Research report · 9 November 2026 · thirtieth check
Thirtieth check: the walking-circuit model re-run with shuffled wiring
Pugliese and colleagues (Nature, 28 Oct 2026) report a small walking-rhythm circuit found by simulating the fly's nerve cord. Until now our ledger could not say whether the real wiring is needed for the simulated rhythm: the authors' saved shuffle result counts circuit patterns in the wiring, and on 31 Oct their model did not fit our test space. On 9 Nov we ran the authors' own model once with the real wiring and once with each of 5 wirings shuffled by the authors' own shuffle, and read the rhythm with the authors' own score. The rule for reading the result was written down before the first install step. We also scanned for new claims from 3 to 9 Nov and re-checked links, videos and tracked repositories.
Short answer
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. It is a separate row in our controls ledger, marked as our own check. It does not test the paper's claims about real flies, and with the activity mismatch it does not settle whether the real wiring is needed for the simulated rhythm.
One simulation took 1.2 minutes and its process peaked at 1.4 GB of memory; the software and wiring data used about 1,130 MB of disk.
What we ran, and why
- The question: does the authors' walking model still produce a leg rhythm when its wiring is shuffled? Their saved notebook shows 21,544 walking-rhythm-type three-neuron circuits in the real front-leg wiring against about 307 in shuffles, but that is a count of patterns in the wiring, not a simulation, so our ledger did not score it.
- The model: the authors' public code (smpuglie/Pugliese_2026, commit 10e7661) and its front-leg nerve-cord wiring data, run on an ordinary 2-processor server without a GPU. We used only the authors' own functions for the simulation, the shuffle and the readouts.
- The stimulus: the authors' standard activation of the descending neuron DNg100 for 2 seconds, with the design values in the authors' configuration files (we re-read them before the runs; they matched what we wrote down).
- The readout: the number of motor neurons whose activity has a regular rhythm by the code's own score (0.5 or more), and, for context, how many neurons were active at all.
- The shuffle: 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 rule, fixed before the run: if at least half of the shuffles still show a rhythm (at least one motor neuron at 0.5 or more), our row reads no difference.
Real wiring against 5 shuffles
| Network | Motor neurons with a regular rhythm | Active motor neurons | Active neurons | Top rate (per second) | Run time |
|---|---|---|---|---|---|
| Real wiring | 3 | 3 | 97 | 16 | 70.1 s |
| Shuffle 1 | 0 | 84 | 1,995 | 231 | 102.3 s |
| Shuffle 2 | 12 | 85 | 1,671 | 231 | 184.5 s |
| Shuffle 3 | 40 | 72 | 1,925 | 226 | 132.2 s |
| Shuffle 4 | 3 | 73 | 1,656 | 227 | 96.2 s |
| Shuffle 5 | 1 | 85 | 1,645 | 231 | 94.2 s |
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 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.
Motor neurons with a regular rhythm
- 3 Real wiring
- 0 Shuffle 1
- 12 Shuffle 2
- 40 Shuffle 3
- 3 Shuffle 4
- 1 Shuffle 5
Active neurons in the network
- 97 Real wiring
- 1,995 Shuffle 1
- 1,671 Shuffle 2
- 1,925 Shuffle 3
- 1,656 Shuffle 4
- 1,645 Shuffle 5
Why the activity matters: in the shuffled circuits 72 to 85 motor neurons were active, against 3 with the real wiring, and many neurons ran near the top rate. A rhythm score in such a busy, near-saturated network is not the same thing as the sparse rhythm of the real wiring. The code can adjust the stimulus to keep activity in a set range, but that option is off by default and we left it off, as we had written down before the run.
What it took to run
One simulation took 1.2 minutes and its process peaked at 1.4 GB of memory; the software and wiring data used about 1,130 MB of disk. The shuffled networks took 1.6 to 3.1 minutes each. The memory figure is the largest simulation process (decimal GB); the whole test container used more. On 31 Oct we had estimated about 706 MB of disk for the software and data; the measured figure was about 60% higher.
Limits
- One run for the real wiring and one for each of 5 shuffles; no statistics.
- No noise, and faster solver settings than the paper (tolerances 1e-4 and 1e-7, against the paper's 2e-6 and 5e-9).
- Not activity-matched: 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.
- Solver status and traces were not saved, so we cannot check whether any shuffled run hit the solver's step limit.
- One helper package version had to be changed for the install to work, before any simulation ran.
- "Front-leg" rests on the wiring data the configuration loads (the T1 segment, which is the front-leg segment) and on the leg motor neurons in the readout; we did not re-read the paper on 9 Nov.
- This is a simulation of a nerve-cord model, not of real flies; it does not test the paper's experiments in behaving flies.
Lead scan, 3 to 9 Nov
9 Nov: we scanned news, videos and new code from 3 to 9 Nov; one widely seen new video, an Urdu and Hindi explainer whose title says a fly's brain was uploaded, got a dated check (its own description says no mind was copied). Two new write-ups about fly-brain games and puzzles and several small new code projects that say they test whether the fly's wiring matters are queued for a first look.
The video now has a dated line in our verdict row on "uploaded" fly brains: A new Urdu and Hindi video, 'Scientists Uploaded a Fly's Brain Into a Computer Are Humans Next?' (Takhti, 3 Nov 2026; about 202,000 views on 9 Nov), uses the word uploaded in its title, while its own description says this does not mean a fly's mind was copied and calls Eon Systems' demos company claims with limits. That matches the primary sources: a wiring map traced from one preserved fly brain, run as a simplified model. We read the title and description, not the video (checked 9 Nov 2026).
Tracking
- Links: as on 2 Nov (fruitfly.live still fails its security certificate; the ascend demo and the FlyGym browser viewer still answer 404).
- Videos: 57 of 57 available.
- Repositories: 9 tracked repositories have new commits since 2 Nov; they are recorded and not yet re-read. The repositories of 13 of the 25 verdict rows were re-checked on 9 Nov.
What comes next
- A shuffle test with activity matched between the real and the shuffled networks would show whether the shuffled rhythms persist in a like-for-like comparison. It needs its own plan, written before the run, and is not yet scheduled.
- The next beginner re-test is due 16 Nov 2026.
Sources: Pugliese et al., Nature, 28 Oct 2026; preprint; code (commit 10e7661). Ledger row: pugliese-2026-shuffle-check; the paper's own entry: pugliese-2026-walking-cpg (not tested). Earlier check: report of 31 Oct 2026.