambled-wiring controls B Author's notebook: looming drives the giant fibre at 163-177 Hz with real wiring, 47-73 Hz with shuffled strengths, 0 with scrambled targets. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author compared the real loom circuit with shuffled synapse strengths and with scrambled targets on giant fibre rate. Real 163-177 Hz, shuffled strengths 47-73 Hz, scrambled targets 0 Hz (3 seeds). Our ledger rules read this as helps, using the midpoints of the reported ranges (computed by us). Caveats: the numbers are only in a lab notebook, with no per-seed values or statistics; not recomputed by us. Partly fair, weak. | 0.35 | Helps | Partly fair null Weak method |
README only commit 12f6c77 checked 13 Oct 2026 |
| Cell-type wiring carries most of a whole-brain Drosophila model's accuracy: the Shiu et al. LIF model on FlyWire against four graded null-wiring families A Measured FlyWire wiring scores 0.879 on 157 published tests; degree-kept nulls 0.501, cell-type-block nulls 0.821 (0 of 100 reach it). Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author re-ran a reimplementation of the Shiu et al. whole-brain LIF model on the measured FlyWire v783 wiring and on 100 draws from each of four null families, scoring each on 157 of 164 published experimental tests (balanced accuracy). In the author's committed results the measured wiring scores 0.879 against null means of 0.501 to 0.821 (0 of 100 draws reach it in each family); no no-brain baseline is reported, and 0.5 is chance. Our ledger rules read this as helps. Stat note: family means over 100 draws from the committed analysis, not re-run by us; the weight-shuffle and degree-preserving nulls leave the network largely silent (a response predicted for 0.2 of 157 items on average, against 54 for the measured wiring; post hoc), and the margin over the cell-type block null is narrow and moves with analysis choices (0.039 to 0.064). | 0.57 | Helps | Fair null Moderate method |
result files commit d9b6a9c checked 21 Oct 2026 |
| fly-brain A Weight-shuffled MaleCNS scores 0.40 (0.56 after refit) vs 0.80 on the fitted 17-assay benchmark; real counts matter. Catalogue entry · Upstream source |
Reflex circuits open-loop |
Benchmark score 0.796 real vs 0.559 with shuffled weights after the readout was re-fitted (12 seeds). | 0.70 | Helps | Partly fair null Moderate method |
result files commit 08cf866 checked 30 Sep 2026 |
| Build your own: sugar to MN9 against four scrambled-wiring nulls (our run) Our run A Our run: real wiring 85 Hz; degree-preserving 0 Hz; weight shuffle 0 Hz; interior-only shuffle 39 Hz; sign shuffle runaway. Build your own: control test · Upstream source |
Reflex circuits open-loop |
MN9 85.2 ± 3.56 Hz with real wiring; degree-preserving and weight shuffles 0 Hz (5 shuffles each); interior-only (boundary-preserving) shuffle 39 Hz (retention 0.46); sign shuffle 214 Hz only because the whole brain ignites. Pre-registered; the gain-matched arm was not matched on 3 shuffles (×2.5 matched shuffle 3 only), and MN9 stayed at 0 Hz at every gain. Brain activity vs real (display only). Activity counted over: the whole brain. degree-preserving 0.291× reduced; weight shuffle 0.388× reduced; sign shuffle 257.637× runaway; boundary-preserving 0.648× comparable |
0.00 | Mixed | Unfair null Weak method |
our run commit 91bdd1e checked 2 Oct 2026 |
| Whole-CNS Drosophila wiring specificity A Steering readout survives shuffles that keep cell type and side (4.11 vs 3.98 Hz) and vanishes under a whole-CNS shuffle (-0.01 Hz). Catalogue entry · Upstream source |
Reflex circuits open-loop |
Steering readout (leg motor neuron separation, MaleCNS v1.0, 5 shuffle seeds per null): 4.11 Hz when wiring is shuffled within cell type and side (real 3.98 Hz), -0.01 Hz under a whole-CNS shuffle. So the result depends on the null: mixed. No no-brain baseline was run. Authors' numbers from a GPU run, not re-run by us. | none | Mixed | Fair null Strong method |
files, recomputed by us commit 213b849 checked 10 Nov 2026 |
| FlyLab: compound-perturbation rate and LIF models on MaleCNS v1.0 cuts, each prediction tested against degree-preserving, weight-permuted and random-graph ensembles A Predicted imidacloprid effect on a MaleCNS cut differs from all four null ensembles (-0.678 vs -0.267 Hz rewired); no performance direction, no biological ground truth. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author predicted the imidacloprid effect on a 5k-cell MaleCNS v1.0 cut with the real graph and with 100 graphs each of degree-preserving rewiring, weight permutation, random graphs and sign permutation. Real -0.678 Hz, rewired -0.267, weight permutation -1.102, random -0.974, sign permutation -0.379 Hz. The metric has no better or worse direction, so our ledger does not score it. Not recomputed by us. | none | Not yet scored | Fair null Strong method |
result files commit 7a4de43 checked 15 Oct 2026 |
| Fly School: a MaleCNS v1.0 whole-brain LIF fly taught by a mushroom-body learning rule, with predeclared exams and a two-fly communication loop B After training, the rewarded odour cuts the chosen output 45-49% in 8 of 8 flies; a never-paired odour moves 1-4%. Catalogue entry · Upstream source |
Reflex circuits open-loop |
The author ran the MaleCNS v1.0 whole-brain LIF model with a dopamine-gated rule on Kenyon-cell to MBON synapses and compared a rewarded odour with a never-paired control odour in 8 simulated flies. In the author's committed results the rewarded odour cuts the chosen output by 45-49% in 8 of 8 flies (midpoint 47%, computed by us), the never-paired odour by 1-4%. No scrambled wiring was compared, so our ledger has no wiring null for this row; the control is within the model. Stat note: from the README and committed per-seed files, not re-run by us. | none | No wiring null | Partly fair null Weak method |
files, recomputed by us commit 75b45f0 checked 21 Oct 2026 |
| Sensory models (15) | ||||||
| malecns-visual-motor: FlyVis optic lobe, a MaleCNS v1.0 spiking network and a FlyGym body, against degree-preserving rewired networks A Author compared original MaleCNS wiring with three degree-preserving rewirings on DNp15 direction contrast: +81.2 Hz vs -3.4 to +1.4 Hz, identical input. Catalogue entry · Upstream source |
Sensory models open-loop |
The author compared the original MaleCNS network with 3 degree-preserving rewired networks on DNp15 direction contrast under identical input. Real +81.2 ± 2.3 Hz, rewired -3.4, +1.4 and -0.2 Hz (3 rewired graphs, 3 seeds each). Our ledger rules read this as helps. Caveats: the optic-lobe front end is pretrained (FlyVis); single author, unpublished manuscript; not recomputed by us. Partly fair, strong. | -0.04 | Helps | Partly fair null Strong method |
result files commit 9bd8336 checked 13 Oct 2026 |
| MelanoGraph (akashacorporation/flybrain): FlyWire LIF substrate with pre-registered rewiring controls A Authors compared real wiring with a degree-preserving rewire on message gain: real 0.4941, null 0 (1 null sample). Our rules read this as helps. Catalogue entry · Upstream source |
Sensory models offline |
The authors compared the real wiring with a degree-preserving rewired copy on the receiver's gain over a resampled message. Real 0.4941, null 0 (1 null sample). Our ledger rules read this as helps. One rewired graph per condition and a two-choice task, so any scrambling drops it to chance: partly fair, weak. | 0.00 | Helps | Partly fair null Weak method |
paper commit 45ad05d checked 11 Oct 2026 |
| flydoom A Smell drives the lateral horn +14 Hz with real wiring, ~0 with a degree-preserving shuffle; in-game the fly narrowly beats a command-matched random agent. Catalogue entry · Upstream source |
Sensory models open-loop |
Smell response +14.06 Hz real vs +0.009 Hz with one degree-preserving shuffle; in the game the fly beats a matched random agent only narrowly. | 0.00 | Helps | Unfair null Weak method |
files, recomputed by us commit 86d99a4 checked 30 Sep 2026 |
| Frozen fly connectome as a motion encoder: rewired wiring loses, a Reichardt detector wins A Real wiring beat a rewired copy on motion, but a classical detector beat both and stricter controls erased the gap. Upstream source |
Sensory models offline |
Motion direction from a frozen male-CNS connectome (4 directions, low contrast): real wiring 90.7% vs 59.8% for a degree-preserving rewired copy (+0.309, 95% CI 0.287-0.332, 3 seeds). A classical Reichardt motion detector scores 99.8%, so the no-brain baseline is not beaten, and the author's stricter nulls remove the gap (input-layer locality and E/I make-up). | 0.53 No-brain baseline: 1.14 |
Helps | Fair null Strong method |
result files commit 014a2ae checked 9 Oct 2026 |
| Fly OCR A Intact MaleCNS digit pilot 82% beats three degree-preserving target rewirings (mean 61%), but raw-pixel linear readout (99%) beats the fly circuit. Catalogue entry · Upstream source |
Sensory models offline |
Digit accuracy 82% intact vs 61% over 3 rewirings (200-image pilot); a plain pixel classifier reaches 99%. | 0.71 | Helps | Fair null Strong method |
files, recomputed by us commit 48cf341 checked 30 Sep 2026 |
| Synaptera BANC: a pre-registered test of whether the real wiring of a BANC nerve-cord subgraph, driven by a FlyWire brain model, helps learn a steering readout A Real nerve-cord wiring fitted the turn command better than degree-preserving shuffled wiring in all 10 seeds (0.9773 against 0.9504), but the difference is small, an unconstrained network fitted better (0.9984), and the shuffled wiring scored higher at the held-out 90 deg/s. Catalogue entry · Upstream source |
Sensory models open-loop |
The author trained the edge gains of a 25,818-neuron BANC nerve-cord subgraph, driven by a FlyWire brain model, to produce a turn command, with real and degree-preserving shuffled wiring (10 seeds, pre-registered, sealed test split). Test R2 was 0.9773 with the real wiring against 0.9504 shuffled (paired difference 0.0268, 95% interval 0.0242 to 0.0291), but the difference is small, an unconstrained two-layer network fitted better (0.9984), and at the held-out 90 deg/s the shuffled wiring scored higher (0.9912 against 0.9838); the shuffled arm was still improving at the end of its training budget. Not re-run by us. A separate project from the Synaptera flight test (catalogue entry synaptera, grade B). | 0.97 | Helps | Partly fair null Moderate method |
result files commit bc2dda3 checked 29 Oct 2026 |
| flywire-connectome step 69: is the real wiring in a flat spot? Odour identification with jittered synapses, real wiring vs 10 strength-preserving random wirings (pre-registered) A With every synapse jittered, real wiring identifies odours about as well as strength-preserving random wiring (0.954 vs median 0.959); no no-brain baseline reported. Upstream source |
Sensory models offline |
The author jittered every projection-neuron to Kenyon-cell synapse (lognormal, sigma 0.3) and measured single-trial odour identification, with the real BANC wiring against 10 random wirings that keep each Kenyon cell's input strength (pre-registered). In the author's committed table (BANC right) the real wiring scored 0.954 and the random wirings 0.959 (median of 10, computed by us; range 0.954-0.964); no no-brain baseline is reported. Our ledger rules read this as no difference. Stat note: by the author's own flatness rule the real wiring is 'flatter' in 5 of 5 hemispheres, but the gap is 0.1-0.2 percentage points and inside the range of the random wirings, and the author does not use it as evidence that the real wiring is special. | 1.01 | No difference | Fair null Moderate method |
result files commit bd51de8 checked 23 Oct 2026 |
| flywire-connectome step 68: do activity-only decompositions recover the PN to KC input structure? Real wiring vs 10 strength-preserving random wirings (pre-registered) A Activity-only NMF recovers KC input axes about equally from real (0.575) and strength-preserving random wiring (median 0.589, 10 draws); chance 0.053, BANC right. Upstream source |
Sensory models offline |
The author simulated mushroom-body activity and asked whether activity-only decompositions recover the projection-neuron input structure, with the real BANC wiring against 10 random wirings that keep each Kenyon cell's input strength (pre-registered). In the author's committed table (BANC right, NMF) recovery was 0.575 on the real wiring and 0.589 on the random wirings (median of 10, computed by us; range 0.557-0.618); chance is 0.053. Our ledger rules read this as no difference. Stat note: 8 of 10 random wirings scored at or above the real one; the other hemispheres are mixed in both directions. | 1.02 | No difference | Fair null Moderate method |
result files commit bd51de8 checked 23 Oct 2026 |
| flywire-connectome (mushroom-body rate model on BANC, real wiring vs null wirings) A Authors compared real wiring with strength-preserving nulls on odour d': real 18.32, null 19.45 (100 null samples). Our rules read this as worse. Catalogue entry · Upstream source |
Sensory models offline |
The authors compared the real wiring with strength-preserving null wirings (median of 100) on odour discrimination d'. Real 18.32, null 19.45 (100 null samples). Our ledger rules read this as worse. Only the task metric is scored; the structure statistic is not. Partly fair (per-null recalibration not stated), strong; single author. | 1.06 | Worse | Partly fair null Strong method |
result files commit 8aea8e8 checked 11 Oct 2026 |
| Fly Brain Simulator (Owen-Silva): a Shiu et al. 2024 re-run on FlyWire with an explorer app and a pre-registered mushroom-body continual-learning test against shuffled and uniform wiring A The real PN-KC wiring learned 110 odours less well (57.5%) than shuffled (64.4%) or uniform (66.8%) wiring, pre-registered, 20 seeds. Catalogue entry · Upstream source |
Sensory models open-loop |
The author trained the mushroom body's Kenyon-cell outputs to learn 110 odours in 10 tasks, with the real projection-neuron to Kenyon-cell wiring of the left brain half against shuffled and uniform wiring (pre-registered, 20 seeds). In the author's committed results the real wiring reached 57.5% final accuracy, shuffled 64.4% and uniform 66.8%; no no-brain baseline is reported and chance is about 0.9% (1 in 110). Our ledger rules read this as worse. Stat note: the right half gives the same order (58.9% against 64.0% and 66.8%); a sequentially trained MLP reached 14.5% and one trained on all odours at once 93.2%. | 1.12 | Worse | Fair null Moderate method |
result files commit 3e49246 checked 23 Oct 2026 |
| Fly Hearing Lab A Real hearing wiring was less rhythm-tuned than degree-preserving shuffled wiring (0.70 vs 2.66 Hz); the author finds no robust effect. Catalogue entry · Upstream source |
Sensory models offline |
The author turned sound into input for the hearing neurons of an 8,000-neuron MaleCNS subnetwork (Shiu-style integrate-and-fire neurons) and measured how strongly song-detector and receptivity cells were tuned to the song's pulse rhythm, with real and degree-preserving shuffled wiring (one shuffle seed, 5 simulation seeds). Rhythm tuning was 0.70 Hz (95% interval 0.68 to 0.72) with the real wiring against 2.66 Hz (2.60 to 2.71) shuffled; every male effect was larger on shuffled wiring, and the author calls all effects not robust. Our ledger reads this as worse, rated partly fair and weak. That a higher tuning is the better direction is our reading of the author's hypothesis. | 3.80 | Worse | Partly fair null Weak method |
result files commit f175eeb checked 31 Oct 2026 |
| fly-connectome-study fly-v4 stage 7: KC code of natural odour statistics (DoOR 2.0), real MaleCNS PN to KC wiring vs shuffled and random wiring, few-shot chemical-class accuracy (pre-registered, 100 worlds) A Real MaleCNS wiring: 0.349 5-shot odour-class accuracy vs 0.332 shuffled, 0.329 random; raw responses without KCs 0.351; chance about 0.11. Upstream source |
Sensory models offline |
The author passed natural odour responses (301 DoOR 2.0 odours) through the MaleCNS projection-neuron to Kenyon-cell wiring and classified 9 chemical classes from 5 examples each, with the real wiring against the author's shuffled wiring and random wiring (pre-registered, 100 worlds). In the author's committed results: real 0.349, shuffled 0.332, random 0.329, and raw responses without the Kenyon-cell layer 0.351; chance is about 0.11. Our ledger rules read this as mixed. Stat note: real minus shuffled +0.017 [-0.001, +0.036], real minus random +0.020 [+0.002, +0.040]; the author's planted positive control checks the method, not fly wiring (real minus shuffled +0.064 [+0.058, +0.071] in planted worlds, with a no-Kenyon-cell linear readout level with real at -0.004 [-0.008, +0.001]). | 0.95 No-brain baseline: 1.01 |
Mixed | Fair null Strong method |
result files commit 2f8f472 checked 23 Oct 2026 |
| fly-connectome-study fly-v4 stage 3b: can a student mushroom body with another fly's real PN to KC wiring learn a teacher built on hemibrain wiring better than shuffled or random wiring? (pre-registered, 100 worlds) A Learning a hemibrain-wired teacher's labels, a student with real MaleCNS wiring scores 0.650 vs 0.648 input-total-kept, 0.628 random; no no-brain baseline. Upstream source |
Sensory models offline |
The author let a student mushroom body with another fly's real wiring learn odour labels produced by a teacher built on hemibrain wiring, against a control that keeps each Kenyon cell's input totals and against random wiring (pre-registered, 100 worlds). In the author's committed results the MaleCNS student scored 0.650, the input-total control 0.648 and random wiring 0.628; no no-brain baseline is reported. Our ledger rules read this as mixed. Stat note: real minus random +0.022 [+0.017, +0.027], real minus the input-total control +0.002 [-0.002, +0.006] (the author calls it equivalent); the author's planted positive control checks the method, not fly wiring. | 1.00 | Mixed | Fair null Strong method |
result files commit 2f8f472 checked 23 Oct 2026 |
| flyvis A Connectome-constrained optic-lobe model beats decoder-alone and random-parameter models; ablation values are figure-only, no wiring shuffle. Project page · Catalogue entry · Upstream source |
Sensory models offline |
Predicts the contrast preference of 32 of 32 cell types and beats decoder-only and random-parameter models; no wiring shuffle, ablation values only in a figure. | none | No wiring null | Partly fair null Weak method |
paper commit 92b3845 checked 3 Oct 2026 |
| Synaptera (pre-registered closed-loop flight test of a fly connectome model) B The model's landing neuron responds to expanding flow, but no more than in a shuffled connectome. Catalogue entry · Upstream source |
Sensory models open-loop |
The model's landing neuron DNp07 fires 9.6 Hz more for expanding flow than for a static grating (5 seeds), but the author's shuffled connectome gave larger differences (null maximum 15.6 Hz); 1 of 21 visual descending-neuron tests named in advance passed. From the author's report only; we did not inspect the shuffle method, so the fixed rules class it as "other" and the row stays unscored (baselines only, no wiring null). | none | No wiring null | Unclear null Weak method |
result files commit e93fcbf checked 8 Oct 2026 |
| Steering a body (14) | ||||||
| Flycraft: frozen MaleCNS v1.0 nerve-cord LIF model drives an 18-joint MuJoCo hexapod through a fixed motor map, vs rewired, shuffled-weight, random and zero-drive controls A Author's test: real nerve-cord graph moved the hexapod 0.85 mm in 8 s; fully rewired graphs 3.5 mm backward; a time-shuffled real signal 0.72 mm. Catalogue entry · Upstream source |
Steering a body open-loop |
The author drove an 18-joint MuJoCo hexapod with a frozen MaleCNS v1.0 nerve-cord graph through a fixed hand-made motor map and compared it with fully degree-preserving rewired graphs, shuffled-weight graphs, a time-shuffled copy of the real signal and zero drive, on forward distance in 8 s. Real graph 0.85 mm forward in 8 s (30 realizations), fully rewired graphs 3.5 mm backward, shuffled weights 2.8 mm backward, zero drive 3.3 mm backward, time-shuffled real signal 0.72 mm (the author's p for real vs time-shuffled is 0.29). Our ledger rules read this as helps. Caveats: every distance is under 1 mm forward and the README says no alternating tripod gait forms; the author's results file is marked pending the author's own review, and the paper's acceptance is the author's statement only. Not recomputed by us. Fair, strong. | -0.07 | Helps Caveat A control that keeps the wiring but scrambles or replays the activity came close to the real result |
Fair null Strong method |
result files commit 61c09b0 checked 17 Oct 2026 |
| Build your own, Add a sense: a looming shadow turns the FlyGym fly (our run) Our run A Our test: a looming shadow turned the fly away in 6 of 6 real-brain runs, 0 of 3 scrambled body runs (+3 floor-identical); our mapping. Our looming test · Chart and caveats · Build your own: Add a sense · Upstream source |
Steering a body open-loop |
A looming shadow on one eye (every LC4 and LPLC2 neuron of that eye at 80 Hz) turned the FlyGym fly away in 6 of 6 real-brain runs (mean 49.9°; 4 on 5 Oct, 2 on 8 Oct; left 3 of 3, right 3 of 3) and in 3 of 6 brainless random-steering runs. The scrambled map turned it away in 0 of 3 scrambled body runs (+3 drives identical to the no-brain floor, whose walk follows by determinism). Scrambled maps turned up until they were as busy as the real one (activity ratio 0.99, 3 shuffles) still sent the floor's drive. With only the inside of the map scrambled, the fly turned away in 0 of 6 runs (1 body run + 5 drives identical to the floor). With the escape neuron silenced, the fly still turned away (2 of 2; in the note, not in the effect). Pre-registered; fair with a strong method by the fixed rules. Brain-level results (how strong the input must be, relays silenced) are in the note only. The body does the walking with its own controller; the rate-to-drive mapping and the constant 0.8 walking drive are ours, and the "away" direction rests on our mapping's DNa02 sign and fly67's claim. Brain activity vs real (display only). Activity counted over: the whole brain. degree-preserving 0.4972× reduced; gain-matched degree-preserving 0.9864× (outside the stimulated input 0.98×) comparable; boundary-preserving 0.6615× comparable |
0.00 No-brain baseline: 0.12 |
Helps | Fair null Strong method |
our run commit 91bdd1e checked 8 Oct 2026 |
| Connectocopter (fly connectome drives a simulated drone-rover) A Real wiring escapes 15 of 15 looming balls and a rewired connectome 0 of 15, but a hand-written controller also escapes 15 of 15. Catalogue entry · Upstream source |
Steering a body closed-loop |
Looming-ball escapes (take-off): 15 of 15 with the real FlyWire wiring vs 0 of 15 with one degree- and sign-preserving rewiring (raw files recounted by us). A hand-written controller also escapes 15 of 15 (README), so the no-brain baseline is not beaten; with the 2 giant fibres silenced, 0 of 15 (README). Flight course and landing: real 10 of 16 vs rewired 0 of 16 (recounted). One rewiring seed; readout thresholds tuned on the real wiring. | 0.00 No-brain baseline: 1.00 |
Helps | Unfair null Weak method |
files, recomputed by us commit fb75ff8 checked 8 Oct 2026 |
| Flight-test the fly A Fly yaw-damper circuit: RMS yaw rate 3.19 deg/s vs 3.92 median of 10 degree-preserving shuffles (best 3.31); classical damper 1.69, bare airframe 3.92. Catalogue entry · Upstream source |
Steering a body closed-loop |
An aircraft yaw damper: RMS yaw rate 3.19 deg/s with the real T4/T5-to-DNa02 subcircuit vs 3.92 for the median of 10 degree-preserving shuffles (best shuffle 3.31) on 20 held-out turbulence seeds. The bare airframe gives 3.92 and a classical yaw damper 1.69, so the no-brain baseline is not beaten. | 0.00 No-brain baseline: 3.07 |
Helps Caveat The authors' own test did not separate the real wiring from a control (p = 0.0909) |
Fair null Strong method |
files, recomputed by us commit d3fa996 checked 5 Oct 2026 |
| Are fruit flies zero-shot adapters? A Scrambled wiring silences DNa02 steering: 0.34 vs 2.96 poles per 10 s for real wiring, barely above unplugged turn neurons (0.12). Catalogue entry · Upstream source |
Steering a body closed-loop |
2.96 poles per 10 s real vs 0.34 scrambled (2 shuffles) and 0.125 with the turn neurons disconnected. | 0.08 | Helps | Partly fair null Moderate method |
files, recomputed by us commit 0c63ba5 checked 30 Sep 2026 |
| FLY-lab: What a fly connectome adds to controlling a body A Connectome beats degree-preserving shuffles on turning (100% vs 53% after readout re-fit, 0% right turns), but a two-line rule ties it; replay fails. Catalogue entry · Upstream source |
Steering a body closed-loop |
Left turns in 100% of episodes real vs 52.7% over 5 shuffles with a re-fitted readout (31.3% without re-fit); a two-line rule also scores 100%. | 0.53 No-brain baseline: 1.00 |
Helps | Fair null Strong method |
files, recomputed by us commit c9ee19f checked 30 Sep 2026 |
| SomaHouse: a trained fly-derived circuit runs a simulated house, tested against degree-preserving rewired, dense and no-hub circuits A The fly-derived circuit cost 0.1589 against 0.1622 for a degree-preserving rewired copy, but dense wiring (0.1538) and a conventional controller without a brain (0.1172) both cost less (lower is better). Catalogue entry · Upstream source |
Steering a body closed-loop |
The author trained a 32-cell circuit per module built from MaleCNS v1.0 neurons to run the heating, ventilation and lighting of a simulated house, alongside a degree-preserving rewired copy, dense artificial wiring and a version without the engineered hub (60 held-out 48-hour scenarios, 3 training seeds). In the author's committed results the fly wiring cost 0.1589 against 0.1622 for the rewired copy (paired difference -0.0033, 95% interval -0.0064 to -0.0008; lower is better), but dense wiring (0.1538) and a conventional controller with no brain (0.1172) both cost less than the fly wiring. Three seeds only; the author notes the intervals condition on them. | none | Helps | Fair null Strong method |
result files commit 7e83715 checked 24 Oct 2026 |
| FlyArm B Frozen MaleCNS arm controller: pick-and-place lift 72% vs 56% for degree-preserving shuffles over 6 seeds (p 0.125); grasp tied, kitchen ahead, dexterous tie. Catalogue entry · Upstream source |
Steering a body closed-loop |
A robot arm driven through a frozen MaleCNS between a trained encoder and decoder: pick-and-place lift 72.2% with the real wiring vs 55.9% for 9 degree-preserving shuffles over 6 seeds (p = 0.125) and 56.9% for a GRU. No difference by our fixed rules on this one metric; the author also says the wiring advantage is not established. | 0.77 No-brain baseline: 0.79 |
No difference Caveat The authors' own test did not separate the real wiring from a control (p = 0.125) |
Fair null Strong method |
files, recomputed by us commit ab3922e checked 5 Oct 2026 |
| Build your own, Add a body: FlyWire brain drives FlyGym walking (our run) Our run A Our test: FlyGym fly walks 13.9 mm with real brain, 14.1 mm scrambled, 14.3 mm with a brainless constant drive; the mapping is ours. Our body test · Build your own: Add a body · Upstream source |
Steering a body open-loop |
FlyGym fly walks 13.93 ± 0.58 mm in 1 s with the real brain (3 seeds), 14.15 mm with 3 degree-preserving shuffles that sent identical drives (in effect one walk), 14.28 mm with a brainless constant drive and 13.96 ± 0.46 mm with brainless random drives. P9 stimulated; the rate-to-drive mapping is ours. Brain activity vs real (display only). Activity counted over: the whole brain. degree-preserving 0.818× comparable |
1.02 No-brain baseline: 1.03 |
No difference | Fair null Strong method |
our run commit 91bdd1e checked 2 Oct 2026 |
| Our check: Pugliese et al. 2026 walking model, real front-leg nerve-cord wiring vs 5 shuffles with the authors' own shuffle (our run, 9 Nov 2026) Our run A Our check, 9 Nov 2026: real wiring 3 rhythmic motor neurons; 4 of 5 shuffles still had some (0, 12, 40, 3, 1), but the shuffles were far more active (1,645 to 1,995 active cells against 97), so not like-for-like. Upstream source |
Steering a body open-loop |
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. By the rule we fixed before the run (if at least half of the shuffles still show a rhythm, our row reads no difference), our own row reads no difference. The catalogue entry for the paper stays not tested. Stat note, in plain words: one real network against 5 shuffled networks, one run each, no noise, and looser solver tolerances than the paper (1e-4 and 1e-7 against 2e-6 and 5e-9); the code's stimulus adjustment was off (its default); solver status and traces were not saved; no statistics. Brain activity vs real (display only). Activity counted over: network units. authors-shuffle 18.33× runaway |
none | No difference | Partly fair null Moderate method |
our run commit 10e7661 checked 9 Nov 2026 |
| therealfly: does the MaleCNS v1.0 ventral nerve cord, as an untuned LIF model, make a walking rhythm under DNa01/DNa02 drive? A preregistered test against a degree-preserving scrambled cord (failed) A Driven MaleCNS cord shows a 5-15 Hz peak on 6 of 6 seeds, a degree-kept scrambled cord on 5 of 6: no wiring-specific walking rhythm. Catalogue entry · Upstream source |
Steering a body open-loop |
The author drove a MaleCNS v1.0 nerve-cord model through two descending neuron types and tested for a 5-15 Hz leg rhythm, against a degree-preserving scrambled cord. Seeds with the rhythm: real 6 of 6, scrambled 5 of 6. Our ledger rules read this as mixed. Stat note: the author's own pre-set criterion needed the scrambled cord to fail on at least 5 of 6 seeds; it failed on 1, so the author's criterion does not credit the rhythm to the wiring; the band peak is a harmonic of a fast runaway oscillation in both arms. Brain activity vs real (display only). Activity counted over: unclear. scrambled 3.57× elevated |
0.83 | Mixed | Fair null Moderate method |
result files commit 366ceee checked 20 Oct 2026 |
| Fly.exe (MaleCNS Virtual Fly) B Regime-matched degree-preserving shuffle still drives the giant fibre but loses left/right selectivity (+2.0 vs -0.1); one seed. Catalogue entry · Upstream source |
Steering a body closed-loop |
Left/right selectivity +2.0 real vs -0.1 shuffled; the giant fibre still fires after shuffling; one seed. | none | Mixed | Partly fair null Weak method |
files, recomputed by us commit cc25411 checked 30 Sep 2026 |
| Flyhard (The Driving Fly) B Trained fly-topology rate network steers a wheel on 100/100 held-out targets vs 0/100 untrained; no wiring null or non-fly baseline. Catalogue entry · Upstream source |
Steering a body closed-loop |
Wheel steering 100 of 100 held-out targets after training vs 0 of 100 untrained; no wiring null or non-fly baseline. | none | No wiring null | Partly fair null Weak method |
result files commit 328906f checked 30 Sep 2026 |
| NeuroCraft Fly C A shuffled-weights run is shown on video, but no numbers or code are published, so the control cannot be checked. Catalogue entry · Upstream source |
Steering a body closed-loop |
A shuffled-weights run is shown in a video; no numbers or code are published. | none | No wiring null | Unclear null Weak method |
catalogue text only commit d121466 checked 29 Sep 2026 |
| Playing games (29) | ||||||
| fruitfly-lab (Fly Marksman): a 3,963-neuron FlyWire v783 LC10a-to-steering LIF circuit aims and fires in a browser game and in ViZDoom, against degree-preserving scrambled wiring A In ViZDoom the real FlyWire steering circuit scores 10.2 kills per episode; degree-kept scrambled wiring, recalibrated, scores 0.7; a no-brain oracle 11.5. Catalogue entry · Upstream source |
Playing games closed-loop |
The author ran a 3,963-neuron FlyWire v783 steering circuit in the Doom game defend_the_center and compared it with degree-preserving scrambled wiring that got its own calibration (10 episodes). Kills per episode: real 10.2, scrambled 0.7, while a no-brain oracle that reads the enemy direction directly gets 11.5 (oracle 11.5 vs fly 10.2 kills). Our ledger rules read this as helps against scrambled wiring; the no-brain baseline does as well. Stat note: the author's p = 0.0019 vs scrambled (test not named in the file); the oracle is not different from the fly (p = 0.35); silencing AOTU025 gives 5.5 kills. | -0.02 No-brain baseline: 1.14 |
Helps | Fair null Strong method |
result files commit 8c27fe8 checked 20 Oct 2026 |
| FlyBrain · Flappy (escape reflex) B Untrained FlyWire whole brain flaps from DNp01: mean score 100.8 over 40 games vs 0.00 and no flaps for shuffled or cut wiring; README only. Catalogue entry · Upstream source |
Playing games closed-loop |
Flappy Bird through the FlyWire looming-escape pathway: mean score 100.8 over 40 games with the real wiring vs 0.00 with shuffled wiring or the direct LC4/LPLC2-to-giant-fibre edges cut (author-reported, README). The two input gains were tuned on the real wiring and not re-tuned for the scrambled graphs, so the null is unfair by the fixed rule. | 0.00 No-brain baseline: 0.00 |
Helps | Unfair null Weak method |
README only commit 7a8b601 checked 6 Oct 2026 |
| flybrain-intransitive: the untrained MaleCNS v1.0 LIF connectome scores rock-paper-scissors-chess moves through looming and small-object inputs, against scrambled wiring A Untrained measured wiring wins 55 of 100 games against a greedy bot, scrambled wiring 0; a no-brain greedy rule also wins 55. Catalogue entry · Upstream source |
Playing games closed-loop |
The author used an untrained MaleCNS v1.0 network to score moves in rock-paper-scissors chess and compared it with scrambled-target wiring, 100 games against a greedy bot. Wins: real 55, scrambled 0, while a greedy rule with no brain also gets 55 (greedy rule 55 vs 55 wins). Our ledger rules read this as helps against scrambled wiring; the no-brain rule does as well. Stat note: the scrambled network gives no approach-minus-escape response to any feature, so its moves are ties resolved at random; one scramble seed. | 0.00 No-brain baseline: 1.00 |
Helps | Partly fair null Weak method |
result files commit f22c07a checked 20 Oct 2026 |
| Is the fly brain actually playing DOOM? (control experiments) A Untrained fly brain beats shuffled wiring in Doom (53 s vs 7 s), but a brainless autopilot nearly matches; trained RL shows no wiring effect. Catalogue entry · Upstream source |
Playing games closed-loop |
Survival 53.3 s real vs 6.8 s over 3 shuffles; the brainless matched autopilot survives 51.9 s, and the best of 12 autopilot settings 53.4 s. | 0.03 No-brain baseline: 0.97 |
Helps | Fair null Strong method |
files, recomputed by us commit 6b22922 checked 30 Sep 2026 |
| FlySnake: MaleCNS v1.0 LIF network plays Snake through a trained descending-neuron readout; held-out test vs rewired graph A Held-out test: real MaleCNS wiring 20.03 food per game, rewired graph 3.63 (100 games); a hand-written rule scores 21.52. Catalogue entry · Upstream source |
Playing games closed-loop |
The author played Snake with a MaleCNS v1.0 LIF network through a readout trained on 1,314 descending-neuron spike counts, and compared it on 100 held-out games with a rewired graph (each neuron's outgoing connections moved to another neuron) that got its own trained readout. Food per game: real 20.03, rewired 3.63 (paired 95% interval of the difference 15.32-17.46); random actions 0.17. A hand-written Snake rule scores higher, 21.52. Our ledger rules read this as helps. Caveats: a trained readout, one rewired graph, single author; an earlier version with a different readout scored 4.83 vs 1.92. Not recomputed by us. Partly fair, moderate. | 0.18 No-brain baseline: 1.07 |
Helps | Partly fair null Moderate method |
result files commit c3e7787 checked 18 Oct 2026 |
| Connectome ping pong (fly tennis) A Untrained MaleCNS flies play ping pong: 54 hits in one 40 s match vs 10 with shuffled wiring and 14 parked; one match each. Catalogue entry · Upstream source |
Playing games closed-loop |
Two untrained MaleCNS flies rally a ball: 54 hits in one 40 s match with the real wiring vs 10 with one degree-preserving shuffle and 14 with a parked fly that does not steer (committed logs). One match per condition, one seed; the logged closed-loop steering check is near zero, so the long rally is not shown to come from tracking. | 0.19 No-brain baseline: 0.26 |
Helps | Partly fair null Weak method |
result files commit c0b5e17 checked 6 Oct 2026 |
| Fly-NAF B On one replayed game night the real wiring answered 7 of 8 threats on the threat's side first; three shuffled brains answered 0, 3 and 3. The game constants were tuned on the real wiring, so we rate the comparison unfair. Catalogue entry · Upstream source |
Playing games open-loop |
The author fed one recorded night of Five Nights at Freddy's (2,618 frames of eye and tablet input) through the FlyWire v783 spiking brain and through three shuffled copies that keep each neuron's in- and out-degree and the weights it sends (open loop). The real wiring answered 7 of 8 hallway threats on the threat's side first; the shuffled brains answered 0, 3 and 3 (mean 2.0 ± 1.73, computed by us), and they fired the giant fibre on 163 to 405 frames against 18. Swapping the eyes gave 0 of 8; cutting the camera inhibitors kept 7. In a second test, after a sighting the real wiring looked back at that side 24 of 24 times, two shuffles 13 of 24 each and a brain without synapses 4 of 24. Our ledger reads this as helps, rated unfair (the game constants were calibrated on the real wiring, so the shuffles ran with settings made for another network) and weak (one night, raw logs not committed, no rule-based baseline without a brain). Brain activity vs real (display only). Activity counted over: unclear. shuffled (frames with a giant-fibre spike, not network-wide) 15.98× runaway |
0.29 | Helps | Unfair null Weak method |
files, recomputed by us commit 64cdef2 checked 31 Oct 2026 |
| Brain Runners A An untrained fly brain survives 82 rows against 28 on shuffled wiring, but a brainless version of the same rule reaches 74. Catalogue entry · Upstream source |
Playing games closed-loop |
Rows survived 82.34 with real wiring vs 27.84 on one shuffled wiring; the same rule with no brain reaches 74.05 (held-out seeds). | 0.34 No-brain baseline: 0.90 |
Helps | Unfair null Weak method |
README or docs commit 79d8039 checked 1 Oct 2026 |
| duth-x-stepmania: a MaleCNS looming-escape circuit as StepMania's second player, with rewired-circuit controls A Authors compared real wiring with degree-preserving rewired copies on StepMania hit rate: real 0.9737, null 0.4013 (3 null samples). Our rules read this as helps. Catalogue entry · Upstream source |
Playing games offline |
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. Caveats: one seed and no statistics; the readout groups (which cells count as up or down presses) are taken from the real wiring and reused on the 3 rewired copies; there is no no-brain baseline. Fair, moderate. | 0.41 | Helps | Fair null Moderate method |
result files commit c6b829d checked 12 Oct 2026 |
| Drosophila Cadet: the MaleCNS v1.0 looming-escape circuit plays 3D Pinball, with six degree-preserving shuffled circuits A The real escape circuit outscores six degree-kept shuffles at pinball, but a hand-written reflex scores higher still (11.48 vs 11.22). Catalogue entry · Upstream source |
Playing games closed-loop |
The author let a 373-neuron MaleCNS v1.0 looming-escape circuit press the flippers of 3D Pinball and compared it with six degree-preserving shuffles, a hand-written reflex and never pressing (96 paired lives, score as log(1 + points)). In the author's committed log the real circuit scores 11.221 against 10.770-10.985 for the shuffles, while a no-brain hand-written reflex scores higher, 11.480 (never pressing 9.853). Our ledger rules read this as helps against scrambled wiring; the no-brain reflex does better. Stat note: values from the committed log, not re-run by us; the author's paired t per shuffle runs from -2.26 to -4.49; a separately trained 444-neuron brain shows no difference (real 11.186, shuffles 11.209-11.311); the author notes the comparison was framed after the fact. | 0.69 No-brain baseline: 1.19 |
Helps | Fair null Strong method |
files, recomputed by us commit ea73efc checked 21 Oct 2026 |
| making-fly-play-chess C FlyWire-patch chess agent ties its rewired twin (0.506, CI spans 0.5) and loses to a material counter; author-labelled code-check run. Catalogue entry · Upstream source |
Playing games closed-loop |
Match score 0.506 vs 0.494 for its rewired twin (the interval spans 0.5); a material-count player scores 0.694. | 0.98 No-brain baseline: 1.37 |
No difference | Fair null Strong method |
files, recomputed by us commit 4ddc9b5 checked 30 Sep 2026 |
| Fly Tic-Tac-Toe: a 98-cell MaleCNS v1.0 circuit with a trained 425-parameter readout picks tic-tac-toe moves, against a degree-preserving rewired graph at equal training budget A A degree-kept rewired 98-cell graph, retrained at equal budget, plays tic-tac-toe as well as the measured MaleCNS circuit (0.850 vs 0.854). Catalogue entry · Upstream source |
Playing games closed-loop |
The author played tic-tac-toe with a 98-cell MaleCNS v1.0 circuit and a trained readout, against a degree-preserving rewired graph retrained with the same budget. Score against a random player (1,000 games): real 0.854, rewired 0.850; a plain board network without the circuit 0.905. Our ledger rules read this as no difference. Stat note: against perfect play the real circuit loses 11 of 200 games and the rewired one 19 of 200; the number of rewiring draws is not stated; one checkpoint seed. | 0.98 No-brain baseline: 1.24 |
No difference | Partly fair null Weak method |
result files commit bc5d32f checked 20 Oct 2026 |
| brainlab (FlyWire whole-brain LIF run with real, shuffled and disconnected wiring) A Authors compared real wiring with a target-permuted copy on decision accuracy: real 92.91, null 92.20 (1 null sample). Our rules read this as no-difference. Catalogue entry · Upstream source |
Playing games offline |
The authors compared the real wiring with a target-permuted copy on held-out decision accuracy (%). Real 92.91, null 92.20 (1 null sample). Our ledger rules read this as no-difference. One seed and one permutation, no uncertainty reported: partly fair, weak. The spiking-activity comparison in the same repository is not scored. | 0.99 No-brain baseline: 1.08 |
No difference | Partly fair null Weak method |
result files commit 1fe1aa4 checked 11 Oct 2026 |
| ChessFly A Chess-learning mushroom body: real FlyWire wiring rates 1680, shuffled claws 1675 and a random design 1850 after 10,000 games (one fly each). Catalogue entry · Upstream source |
Playing games closed-loop |
Chess rating after 10,000 self-play games: real FlyWire mushroom-body wiring 1680 Elo, shuffled claws 1675, the original random design 1850 (one fly and one shuffle per arm, 40-game exams). A clean negative result. | 1.00 | No difference | Partly fair null Weak method |
result files commit 52276aa checked 3 Oct 2026 |
| MaleCNS_training: a trainable recurrent network on a MaleCNS v1.0 subgraph compared with a rewired graph, an RNN and a GRU on 12x12 gomoku and memory tasks A Trained MaleCNS-subgraph network reaches 0.787 held-out gomoku tactical accuracy vs 0.785 rewired, 0.788 RNN and 0.790 GRU over three seeds. Catalogue entry · Upstream source |
Playing games open-loop |
The author compared a trainable recurrent network on a MaleCNS v1.0 subgraph with a rewired graph, an RNN and a GRU on held-out 12x12 tactical positions. Accuracy: fly 0.787, rewired 0.785, RNN 0.788, GRU 0.790. Our ledger rules read this as no difference. We did not read the rewiring code. Not recomputed by us. | 1.00 No-brain baseline: 1.00 |
No difference | Partly fair null Moderate method |
result files commit a1f8bbf checked 15 Oct 2026 |
| Fruit Fly Plays Fruit Ninja: a 4,386-neuron MaleCNS circuit with a fitted readout, against a rewired circuit and a pixel heuristic A Real and rewired circuits score alike at Fruit Ninja (24.85 vs 24.80), level with a no-brain pixel rule (24.95). Catalogue entry · Upstream source |
Playing games closed-loop |
The author played Fruit Ninja with a 4,386-neuron MaleCNS v1.0 rate circuit and a fitted readout, against a target-permuted circuit with its own readout and a no-brain pixel rule (20 test episodes). In the author's committed results: real 24.85, rewired 24.80, pixel rule 24.95 mean reward. Our ledger rules read this as no difference. Stat note: values from the committed evaluation file, not re-run by us; one rewired graph only; blinded input scores 0. | 1.00 No-brain baseline: 1.00 |
No difference | Partly fair null Moderate method |
result files commit 8b7e297 checked 21 Oct 2026 |
| Flybrain (yyk2007): a MaleCNS v1.0 whole-CNS rate model with a trained logistic flap decoder flies a 3D gate course, with silenced and presynaptic-shuffled checks A Both measured and presynaptic-shuffled MaleCNS wiring clear all 15 gates with the same decoder; the capped task cannot tell them apart. Catalogue entry · Upstream source |
Playing games closed-loop |
The author flew a game character through gates with a MaleCNS v1.0 rate model and a trained flap decoder, and compared it with presynaptic-ID shuffled wiring on held-out courses. Gates cleared (cap 15): real 15, shuffled 15. Our ledger rules read this as no difference. Stat note: both arms hit the 15-gate cap on every course, so the task cannot separate them; the decoder was trained on the real wiring only; the author states the result does not establish a wiring advantage. | 1.00 | No difference | Partly fair null Weak method |
result files commit 8ab0452 checked 20 Oct 2026 |
| doomfly-rl A Doom agent with MaleCNS-49k backbone scores like its degree-preserving shuffle and below a no-connectome model after GRPO; negative result. Catalogue entry · Upstream source |
Playing games closed-loop |
Episode return 6.57 real vs 6.71 with one shuffle; a model without the connectome scores 7.39. | 1.02 No-brain baseline: 1.13 |
No difference | Partly fair null Moderate method |
files, recomputed by us commit 334d915 checked 30 Sep 2026 |
| Fly Touhou: a MaleCNS v1.0 rate circuit with a 320-parameter evolved readout plays a Touhou bullet-hell, against a degree-preserving rewired circuit A A rewired circuit dodges bullets as long as the real one (1,398 vs 1,386 frames); standing still lasts 886. Catalogue entry · Upstream source |
Playing games closed-loop |
The author dodged bullets in a Touhou-style game with a 600-neuron MaleCNS v1.0 rate circuit and an evolved readout, against one degree-preserving rewired circuit trained the same way (20 test seeds). In the author's committed results: real 1,386 frames survived, rewired 1,398, standing still 886. Our ledger rules read this as no difference. Stat note: p is the author's; not re-run by us; a random readout survives 685 frames, and re-training with another seed moved the mean by 63 frames. | 1.02 No-brain baseline: 0.00 |
No difference Caveat The authors' own test did not separate the real wiring from a control (p = 0.9351) |
Partly fair null Moderate method |
result files commit a7984c9 checked 21 Oct 2026 |
| FlyAim A Whole MaleCNS LIF aiming a crosshair: 412.5 px mean target distance vs 430.9 shuffled (p 0.66); both worse than a random walker (251.7). Catalogue entry · Upstream source |
Playing games closed-loop |
Aiming a crosshair with the whole MaleCNS, pre-registered: mean distance to the target 412.5 px with the real wiring vs 430.9 px with one shuffle (lower is better; not different, p = 0.66). A uniform random walker reaches 251.7 px and a PID controller 155.8 px. | none | No difference Caveat The authors' own test did not separate the real wiring from a control (p = 0.66) |
Partly fair null Moderate method |
files, recomputed by us commit 15d800f checked 5 Oct 2026 |
| alpha-fly: the full MaleCNS v1.0 wiring with trained per-synapse gains and linear input and output maps, taught chess by AlphaZero-style self-play, with a degree-preserving shuffled-partner control run A In chess self-play, the shuffled-partner control reached higher Elo than the measured MaleCNS wiring at every matched game count (958 vs 746). Catalogue entry · Upstream source |
Playing games closed-loop |
The author trained per-synapse gains on the full MaleCNS v1.0 wiring to play chess by self-play, and ran the same training with shuffled presynaptic partners (degree kept). At about 39,000 games: real 746 Elo, shuffled 958. Our ledger rules read this as worse. Stat note: rows read by us at matched game counts with the author's two-standard-error rule (difference -212 Elo, combined standard error about 35); the shuffled run also leads at 10,000, 20,000 and 30,000 games; the control stopped at about 12% of the main run's length; one shuffle only. | 1.28 | Worse | Partly fair null Moderate method |
result files commit e3b1fe9 checked 20 Oct 2026 |
| fly-chess: the MaleCNS v1.0 brain as a chess player, trained by backprop or by mushroom-body dopamine learning B Under the same readout probe the shuffled wiring matched Stockfish's move more often than the real MaleCNS wiring (20.9% against 18.0%; board-blind 11.7%). Catalogue entry · Upstream source |
Playing games offline |
The author froze a MaleCNS v1.0 rate model, fed the chess board through a photoreceptor grid and trained only a linear move readout on the descending neurons, then ran the same short readout probe with the connection targets shuffled and with brain regions cut (4,096 held-out positions). The shuffled wiring matched Stockfish's best move more often than the real wiring (20.9% against 18.0% top-1; board-blind baseline 11.7%); cutting the central brain gave 17.9% and cutting the optic lobe 15.1%. On the author's Elo ladder the shuffled-wiring fly rated 682 against 653 and 631 for comparable real-wiring flies. Single runs, one shuffle seed; the longest-trained fly has no shuffled counterpart. | 1.46 | Worse | Partly fair null Weak method |
result files commit 5e4d18e checked 24 Oct 2026 |
| Fly Self Driving A Trained fly graph drives streets 97% vs 80% rewired and 90% for a small MLP; one seed, and the flat-road task showed no wiring advantage. Catalogue entry · Upstream source |
Playing games closed-loop |
Streets completed: 96.7% real vs 80% rewired (one seed) and 90% for a small MLP. | 0.83 No-brain baseline: 0.93 |
Mixed | Partly fair null Weak method |
README only commit 3516a09 checked 30 Sep 2026 |
| Fly-Racer C PPO car racer on a 3,350-neuron MaleCNS core: 903 mean score vs 884 shuffled, 897 sign-shuffled, 894 MLP; one seed each, README only. Catalogue entry · Upstream source |
Playing games closed-loop |
CarRacing score 903.2 with a 3,350-neuron MaleCNS steering-pathway core trained by PPO vs 883.7 for a degree-preserving shuffle, 896.9 for shuffled signs and 894.0 for a plain MLP (100 held-out episodes, one training seed per core). README numbers only; the run files are not committed. | 0.98 No-brain baseline: 0.99 |
Mixed | Partly fair null Weak method |
README only commit 71ede00 checked 6 Oct 2026 |
| fly-plays-games (Pokémon Red chapter; formerly fly-plays-pokemon) C In Arkanoid a weight-shuffled fly brain plays as well as the real one (both hit the 1500-frame cap); rewired or silenced wiring fails by 204. Catalogue entry · Upstream source |
Playing games closed-loop |
Arkanoid: 1,500 frames (the cap) with real and weight-shuffled wiring; rewired or silenced wiring fails at 204; a scripted tracker reaches 1,372. | 1.00 No-brain baseline: 0.90 |
Mixed | Partly fair null Weak method |
README only commit ecb95f7 checked 30 Sep 2026 |
| DOOMFLY B DOOMFLY's controls only show that its outputs depend on edges and input (forward 16.7 intact, 0 without edges); wiring specificity untested here. Project page · Catalogue entry · Upstream source |
Playing games closed-loop |
Forward command 16.7 intact vs 0 without edges in one 500 ms trial: the output depends on the edges, but wiring specificity is untested. | none | No wiring null | Unclear null Weak method |
result files commit 71ecf53 checked 30 Sep 2026 |
| Fly Dino (flyjump) C 80-neuron fly circuit plus trained readout survives 179 s on held-out Dino courses; a 243-parameter network without the connectome survives 180 s. Catalogue entry · Upstream source |
Playing games closed-loop |
Survives 179.37 s on 100 held-out 180 s courses with the 80-neuron circuit and a trained readout; a 243-parameter network with no connectome, trained the same way, survives 180.00 s (147 parameters: 176.89 s); a hand-written rule 46.49 s. No wiring null. Corrected on 6 Oct 2026: the no-brain baseline is not beaten (it was listed as beaten). | none No-brain baseline: 1.00 |
No wiring null | Partly fair null Moderate method |
files, recomputed by us commit e34c661 checked 6 Oct 2026 |
| Fly Worker A After a bias fix the fly covers 50% of a game map vs 63% for smoothed noise; a brain shown the wrong screen scores 66%. Catalogue entry · Upstream source |
Playing games closed-loop |
Author's numbers after a steering-bias fix (10 Nov 2026 re-read): map coverage 50% for the whole-brain fly vs 63% for smoothed noise and 49% for uniform noise; a fly brain shown a different board scores 66%. The author withdraws the earlier -3 point paired result. No wiring null; not re-run by us. | none No-brain baseline: 1.26 |
No wiring null | Unclear null Weak method |
README only commit c1bed69 checked 10 Nov 2026 |
| Haltere C Connectome brain finished 0 of 2 frozen races; a plain PD controller finished 1 of 2. No wiring null. Catalogue entry · Upstream source |
Playing games closed-loop |
Races finished: 0 of 2 for the connectome brain, 1 of 2 for a PD controller; no wiring null. | none | No wiring null | Unclear null Weak method |
README only commit 3e3e7b5 checked 30 Sep 2026 |
| Machine-learning benchmarks (13) | ||||||
| fly-cartpole A Fly flight circuit balances CartPole (499.9 steps) and shuffled wiring drops it to 165.9 (p = 0.0001), but its own linear map also reaches 499.6. Catalogue entry · Upstream source |
Machine-learning benchmarks closed-loop |
Redesigned on 3 Oct 2026 (commit a8c6257) around a MaleCNS flight-stabilisation circuit: 499.9 steps with the real circuit vs 165.9 with degree-preserving shuffled wiring (20 seeds, p = 0.0001). But the circuit's own linear map, tuned the same way, scores 499.6 and LQR 500: the no-brain baseline is not beaten. The real arm sits at the 500-step ceiling. The earlier mushroom-body result (392.5 vs 390.3, p = 0.48) is still in the repository. | 0.30 No-brain baseline: 1.00 |
Helps Caveat A control without the fly wiring came close to the real result |
Fair null Strong method |
result files commit a8c6257 checked 3 Oct 2026 |
| Maxwell's Demon: Inside the Fly Brain (FlyWire FAFB v783 spiking network as a door-keeping demon, against a degree-preserving shuffled wiring) A Real FlyWire wiring keeps the demon's door slightly better than shuffled wiring in the author's short runs, but random door decisions score higher than both. Catalogue entry · Upstream source |
Machine-learning benchmarks closed-loop |
The author ran a FlyWire v783 spiking network with dopamine-gated plasticity as a Maxwell's-demon door between a hot and a cold gas chamber, against a degree-preserving shuffled wiring and random decisions, the no-brain baseline. In the author's committed benchmark summary (3 seeds) the test-phase good:bad crossing ratio (higher is better by the author's write-up) was 0.645 for the real wiring, 0.597 for the shuffled wiring and 1.035 for random decisions, so random decisions scored above both wirings. Our ledger rules read this as no difference. Stat note: not re-run by us; in a 2-seed run the ratios were 0.893, 0.391 and 1.091; the author reduced the sensory drive because real and shuffled wiring behave the same at full drive. | 0.93 | No difference | Fair null Strong method |
result files commit d56568a checked 22 Oct 2026 |
| FlyGateDemo: a MaleCNS v1.0 olfactory circuit whose readout weights change to block scanner probes on a real blog, with a shuffled and random wiring control A Real olfactory wiring blocks 59% of probe families after one lesson; shuffled wiring 54-59%, random 64-68% with more readers blocked. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
The author's C# rate model of the MaleCNS v1.0 olfactory circuit blocks scanner probes on a blog after one punishment per probe family, compared with shuffled and random projection-neuron to Kenyon-cell wiring (3 seeds each). By the numbers reported in the README: real 59% of later probes blocked, shuffled 54-59%, random 64-68% (with more readers blocked). Our ledger rules read this as no difference. Stat note: no result file is committed and we did not re-run it; a no-brain baseline, cosine similarity on the 53 receptor channels, catches 62-98% of probes per family at the same reader false-alarm rate (a per-family range, so it is not entered as an arm). | 0.96 | No difference | Fair null Weak method |
files, recomputed by us commit a6ea6b5 checked 21 Oct 2026 |
| malecns-keratitis-pilot: a frozen MaleCNS v1.0 visual circuit as image features for keratitis classification, against rewired circuits A Author compared a frozen MaleCNS visual circuit with 10 rewired circuits on keratitis images: AUROC 0.479 vs 0.464, both near chance. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
The author compared the real circuit with 10 degree-preserving rewired circuits on telling bacterial from fungal keratitis (AUROC, 0.5 is chance). Real 0.479, rewired mean 0.464; the paired 95% interval of the difference is -0.037 to +0.064, so it spans 0. Our ledger rules read this as no difference. Both are near or below chance, and an image model without the circuit scored 0.754. The clinical data are private, so we could not recompute it. Fair, strong; a negative result reported by the author. | 0.97 No-brain baseline: 1.57 |
No difference | Fair null Strong method |
result files commit bea3cb1 checked 13 Oct 2026 |
| Does the larval connectome beat its own shuffles? (connectome-null-models) A Larval connectome reservoir ties its degree-preserving shuffles, Erdos-Renyi and weight shuffles on CIFAR-10 and MNIST; only removing recurrence hurts. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
CIFAR-10 accuracy 43.04% real vs 42.95% over 15 degree-preserving shuffles (p = 0.51). | 1.00 | No difference Caveat The authors' own test did not separate the real wiring from a control (p = 0.51) |
Fair null Strong method |
files, recomputed by us commit be301d9 checked 30 Sep 2026 |
| Boltzmann-fly: a Deep Boltzmann Machine world model with couplings masked to the MaleCNS v1.0 mushroom-body wiring, against rewired masks A Masking a Boltzmann world model to mushroom-body wiring predicts as well as a rewired mask or a plain MLP (AUC 0.712, 0.712, 0.713). Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
The author masked the couplings of a Deep Boltzmann Machine world model to the MaleCNS v1.0 mushroom body and compared it with a degree-preserving rewired mask and a plain MLP on predicting visits. In the author's results: AUC real 0.712, rewired 0.712, MLP 0.713. Our ledger rules read this as no difference. Stat note: one seed for data, mask and training; values from the README table and committed run files, not re-run by us. | 1.00 No-brain baseline: 1.00 |
No difference | Partly fair null Weak method |
result files commit ac78a51 checked 21 Oct 2026 |
| NeuroWeave C 150-neuron MaleCNS mask scores 98% versus 99-100% for configuration-model, random and dense masks on a ceiling task; single seed. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
Accuracy 98% for the fly mask vs 99% for a configuration-model null; one seed, near the ceiling. | 1.01 | No difference | Partly fair null Weak method |
result files commit 518f162 checked 30 Sep 2026 |
| MaleCNS Cube: trainable network on a MaleCNS v1.0 subgraph estimates Rubik's cube distance for a budgeted search, against a degree-preserving rewired graph, a frozen graph and an MLP A On a 256-node MaleCNS subgraph the rewired graph (MAE 0.338) and a matched MLP (0.325) do as well as the measured wiring (0.351), 3 seeds. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
The author trained a recurrent network on a 256-node MaleCNS v1.0 subgraph to estimate how many face turns a Rubik's cube state is from solved, and compared it with a degree-preserving rewired graph trained the same way, the real graph with frozen connection strengths, and an MLP with a matched parameter count. Test error (MAE in moves, lower is better; 3 seeds): real 0.3509, rewired 0.3380, frozen 0.5275, MLP 0.3245; all four solve 12 of 12 search tests. Our ledger rules read this as no difference. Caveats: a small trained subgraph, not a simulation of fly dynamics; 3 seeds; a course project; the deployed 2,048-node model has no wiring control. Not recomputed by us. Fair, strong. | none | No difference | Fair null Strong method |
result files commit 1917deb checked 19 Oct 2026 |
| The Fly's Hash Function A Real hemibrain PN-to-KC wiring hashes worse than degree-preserving and random FlyHash nulls on MNIST, Fashion-MNIST and correlated odours; beats SimHash. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
Precision@16 0.499 real vs 0.552 degree-preserving: the real wiring hashes worse; it beats SimHash (0.241). | 1.11 No-brain baseline: 0.48 |
Worse | Fair null Strong method |
files, recomputed by us commit 271cc85 checked 30 Sep 2026 |
| flybench A Benchmark of 28 circuit tasks: real FlyWire scores 0.60 vs 0.40 on one degree-preserving rewiring (specificity +0.20). Catalogue entry · Upstream source |
Machine-learning benchmarks open-loop |
Mean task score 0.599 real vs 0.399 for one rewiring, over 28 circuit tasks. | 0.67 | Mixed | Partly fair null Weak method |
files, recomputed by us commit 3052ce5 checked 30 Sep 2026 |
| flywire-snn-stdp: FlyWire v783 Johnston's organ subgraph (1,239 neurons) as an STDP spiking classifier of courtship-song rhythms vs Erdos-Renyi and configuration-model wiring A FlyWire ear-circuit wiring with STDP: 0.88 accuracy, configuration model 0.72, random graph 0.43 (5 seeds); without STDP no advantage. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
The author trained a FlyWire v783 hearing-circuit subgraph (1,239 neurons) with a local learning rule (STDP) to tell apart five courtship-song rhythms, against a configuration-model graph and a random graph with the same neurons, edge count and weights. Test accuracy with learning: real 0.88, configuration model 0.724, random graph 0.428 (5 seeds each). Without learning the real wiring is not ahead (0.566 vs 0.604), and a random input mapping with learning reaches 0.920. Our ledger rules read this as mixed. Not recomputed by us. Fair, strong. | 0.77 | Mixed | Fair null Strong method |
result files commit 4ae5b8c checked 18 Oct 2026 |
| fly-connectome-study: preregistered tests of MaleCNS v1.0 mushroom-body wiring against shuffled and random sparse wiring in foraging, meta-learning, continual learning and TSP A Author's preregistered foraging test: real MaleCNS PN to KC wiring 44.0 vs shuffled 44.4 and random sparse 50.9 net reward, 200 worlds. Catalogue entry · Upstream source |
Machine-learning benchmarks closed-loop |
The author pre-registered a foraging test (200 worlds) comparing a MaleCNS v1.0 mushroom-body learner with shuffled (labelled degree-preserving) and random sparse PN-to-KC wiring. Real 44.0, shuffled 44.36, random 50.905, a linear learner 46.17, floor 0.485. Our ledger rules read this as mixed. Not recomputed by us. Re-checked 22 Oct 2026 (the author's numbers, not re-run by us): in the author's positive control, where the teacher is the real wiring with 30% of its edges rearranged, real minus shuffled wiring was +0.064 [+0.058, +0.071] and real minus random wiring +0.066 [+0.059, +0.074], while the no-brain linear readout (no Kenyon cells) was level with the real wiring at −0.004 [−0.008, +0.001]. | 1.01 No-brain baseline: 1.05 |
Mixed | Partly fair null Strong method |
result files commit 2f8f472 checked 22 Oct 2026 |
| fly-novelty-scada: a mushroom-body novelty detector with FlyWire wiring on wind-turbine fault data, against shuffled and random wiring A The detector found almost no turbine faults; real wiring scored no better than shuffled wiring (0.30 vs 0.43; interval includes zero), and random wiring scored 0.44, so the wiring nulls disagree and our rules mark the row mixed. Catalogue entry · Upstream source |
Machine-learning benchmarks offline |
The author used the real FlyWire mushroom-body wiring (149 projection neurons, 2,364 Kenyon cells) in a novelty detector for wind-turbine faults on the CARE to Compare benchmark, against degree-preserving shuffled wiring, uniform random wiring and standard detectors (3 seeds). CARE score (higher is better): FlyWire 0.30, shuffled 0.43, random 0.44, k nearest neighbours about 0.51; the FlyWire-minus-shuffled difference was −0.13 (95% interval −0.29 to +0.03). The detector found almost no faults with the alarm settings used. Our ledger rules read the headline comparison as no difference (the interval includes zero) and mark the row mixed because the wiring nulls disagree. Not re-run by us. | 1.43 No-brain baseline: 1.70 |
Mixed | Fair null Strong method |
result files commit eb03e9f checked 30 Oct 2026 |
| Reservoir computing (8) | ||||||
| fly-connectome-reservoir: frozen MaleCNS LIF network drives escape in a fly-swatter game through a hand-designed decoder; reservoir test vs target-permuted wiring B Author's reservoir test: real MaleCNS wiring 100%, target-permuted 83.9%, no connections 50% on a left/right task (3 seeds); a raw-stimulus readout also reaches 100%. Catalogue entry · Upstream source |
Reservoir computing open-loop |
The author fitted a linear readout on descending-neuron activity of a frozen MaleCNS-derived LIF network and compared real wiring with a global target permutation and with no connections, on a left/right looming task. Test accuracy: real 100%, target-permuted 83.9%, no connections 50% (3 seeds, 60 test trials each); a raw-stimulus readout also reaches 100%. Our ledger rules read this as helps. Caveats: a trained readout on a decoding task with 3 seeds; the permutation moves inputs and outputs to other neurons; the escape behaviour in the game has no wiring control. Not recomputed by us. Partly fair, strong. Brain activity vs real (display only). Activity counted over: network units. target-permuted 0.957× comparable; no connections 0.088× reduced |
0.68 No-brain baseline: 1.00 |
Helps Caveat A control without the fly wiring came close to the real result |
Partly fair null Strong method |
result files commit 2210e62 checked 17 Oct 2026 |
| The Fly Hears: spoken digits through a 30,000-neuron MaleCNS LIF reservoir with a trained reader, against scrambled wiring and audio features A Spoken-digit accuracy: real wiring 70.0%, one scrambled graph 69.7%, audio features alone 66.9%. Catalogue entry · Upstream source |
Reservoir computing offline |
The author fed spoken digits through a 30,000-neuron MaleCNS v1.0 LIF reservoir with a logistic-regression reader, against one scrambled graph and a reader on audio features alone (3,000 clips, 6 speakers). In the author's committed results: real 70.0%, scrambled 69.7%, audio alone 66.9%. Our ledger rules read this as no difference. Stat note: values from the README table backed by the committed results file, not re-run by us; one scrambled graph; with synthetic noise the scrambled graph is ahead (70.0% vs 68.2%). | 1.00 No-brain baseline: 0.96 |
No difference | Partly fair null Weak method |
result files commit 6f867b7 checked 21 Oct 2026 |
| Oruk Labs: a 499-neuron MaleCNS circuit with a trained readout labels emotion in human voices, tested against scrambled wiring (write-up only) U The authors report that a 499-neuron fly circuit with a trained readout scored 16.84% mAP on held-out voices, and the same circuit with scrambled wiring scored 16.88% (paired difference −0.04 points, 95% interval −0.16 to +0.07; 6 Sep 2026). We did not re-run it. Catalogue entry · Upstream source |
Reservoir computing offline |
The authors report that a 499-neuron fly circuit with a trained readout scored 16.84% mAP on held-out voices, and the same circuit with scrambled wiring scored 16.88% (paired difference −0.04 points, 95% interval −0.16 to +0.07; 6 Sep 2026). We did not re-run it. Write-up only, no code; the readout also receives pooled audio features directly. | 1.01 | No difference | Partly fair null Weak method |
catalogue text only checked 29 Oct 2026 |
| fly-arithmetic: a 2,000-neuron FlyWire rate network as a reservoir for XOR and digit arithmetic, vs a random network A Authors compared real wiring with a same-size random network on held-out addition: real 48, null 98 (10 null samples). Our rules read this as worse. Catalogue entry · Upstream source |
Reservoir computing offline |
The authors compared the real wiring with a random network of the same size on held-out addition accuracy (%). Real 48, null 98 (10 null samples). Our ledger rules read this as worse. The random network does not keep each neuron's degree; a linear model on the raw numbers scored 94. Partly fair, strong. | 2.04 No-brain baseline: 1.96 |
Worse | Partly fair null Strong method |
result files commit e36dd2b checked 11 Oct 2026 |
| flybrain-reservoir A Whole fly CNS reservoir remembers far less than its degree-preserving shuffle (2.2 vs 15.5); spectral-radius scaling around an antennal-lobe hot spot explains it. Catalogue entry · Upstream source |
Reservoir computing offline |
Memory capacity 2.2 real vs 15.5 for a degree-preserving shuffle (10 seeds). | 7.05 | Worse | Fair null Weak method |
README only commit 6da6325 checked 30 Sep 2026 |
| QedSDL Endpoint: FlyWire v783 mushroom-body (PN to KC) code as a novelty and value model in simulated materials-lab tests, against degree-preserving shuffled and random wiring over 20 seeds A Measured FlyWire PN to KC wiring matches its degree-preserving shuffle (AUROC 0.9505 vs 0.9505, 20 seeds) and edges out uniform random wiring (0.9391). Catalogue entry · Upstream source |
Reservoir computing offline |
The authors used the FlyWire v783 wiring from projection neurons to Kenyon cells as the fixed input layer of a novelty detector for simulated Raman spectra (a materials-lab task), against a degree-preserving shuffle and a random sparse matrix with the same synapse count. AUROC over 20 seeds: real 0.9505, shuffled 0.9505, random 0.9391 (means recomputed by us from the 20 committed per-seed files). Our ledger rules read this as mixed. Caveats: only the mushroom-body input layer is used, on simulated data; one null draw per seed. Fair, strong. | 1.00 | Mixed | Fair null Strong method |
files, recomputed by us commit a131dbd checked 19 Oct 2026 |
| RiskFly: a 128-neuron FlyWire reservoir imitating a rule-based stock and bond allocation, against degree-preserving rewired wiring A Test error (MAE) 0.00576 with the fly wiring against 0.00723 rewired, fly better in 5 of 5 seeds, author's p = 0.0625, called exploratory, while a dense network did best (0.00537); our rules read the rewired comparison as no difference and mark the row mixed because a random sparse reservoir did clearly worse. Catalogue entry · Upstream source |
Reservoir computing offline |
The author used a 128-neuron FlyWire subgraph as a fixed reservoir, with trained input and readout layers, to copy a hand-written stock and bond allocation rule, against degree-preserving rewired wiring, a random sparse reservoir and a dense network (5 seeds). Mean test error (MAE, lower is better) was 0.00576 with the fly wiring against 0.00723 rewired, fly better in 5 of 5 seeds, author's p = 0.0625, called exploratory, while a dense network did best at 0.00537 (random sparse 0.00857). Our ledger rules read the rewired comparison as no difference (p of 0.05 or more) and mark the row mixed because the random sparse reservoir did clearly worse. The target is a formula, not an investment strategy. Not re-run by us; no licence file. | none | Mixed Caveat The authors' own test did not separate the real wiring from a control (p = 0.0625) |
Fair null Strong method |
result files commit 15d0995 checked 30 Oct 2026 |
| ENSOphilia: an 80-neuron MaleCNS v1.0 visual-to-motor circuit as a reservoir for El Nino (Nino 3.4) forecasts, against 100 degree-preserving rewires A The measured 80-neuron fly circuit forecasts El Nino worse than 85 of 100 degree-preserving rewires (RMSE 0.807 vs median 0.742). Catalogue entry · Upstream source |
Reservoir computing offline |
The author used an 80-neuron MaleCNS v1.0 circuit as a reservoir to forecast the El Nino index (Nino 3.4) six months ahead, against 100 degree-preserving rewires, each with its own fitted readout. Held-out error (RMSE, lower is better): real 0.807, rewire median 0.742; a standard echo-state network 0.761 and a plain regression with no network 0.573. Not yet scored by our ledger rules. Stat note: lower RMSE is better; the author's paired 95% interval of real minus rewire median is -0.047 to +0.148, so it includes no difference; the measured circuit is worse than 85 of 100 rewires. | none | Not yet scored | Fair null Moderate method |
result files commit 4530aa9 checked 20 Oct 2026 |
| Language models (4) | ||||||
| DrosoCode A Rewired mushroom-body wiring learned to predict code tokens better than the real wiring (18.1% vs 12.6%; one run each). Catalogue entry · Upstream source |
Language models offline |
The author used the FlyWire mushroom body (2,597 Kenyon cells, projection neurons, APL and 48 output neurons) as a fixed network whose Kenyon-cell-to-output synapses learn to predict the next token of small Python programs, against a rewired copy of the same size that keeps each target's number of inputs and their weights. After 1,200 lessons next-token accuracy was 12.6% with the real wiring against 18.1% rewired (one run each; a majority-token baseline scores 9.2%). Our ledger reads this as worse, rated partly fair and weak. Brain activity vs real (display only). Activity counted over: unclear. rewired 0.964× comparable |
2.62 | Worse | Partly fair null Weak method |
result files commit 4899921 checked 31 Oct 2026 |
| Flybrain (prove-ai): seven MaleCNS v1.0 projects, from a connectome-driven garden game to a connectome language model tested against degree-rewired graphs A Real MaleCNS subgraph language model scores 2.639 bits per byte vs 2.629 for degree-rewired graphs; the paired interval includes zero. Catalogue entry · Upstream source |
Language models offline |
The author compared a 512-neuron MaleCNS v1.0 subgraph language model with degree-rewired graphs on bits per byte (lower is better). Fly 2.6388, degree-rewired 2.6286 (15 graphs), random graph 2.5827, a parameter-matched GRU 2.5575. Our ledger rules read this as mixed. The grade rests on this sub-project of the repository. Not recomputed by us. | 1.00 No-brain baseline: 1.02 |
Mixed | Fair null Strong method |
result files commit f42f586 checked 15 Oct 2026 |
| fly-distill (larval connectome as the frozen recurrent substrate of a byte-level language model) A Authors compared real wiring with a degree-preserving shuffle on bits per byte: real 3.9572, null 3.9171 (4 null samples). Our rules read this as mixed. Catalogue entry · Upstream source |
Language models offline |
The authors compared the real wiring with a degree-preserving shuffle on bits per byte (lower is better). Real 3.9572, null 3.9171 (4 null samples). Our ledger rules read this as mixed: an Erdos-Renyi random graph scored 3.5234, better than both. Fair, strong (pre-registered decision rules, 4 seeds per arm); single author. | none | Mixed | Fair null Strong method |
result files commit ad5aef6 checked 11 Oct 2026 |
| FLM - Fly Language Model A Fly-graph residual lowers LLM loss slightly, but a parameter-matched direct-input adapter does as well or better; no refit wiring null. Catalogue entry · Upstream source |
Language models offline |
Loss 1.3598 nats per token with the fly graph vs 1.3593 for a parameter-matched adapter without it; no re-fitted wiring null. | none No-brain baseline: 1.02 |
No wiring null | Partly fair null Weak method |
paper commit 7251a89 checked 30 Sep 2026 |
| Forecasting (4) | ||||||
| FlyTrader: a reinforcement-learning stock trader with a FlyWire-wired feature circuit, tested against a random-weight circuit A Excess return over buy-and-hold (0 by definition) was +280.3 points with the FlyWire circuit and +211.0 with random weights; the interval spans zero. Catalogue entry · Upstream source |
Forecasting closed-loop |
The author fed market features through a 1,024-neuron FlyWire v783 spiking circuit into a reinforcement-learning trading policy, against a random-weight circuit of the same size with the same recipe, seeds and read-out (3 seeds, 10 unseen tickers). Mean excess return over buy-and-hold (the no-brain baseline, 0 by definition) was +280.3 points with the fly wiring and +211.0 with random weights (paired difference +69.3, 95% interval -186.6 to +325.3); the 2022 and 2024 tests were similar, and a pinned rebuild gave 251.2 against 211.0, also spanning zero. Back-test only. | 0.75 No-brain baseline: 0.00 |
No difference | Partly fair null Strong method |
result files commit d59eb36 checked 24 Oct 2026 |
| fly-connectome-forecast: a FlyWire reservoir forecasting Polish wind and solar power, against random and rewired reservoirs A The fly reservoir forecast wind and solar power as well as random and rewired reservoirs, not better (mean skill 0.355 vs 0.353 rewired, 0.355 random). Catalogue entry · Upstream source |
Forecasting offline |
The author used the 1,000 most connected FlyWire v783 neurons as an echo-state reservoir to forecast Polish wind and solar power 1, 6 and 24 hours ahead, against degree-preserving rewired and weight-shuffled fly wiring, a random reservoir of the same size and density, no recurrence and gradient boosting (5 seeds; a 12-month held-out test year). Mean forecast skill over six tasks (higher is better; averaged by us from the author's table): fly 0.355, rewired 0.353, random 0.355, shuffled weights 0.345, no recurrence 0.334, gradient boosting 0.334. Our ledger rules read this as no difference. Not re-run by us. | 0.99 No-brain baseline: 0.94 |
No difference | Fair null Moderate method |
result files commit 750fb92 checked 30 Oct 2026 |
| FlyBrain Trader: a FlyWire v783 mushroom-body subgraph with reward-modulated KC to MBON plasticity as a crypto position-size multiplier, against degree-preserving rewired and Erdos-Renyi graphs A In six walk-forward windows the measured mushroom-body wiring trades no better than degree-kept rewired or random graphs (mean returns 10.2, 10.6, 11.5 %). Catalogue entry · Upstream source |
Forecasting closed-loop |
The author used a FlyWire v783 mushroom-body subgraph to size positions in a simulated crypto portfolio and compared it with a degree-preserving rewired graph and a random graph, trained the same way. Mean test-window return over 6 walk-forward windows: real 10.15%, rewired 10.58%, random graph 11.48%. Our ledger rules read this as no difference. Stat note: means and spreads computed by us from the author's 6-window table; no raw result files are committed and nothing was re-run; the author reads the gap as within noise. | 1.04 | No difference | Fair null Moderate method |
files, recomputed by us commit ad986f8 checked 20 Oct 2026 |
| BioReservoir B Fly-brain oracle answers sit at p(yes)~0.50 like the no-brain baseline; controls exist but no forecast outcomes are scored yet. Catalogue entry · Upstream source |
Forecasting offline |
Mean forecast 0.4996 real vs 0.5009 rewired and 0.50 without a brain; the questions resolve on 15 Dec 2026. | none | Not yet scored | Partly fair null Moderate method |
files, recomputed by us commit ac0253e checked 30 Sep 2026 |
| Graph analysis, no simulation (4) | ||||||
| Wired Different (ConnectomeLens) n/a Real MaleCNS topology predicts sex-related cell types better than all 500 degree-preserving rewirings, though neuropil location carries most of the signal. Catalogue entry · Upstream source |
Graph analysis, no simulation offline |
AUC-PR 0.759 real vs 0.712 over 500 rewirings (p = 0.002); brain-region location alone gives 0.722. | 0.93 No-brain baseline: 0.95 |
Helps Caveat A control without the fly wiring came close to the real result |
Fair null Strong method |
files, recomputed by us commit 03d9c5c checked 30 Sep 2026 |
| Reviving a model fly brain: anoxia, brain death and resuscitation in a FlyWire FAFB v783 LIF model, with a degree-preserving rewired null for the structural analysis A Hub-ordered removal collapses reachability after 3.5% of neurons in the real FAFB wiring, against 26.6% in a degree-preserving rewired copy; we read a higher critical fraction as a sturdier network; the author does not present it as a score. Catalogue entry · Upstream source |
Graph analysis, no simulation open-loop |
The author removed neurons of the FlyWire FAFB v783 wiring in order of connectivity and measured when sensory-to-motor reachability collapses, against one degree-preserving rewired copy. In the author's committed table the real wiring collapsed after 3.5% of neurons were removed, the rewired copy after 26.6% (random removal: 80.0% against 89.3%). We read a higher critical fraction as a sturdier network; the author does not present it as a score; on that reading our ledger rules read this as worse. Stat note: one rewired copy; a structural measure with no no-brain baseline and no chance level; the anoxia and reperfusion results behind the main claim have no wiring control. | 7.60 | Worse | Partly fair null Weak method |
result files commit 48294c0 checked 23 Oct 2026 |
| Pugliese et al. 2026 (Nature): nerve-cord connectome simulations that find a walking rhythm circuit A In the authors' saved notebook output the real front-leg nerve-cord wiring held 21,544 walking-rhythm-type three-neuron circuits, against about 307 across 100 shuffles that keep each neuron excitatory or inhibitory; a count of circuit patterns in the wiring, not a walking test with shuffled wiring (not re-run by us). Catalogue entry · Upstream source |
Graph analysis, no simulation offline |
The authors simulated rate models of the fly's ventral nerve cord, built from four wiring maps (MANC, FANC, BANC and the male CNS), and pruned them to a small walking-rhythm circuit: two excitatory neurons in a chain and one inhibitory neuron inhibiting both. The authors' code includes a test that shuffles the wiring while keeping the balance of excitatory and inhibitory connections; their saved notebook output shows the number of three-neuron circuits of the walking-rhythm type among front-leg nerve-cord neurons: 21,544 with the real wiring and 306.95 ± 29.10 (mean ± SD) across 100 shuffles. We did not re-run it (checked 30 Oct 2026). This is a count of circuit patterns in the wiring, not a walking test with shuffled wiring, so our ledger does not score it (not tested). Update, 31 Oct 2026: We checked whether the authors' walking model can be re-run with shuffled wiring on an ordinary 2-processor server without a GPU (31 Oct 2026): its software would need an estimated 620 MB once installed, plus 86 MB of wiring data, more than our 600 MB test space. The test was not run. 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. Whether the real wiring is needed for the simulated walking rhythm stays not tested. Peer-reviewed (Nature, 28 Oct 2026); a nerve-cord study, not a whole-brain or body simulation. | 0.01 | Not yet scored | Partly fair null Strong method |
result files commit 10e7661 checked 30 Oct 2026 |
| Does fly brain wiring shape signal spread? A In a sampled 1,338-neuron FlyWire circuit, real wiring activated about 137 downstream neurons vs about 107 after degree-preserving rewiring (10 of 10 pairs). Catalogue entry · Upstream source |
Graph analysis, no simulation offline |
In a sampled 1,338-neuron FlyWire circuit, real wiring spread activity further than degree-preserving rewiring: about 137 vs 107 active downstream neurons (10 of 10 paired trials). This measures how far activity spreads, not a task, so it is not scored as the wiring helping. No no-brain baseline was run. Small subnetworks, not the whole brain; authors' numbers, not re-run by us. | 0.78 | Not yet scored | Fair null Strong method |
files, recomputed by us commit b65eaa1 checked 10 Nov 2026 |
| Evolved controllers (2) | ||||||
| A Drosophila connectome substrate inside an LLM agent (Mcking1983/flybrain): MaleCNS v1.0 gate in a foraging economy, with no-brain, weight-permuted and coin-flip control arms on three seeds A Foraging economy: the weight-permuted MaleCNS graph ends with more energy than the measured graph on all 3 seeds (743 vs 360); zeroed weights starve. Catalogue entry · Upstream source |
Evolved controllers closed-loop |
The author ran a MaleCNS v1.0 network (166,700 neurons) in an evolving foraging economy of 6 flies over 6 generations and compared it with weight-permuted wiring (degree kept), zeroed connectome weights and a coin-flip decision. Final-generation mean energy (higher is better; 3 seeds): real 360.27, weight-permuted 742.93, zeroed -151.17, coin flip -110.33 (recomputed by us from the committed ledgers). Our ledger rules read this as worse: the permuted wiring ends with more energy on all 3 seeds. Caveats: the simulation kernel is not in the repository; a single-author preprint; 3 seeds. Partly fair, moderate. | 2.06 | Worse | Partly fair null Moderate method |
files, recomputed by us commit d961add checked 19 Oct 2026 |
| Null-model treatment of the sensory-motor boundary changes an evolutionary connectome comparison A Standard nulls beat the connectome after evolution, but boundary-preserving nulls tie it: the choice of null flips the verdict. Catalogue entry · Upstream source |
Evolved controllers closed-loop |
Fitness 1.57 for the connectome vs 1.78 and 1.77 for standard nulls, and 1.58 and 1.62 for nulls that keep the sensory-motor boundary. | 1.13 | Mixed | Fair null Strong method |
result files commit 2f5683d checked 30 Sep 2026 |
Which null is fair?
"Scrambled wiring" can mean many things, and the choice changes the answer. A weight shuffle keeps who connects to whom and only moves the synapse strengths, so it tests weights, not wiring. A random graph of the same size destroys hubs and structure, so almost anything beats it. A degree-preserving rewiring keeps each neuron's number of inputs and outputs and changes only its partners. A boundary-preserving null keeps every sensory-input and motor-output connection and shuffles only the wiring in between: the fairest test of the central brain, but a hard one.
| Null | Keeps | Destroys | Pitfall | Studies |
|---|---|---|---|---|
| Weight shuffle | Who connects to whom, degrees, the weight distribution | Which synapse strength sits on which connection | Tests weights, not wiring; papers often do not say whether the graph was kept. | 18 |
| Degree-preserving rewiring | Every neuron's number of inputs and outputs (often weights and signs too) | Which partners each neuron has | Rewires the sensory and motor interface too, which can open input-to-output shortcuts; one shuffle is not enough. | 76 |
| Random graph | Neuron and connection counts | Hubs, degree structure, modules, partners | Easy to beat, so a win overstates how specific the wiring is. | 25 |
| Target permutation | Each neuron's outputs, weights and signs | Where each output goes; input counts change | Activity levels shift, not only routing. | 12 |
| Sign shuffle | The graph and synapse sizes | Which synapses excite and which inhibit | A loss can come from runaway or silent activity, not lost computation. In our test it caused runaway. | 4 |
| Boundary-preserving | All sensory-input and motor-output connections | Only the central wiring | Fairest for "does the central brain matter?", but it can erase effects other nulls find. | 6 |
Five cases where the choice of null changed the result
-
Evolved controllers: the verdict flips
flyconnectome-nulls evolves controllers on a compressed FlyWire connectome. Standard nulls do better than the real wiring (1.78 and 1.77 against 1.57), because rewiring opens direct smell-to-motor shortcuts. Nulls that keep the sensory-motor boundary tie it (1.58 and 1.62). "Worse" becomes "no difference" with a fairer null.
flyconnectome-nulls (grade A), fitness after 600 generations, the author's committed results (HEAD 2f5683d). The vertical line marks the connectome. Standard nulls do better than the real wiring (95% intervals of the difference [-0.34, -0.06] and [-0.41, -0.11]); nulls that keep every sensory-input and motor-output edge tie it (intervals [-0.04, +0.05] and [-0.10, +0.05]). -
Arkanoid: weights vs wiring
In fly-plays-games, a weight shuffle keeps the graph and plays as long as the real brain (both hit the 1,500-frame cap). A random rewiring fails at 204 frames, the same as a silenced brain. The weight shuffle tested synapse strengths, not wiring.
-
A body task: give the null the same fitting
In FLY-lab, five shuffled brains turn left in 31.3% of test episodes with a fixed readout, against 100% for the real wiring. Re-fit the readout for each shuffle, as was done for the real brain, and they reach 52.7%. The gap shrinks from 69 to 47 points: still real, but a null that skips the fitting step looks worse than it is.
-
Reservoirs: compare at the same gain
The FlyBrain reservoir remembers far less than its shuffle (2.2 against 15.5), because scaling the whole matrix lets one dense hot spot dominate the real wiring. The larva study fixes the gain for every network and finds no difference (43.04% against 42.95%, p = 0.51). fly-cartpole shows the same for a learning circuit: shuffled mushroom bodies learn CartPole about as well (390 against 392 steps, p = 0.48).
-
Doom: scrambled brain vs no brain
In the Doom control study, shuffled wiring collapses survival from 53.3 s to 6.8 s (3 shuffles). But a brainless autopilot with the fly's average commands survives 51.9 s, and the best of 12 autopilot settings survives 53.4 s. More in the Doom test on Is it real?
A sixth worked example is our own test: on one reflex, four nulls give four different answers, from silent (0 Hz) through partly (0.46) to runaway (2.51). And our walking test shows the opposite trap: a null can tie the real wiring because the readout (here our capped mapping) leaves the wiring little room to matter.
A four-point checklist for a fair control
Use it on any fly brain demo or paper. Since 6 Oct one study in the ledger passes all four and shows the real wiring beating both its null and its no-brain baseline on a behaviour: our own looming test, with 6 real runs, one model and our own mapping, and with scrambled drives that take the no-brain walk by determinism rather than in new body runs. No study by other people does yet.
- Keep the degrees and signs. The scrambled brain should keep each neuron's number of inputs and outputs and whether it excites or inhibits. A random graph or a weight shuffle answers a different, easier question.
- Give the null the same chance. Same inputs and readout, same gain, and the same training or readout fitting as the real wiring. If only the real brain was tuned, a win shows little.
- Use several nulls and report the spread. At least 3 independent shuffles, with a standard deviation, interval or p-value. One shuffle is an anecdote.
- Include a no-brain baseline, and say whether the fly beat it. A scripted rule, autopilot, random policy or plain model on the same task shows whether the task needs a brain at all.
Replayed or live?
A recording of brain activity played back into a body can look like a working controller. Two studies test that directly, and in both the live closed loop is what works.
FLY-lab: brain and body
- 100%turning success with the brain live in the loop
- 0 of 30when the recorded brain output is replayed
Zero-shot steering
- 2.96poles per 10 s, live, real wiring
- 0.34 / 0.125scrambled wiring / turn neurons disconnected
No study in the ledger shows a replayed sequence matching live performance. Replay is an honest control only when it is reported next to the live run.
Brain vs no brain
24 studies compare the fly network with a controller that has no brain: a scripted rule, an autopilot, noise, or a plain model without the fly graph. The fly network beats it in only 4: our looming test (against random steering), fly tennis and FlyBrain Flappy (one match or one shuffle each, weak method), and the hash function, which loses to its own wiring nulls. On 6 Oct Fly Dino moved to the other side: a network with no connectome does as well. On 8 Oct Connectocopter joined that side: its real wiring escapes 15 of 15 looming balls, and so does a hand-written controller. On 9 Oct FlyVL joined it: a classical motion detector scores 99.8% against the connectome's 90.7%. The newest, Brain Runners (1 Oct 2026), fits the pattern: its untrained fly brain survives 82 rows against 28 on shuffled wiring, but the same rule with no brain reaches 74. Our own walking test adds one more: a brainless constant drive walks 14.3 mm against the brain's 13.9 mm. And fly-cartpole, which now beats its shuffled wiring (499.9 vs 165.9 steps), ties its own linear map (499.6) and LQR (500).
- Fly wins Build your own, Add a sense: a looming shadow turns the FlyGym fly (our run): brainless random steering turned away in 3 of 6 runs (mean 6.2°) vs the real brain's 6 of 6 (mean 49.9°).
- Fly wins Connectome ping pong (fly tennis): a parked fly that does not steer gets 14 hits vs 54 (one match each).
- Fly wins Fly Touhou: a MaleCNS v1.0 rate circuit with a 320-parameter evolved readout plays a Touhou bullet-hell, against a degree-preserving rewired circuit: standing still survives 886 frames vs the real circuit's 1,386.
- Fly wins FlyBrain · Flappy (escape reflex): no control (free fall) scores 0 vs 100.8.
- Fly wins The Fly's Hash Function: beats SimHash (0.499 vs 0.241) while losing to its wiring nulls.
- Fly wins FlyTrader: a reinforcement-learning stock trader with a FlyWire-wired feature circuit, tested against a random-weight circuit: excess return is measured against buy-and-hold (0 by definition): +280.3 points with the fly wiring, +211.0 with random weights; the interval of the difference spans zero.
- Fly wins A Drosophila connectome substrate inside an LLM agent (Mcking1983/flybrain): MaleCNS v1.0 gate in a foraging economy, with no-brain, weight-permuted and coin-flip control arms on three seeds: a coin-flip decision ends with -110.33 energy vs the fly network's 360.27 (3 seeds).
- Baseline matches or wins fly-connectome-study fly-v4 stage 7: KC code of natural odour statistics (DoOR 2.0), real MaleCNS PN to KC wiring vs shuffled and random wiring, few-shot chemical-class accuracy (pre-registered, 100 worlds): raw odour responses without the Kenyon-cell layer score 0.351 vs the real wiring's 0.349 (chance about 0.11).
- Baseline matches or wins Frozen fly connectome as a motion encoder: rewired wiring loses, a Reichardt detector wins: a classical Reichardt motion detector scores 99.8% vs the connectome's 90.7%.
- Baseline matches or wins Build your own, Add a body: FlyWire brain drives FlyGym walking (our run): a brainless constant drive walks 14.3 mm vs the brain's 13.9 mm.
- Baseline matches or wins Connectocopter (fly connectome drives a simulated drone-rover): a hand-written controller also escapes 15 of 15 looming balls (README).
- Baseline matches or wins Flight-test the fly: a classical yaw damper 1.69 deg/s vs the fly circuit's 3.19 (lower is better).
- Baseline matches or wins FLY-lab: What a fly connectome adds to controlling a body: a two-line rule also turns correctly in 100% of episodes.
- Baseline matches or wins FlyArm: a GRU with the same trained interface 56.9% vs 72.2% lift; not significant over 6 seeds (p = 0.125), so not counted as beaten.
- Baseline matches or wins SomaHouse: a trained fly-derived circuit runs a simulated house, tested against degree-preserving rewired, dense and no-hub circuits: a conventional balance controller with no brain costs 0.1172 vs the fly wiring's 0.1589 (lower is better); dense wiring costs 0.1538.
- Baseline matches or wins Brain Runners: the same rule with no brain reaches 74.05 rows vs 82.34.
- Baseline matches or wins brainlab (FlyWire whole-brain LIF run with real, shuffled and disconnected wiring): a fine-tuned 2B language-model decider scores 100% (in-file) vs the scorer's 92.91%.
- Baseline matches or wins Is the fly brain actually playing DOOM? (control experiments): matched autopilot 51.9 s vs 53.3 s; best autopilot setting 53.4 s.
- Baseline matches or wins doomfly-rl: a model without the connectome: 7.39 vs 6.57.
- Baseline matches or wins Drosophila Cadet: the MaleCNS v1.0 looming-escape circuit plays 3D Pinball, with six degree-preserving shuffled circuits: a no-brain hand-written reflex scores 11.480 vs the real circuit's 11.221.
- Baseline matches or wins Fly Dino (flyjump): a 243-parameter network with no connectome survives 180.0 s vs the circuit's 179.4 s (corrected 6 Oct; no wiring null).
- Baseline matches or wins fly-plays-games (Pokémon Red chapter; formerly fly-plays-pokemon): a scripted tracker 1,372 vs 1,500 frames, within the spread of 4 runs.
- Baseline matches or wins Fly-Racer: a plain MLP scores 894.0 vs 903.2 (one seed each).
- Baseline matches or wins Fly Self Driving: a small MLP 90% vs 97% (one seed).
- Baseline matches or wins Fly Tic-Tac-Toe: a 98-cell MaleCNS v1.0 circuit with a trained 425-parameter readout picks tic-tac-toe moves, against a degree-preserving rewired graph at equal training budget: a plain board network without the circuit scores 0.905 vs the circuit's 0.854 against a random player.
- Baseline matches or wins Fly Worker: smoothed noise covers 63% of the map vs 50% (author, 10 Nov).
- Baseline matches or wins FlyAim: a uniform random walker gets closer to the target (251.7 px vs 412.5 px).
- Baseline matches or wins flybrain-intransitive: the untrained MaleCNS v1.0 LIF connectome scores rock-paper-scissors-chess moves through looming and small-object inputs, against scrambled wiring: a greedy rule with no brain wins 55 of 100 games, the same as the fly network's 55.
- Baseline matches or wins FlySnake: MaleCNS v1.0 LIF network plays Snake through a trained descending-neuron readout; held-out test vs rewired graph: a hand-written Snake rule scores 21.52 vs the fly network's 20.03 food per game.
- Baseline matches or wins Fruit Fly Plays Fruit Ninja: a 4,386-neuron MaleCNS circuit with a fitted readout, against a rewired circuit and a pixel heuristic: a no-brain pixel rule scores 24.95 vs the real circuit's 24.85.
- Baseline matches or wins fruitfly-lab (Fly Marksman): a 3,963-neuron FlyWire v783 LC10a-to-steering LIF circuit aims and fires in a browser game and in ViZDoom, against degree-preserving scrambled wiring: a no-brain oracle that reads the enemy direction gets 11.5 kills vs the fly circuit's 10.2.
- Baseline matches or wins Haltere: a PD controller finished 1 of 2 races vs 0 of 2.
- Baseline matches or wins making-fly-play-chess: loses to a material-count player (0.694).
- Baseline matches or wins MaleCNS_training: a trainable recurrent network on a MaleCNS v1.0 subgraph compared with a rewired graph, an RNN and a GRU on 12x12 gomoku and memory tasks: an RNN 0.788 and a GRU 0.790 vs the fly graph 0.787 held-out accuracy.
- Baseline matches or wins Boltzmann-fly: a Deep Boltzmann Machine world model with couplings masked to the MaleCNS v1.0 mushroom-body wiring, against rewired masks: a plain MLP reaches AUC 0.713 vs the fly-masked model's 0.712.
- Baseline matches or wins fly-cartpole: the circuit's own linear map 499.6 and LQR 500 vs the circuit's 499.9 steps.
- Baseline matches or wins fly-connectome-study: preregistered tests of MaleCNS v1.0 mushroom-body wiring against shuffled and random sparse wiring in foraging, meta-learning, continual learning and TSP: a linear learner scores 46.17 vs the real wiring 44.0.
- Baseline matches or wins fly-novelty-scada: a mushroom-body novelty detector with FlyWire wiring on wind-turbine fault data, against shuffled and random wiring: a k-nearest-neighbours detector scores 0.51 vs the fly detector's 0.30 CARE score.
- Baseline matches or wins MaleCNS Cube: trainable network on a MaleCNS v1.0 subgraph estimates Rubik's cube distance for a budgeted search, against a degree-preserving rewired graph, a frozen graph and an MLP: a parameter-matched MLP scores 0.3245 vs the fly graph's 0.3509 test error (lower is better).
- Baseline matches or wins malecns-keratitis-pilot: a frozen MaleCNS v1.0 visual circuit as image features for keratitis classification, against rewired circuits: does not beat an image model without the circuit (AUROC 0.479 vs 0.754) and ties the raw input (0.464).
- Baseline matches or wins ENSOphilia: an 80-neuron MaleCNS v1.0 visual-to-motor circuit as a reservoir for El Nino (Nino 3.4) forecasts, against 100 degree-preserving rewires: a plain regression with no network forecasts better (RMSE 0.573 vs the circuit's 0.807; lower is better).
- Baseline matches or wins fly-arithmetic: a 2,000-neuron FlyWire rate network as a reservoir for XOR and digit arithmetic, vs a random network: a linear model on the raw numbers scores 94% vs the network's 48% on held-out addition.
- Baseline matches or wins fly-connectome-reservoir: frozen MaleCNS LIF network drives escape in a fly-swatter game through a hand-designed decoder; reservoir test vs target-permuted wiring: a linear readout on the raw stimulus history also reaches 100%, the same as the real wiring.
- Baseline matches or wins The Fly Hears: spoken digits through a 30,000-neuron MaleCNS LIF reservoir with a trained reader, against scrambled wiring and audio features: audio features alone reach 66.9% vs the real reservoir's 70.0%.
- Baseline matches or wins RiskFly: a 128-neuron FlyWire reservoir imitating a rule-based stock and bond allocation, against degree-preserving rewired wiring: a dense network with no fly graph reaches 0.00537 test error vs the fly wiring's 0.00576 (lower is better).
- Baseline matches or wins FLM - Fly Language Model: a parameter-matched direct-input adapter does as well (1.3593 vs 1.3598).
- Baseline matches or wins Flybrain (prove-ai): seven MaleCNS v1.0 projects, from a connectome-driven garden game to a connectome language model tested against degree-rewired graphs: a parameter-matched GRU 2.5575 vs the fly graph 2.6388 bits per byte (lower is better).
- Baseline matches or wins fly-connectome-forecast: a FlyWire reservoir forecasting Polish wind and solar power, against random and rewired reservoirs: gradient boosting on lagged features reaches 0.334 mean skill vs the fly reservoir's 0.355 (higher is better).
- Baseline matches or wins Wired Different (ConnectomeLens): brain-region location alone gives 0.722 vs 0.759.
The clearest case: Doom
In the Doom control study, a brainless autopilot that turns, walks and shoots at the fly brain's average rates survives 51.93 s against the fly brain's 53.34 s, and the best of 12 autopilot settings survives 53.44 s. Scrambled wiring collapses the fly brain to 6.75 s, so this brain needs its wiring, but the task does not need a brain.
All 19 rows of the source file
| Arm | Key in the file | Seconds survived (mean) | 95% interval | Episodes | Kills (mean) |
|---|---|---|---|---|---|
| DOOMFLY, real wiring | real | 53.34 | 49.73–56.78 | 20 | 12.1 |
| Matched spray autopilot (no brain) | spray_matched | 51.93 | 45.32–57.28 | 20 | 11.75 |
| Spray autopilot setting t0_f0 | spray_t0_f0 | 6.11 | 5.63–6.67 | 16 | 1.625 |
| Spray autopilot setting t0_f10 | spray_t0_f10 | 8.59 | 7.65–9.57 | 16 | 1.625 |
| Spray autopilot setting t0_f20 | spray_t0_f20 | 9.63 | 8.47–10.89 | 16 | 1.875 |
| Spray autopilot setting t1_f0 | spray_t1_f0 | 5.68 | 5.35–6 | 16 | 0.9375 |
| Spray autopilot setting t1_f10 | spray_t1_f10 | 26.7 | 21.91–31.43 | 16 | 6.6875 |
| Spray autopilot setting t1_f20 | spray_t1_f20 | 53.44 | 48.81–57.31 | 16 | 13.25 |
| Spray autopilot setting t2_f0 | spray_t2_f0 | 5.46 | 5.18–5.76 | 16 | 0.25 |
| Spray autopilot setting t2_f10 | spray_t2_f10 | 9.65 | 8.29–10.99 | 16 | 2.625 |
| Spray autopilot setting t2_f20 | spray_t2_f20 | 21.09 | 16.55–26.08 | 16 | 5.0625 |
| Spray autopilot setting t4_f0 | spray_t4_f0 | 5.28 | 5.06–5.53 | 16 | 0.25 |
| Spray autopilot setting t4_f10 | spray_t4_f10 | 7.6 | 6.73–8.52 | 16 | 1.5 |
| Spray autopilot setting t4_f20 | spray_t4_f20 | 11.41 | 9.39–13.59 | 16 | 3.125 |
| Shuffled wiring, seed 0 | shuf0 | 5.48 | 4.99–5.97 | 4 | 0.25 |
| Shuffled wiring, seed 1 | shuf1 | 9.29 | 7.1–11.47 | 4 | 1.5 |
| Shuffled wiring, seed 2 | shuf2 | 5.48 | 4.99–5.97 | 4 | 0.25 |
| No connections | noconn | 5.5 | 5.24–5.76 | 8 | 0.375 |
| No vision (blind) | blind | 10.46 | 8.84–12.83 | 8 | 2.5 |
What would change the answer
- A study by other people that passes all four checks. A behavioural task with 3 or more degree-preserving shuffles, tuned the same way, and a no-brain baseline, in which the real wiring beats both. Only our own looming test does so far, with few runs and one model. An independent one would turn "sometimes" into a clear yes for that task.
- Boundary-preserving nulls on the reflex studies. If they erase the reflex effects, as they erased the null advantage in the evolution study, "helps" would shrink to "the input and output wiring matters".
- Independent replications. Every study except ours is a single author's, and 41 of 43 are not peer-reviewed.
- Finishing our own reflex test. Our reflex test still leaves the excitability question formally open: ×2.5 matched one shuffle but not three. A per-shuffle gain search, with one gain for each shuffle inside the band and written down before it runs, would settle whether the scrambled brains lose only because they are quieter (they stayed at 0 Hz at every gain so far). A selectivity measure next to the MN9 rate would tell a reflex from a runaway, but needs a new pre-registration.
- A body test where the wiring could matter. Our walking test capped the drive, so speed could hardly differ. A pre-registered mapping without the cap, P9 at lower rates, longer walks and a coupled run in which the body feeds back to the brain (not yet tested) would give the wiring room to matter.
- Finishing our looming test. Our looming test now has 6 real body runs, and the pending drives walked on 8 Oct. Open: which longer paths carry the turn (silencing PVLP141 or all two-step relays left the turn rule passing), a rate-based measure next to the saturating ratio, a measured real-fly input rate instead of a rate we set, and a coupled run in which the fly's movement changes what the eye sees (not yet tested).
- Pending results. Open Fly has pre-registered a shuffled control with no result yet, fly-walking-wiring compares real fly leg networks with six rewired families under coded rules but has not published its results, and bioreservoir's forecast questions resolve on 15 Dec 2026.
Method and limits
How the ledger was built
- One study per catalogue entry that compares its fly wiring with a wiring null or a no-brain baseline, plus our own three Build your own tests. Read on 30 Sep 2026 (our reflex test and Brain Runners: 1 Oct; our body test: 2 Oct; ChessFly, fly-cartpole at commit a8c6257 and flyvis: 3 Oct; our looming test, Flight-test the fly, FlyAim and FlyArm: 5 Oct; fly tennis, Fly-Racer, FlyBrain Flappy and Fly Dino's correction: 6 Oct; Connectocopter, Synaptera and our looming row's new runs: 8 Oct; one study: 29 Sep) at the commits listed in the data file.
- Numbers come from committed files (or, for our own tests, our run files) for 39 studies, papers for 3 (Shiu et al., FLM, flyvis) and catalogue text for 1 (NeuroCraft Fly, which reports no numbers).
- README-only numbers. The data file has no separate value for numbers that come only from a README. The convention:
extraction: "files"together withquality.results_in_repo_files: falsemeans the numbers come only from the README (or a docs page), not from committed result files. 9 studies are marked that way (the table says "README only" unless we recomputed the numbers); Brain Runners' README table is read as well, although its result files are committed. - "Helps" or "worse" needs a gap larger than twice the combined standard error. Spreads "across tasks" do not count as uncertainty. One label was set by hand: Fly.exe is "mixed", because its selectivity index can be negative.
- Fairness (fair 15, partly fair 18, unclear 5, unfair 5) and quality (strong 13, moderate 6, weak 24) follow fixed yes/no rules: degrees kept, input-output boundary kept, tuned equally, number of null samples, no-brain baseline, held-out test, reported uncertainty, results in files. The rules did not change on 2, 3, 5, 6 or 8 Oct.
- Since 2 Oct 2026 every arm also carries
activity_ratioandactivity_class(whole-brain activity of the null against the real wiring: reduced, comparable, elevated, runaway or not reported). They are display only and never enter a score; only our three tests report them so far.
Limits
- We re-ran none of the other studies; our own three tests are the only ones we ran. The Shiu paper result is "weak" only because the paper text does not give the shuffle count or spread.
- The flyvis ablation values sit in a figure and were not extracted, so flyvis has no retention.
- Retention compares very different measurements on one scale; it shows direction and size, not significance.
- The ledger covers our catalogue only. Studies we have not found or admitted are missing, and new ones are added as the catalogue grows.
Data: the full ledger, with every arm, source file and rule output, is published as controls.json with its JSON Schema; field notes are on For AI agents. Catalogue records carry the matching controls and wiring_effect fields. The method and the sources are in the research report of 30 Sep 2026; our own tests are in the report of 1 Oct 2026 (reflex) and the report of 2 Oct 2026 (body, gain check).