Research report · 22 October 2026 · twenty-third check
Twenty-third check: the four-week re-check of our grades, and two new ledger rows
We re-checked every project we last graded on 28 Sep against its current code, re-read three carried projects at their new commits, graded two new projects with wiring controls, and watched the neural-canvas demo for about three minutes after pressing its start button. Nothing was installed or simulated by us in this check.
Short answer
22 Oct: we re-checked the 34 projects we last graded on 28 Sep against their current code: 33 keep their grade (3 of them after a full re-read of changed code), and 1 could not be re-checked because the source is no longer online.
Fly wiring beat scrambled wiring in 31 of 74 studies. Of the two new rows, one helps without a no-brain baseline, and in the other random decisions scored above both wirings on our reading of its metric.
neural-canvas downloaded about 79 MB after a long preparation, but its status line read "0 spikes", so we record its brain simulation as not seen. Numbers in the ledger are the authors' own and were not re-run by us.
The four-week re-check: 34 projects graded on 28 Sep
Every record graded A, B or C on 28 Sep 2026 in the 21 Oct catalogue (15 A, 7 B, 12 C), checked against the commit it was graded at and today's code
| Outcome | Projects | What we found |
|---|---|---|
| Kept, code unchanged (25) | bioreservoir, doom-fly-control, doomfly-rl, dotfly, drosophila-brain-mlx, embodied-fly-brain-erojas, flm, fly-api, fly-connectome-adds-to-body, fly-exe, fly-self-driving, flyarena-banc, flybrain-app-snedea, flycraft, flydrones, flys-hash-function, fly-worker, infinite-sugar, larva-vs-shuffles, pianist-fly, same-smell, shiu-drosophila-brain-model, vial, webgpu-fly, neural-canvas | Every cited file is still there; neural-canvas is unchanged at its graded commit. |
| Kept, code moved without a material change (5) | fly-brain-lulzx: new nerve-cord experiments; no cited file changed; flybrain-reservoir: a DOI badge; flyconnectome-nulls: affiliation changes and new research-log entries; the cited content is unchanged (a judgement call); haltere: a README row the catalogue already reports; help-the-fly-escape: tooling files only | No cited result changed. |
| Re-read in full, grade kept (3) | flydoom-mutkuoz A: new feature tests; the shuffle and random-agent controls are unchanged; fly-dino C: new controllers without the connectome completed 100 of 100 courses against the connectome's 99 of 100; still no rewired-circuit comparison; flyvis A: the analysis matches cells by type, and bug fixes leave the default results unchanged | Their ledger rows were rebuilt with unchanged values. |
| Not re-checkable (1) | flydoom | The repository asks for credentials (not found since 1 Oct 2026); grade C and its date are unchanged. |
Carried re-checks
- 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.
- fly-with-me stays B: the brain still runs the Shiu et al. model on FlyWire v783 with hand-made ear bands and dance moves, and there is no wiring comparison; the author reports a trained brain at 0.498 macro-F1 against 0.485 for the ear input alone (difference +0.013 ± 0.028; the results file is not in the repository).
- FlyGateDemo stays A: we read the code behind its loader, its wiring comparison and its no-brain cosine baseline; its numbers are reported in the README, printed by the author's program, and were not re-run by us.
- adfly-connectome: links now point at the current commit (the earlier commit is gone; code and results unchanged), and the claim cites the authors' arXiv preprint (arXiv:2610.19153, 13 Oct 2026); grade B unchanged.
2 new entries and 2 new ledger rows
- Helps matcha-fly A, MaleCNS v1.0 LIF model, pre-registered: 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.
- No difference maxwells-demon-fly-brain A, FlyWire v783: 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. In a 2-seed run the ratios were 0.893, 0.391 and 1.091.
Three more repositories are carried to the next check, six were excluded as not fly-brain projects or not projects, and one could not be reached. Two new videos were not admitted to the gallery: one looks like a re-upload of a project already listed, and the other names no project.
neural-canvas, watched for about three minutes
One load · headless Chromium, Linux, no GPU · procedure written down before the load · screenshots and network log only
22 Oct: in a headless browser without a GPU, after we pressed 'Download & start', the page prepared its scene for about 73 seconds, then started a background worker, which downloaded about 79 MB within about 94 seconds of the click; the picture changed between our later screenshots, and at the end of our about 187-second watch the page's status line read 'JavaScript · 0.02× realtime · 0 spikes'. Screenshots cannot show whether the brain drives the movement.
- 0 s: Page opened
- 30 s: We pressed 'Download & start'
- 38-81 s: The same loading screen in our screenshots
- about 103 s: A background worker started
- 105-125 s: The worker downloaded 29 files, about 79 MB (page total about 86 MB)
- 127-217 s: The picture changed between our later screenshots
- end of watch: Status line: 'JavaScript · 0.02× realtime · 0 spikes'
Catalogue status: loads; brain simulation not seen in our headless browser without a GPU (checked 2026-10-22)
. The files come from a content delivery network under content-hash names, so the log alone does not tell us which download is which; we do not label them as connectome data.
Correction. On 6 and 20 Oct our check pressed 'Download & start'; in the about 30 seconds we watched afterwards, the page fetched about 6 MB and started no background worker, and our screenshots show only the start card and a loading screen. Our 21 Oct note said our check had not pressed the button.
Browser count: in our checks of 3, 6, 11, 17, 18, 20 and 22 Oct 2026, 9 of 13 demos ran by our check's rule at their latest check; neural-canvas is one of the 9 by that rule, without a seen brain simulation.
Tracking
- Repositories of 141 catalogue entries checked: 134 unchanged, 6 changed and 1 gone; two moved projects (fly-chess and flybench) get a material-change check next time.
- Link rows: 288 project links (261 answered, 10 redirected, 15 blocked our automated check, 2 errors) and 160 source links.
- 14 gallery videos have view counts read on 22 Oct 2026 and 3 on 21 Oct 2026 (approximate, from YouTube search results pages, linked on each card). The other view counts keep their earlier dates.
- Odyssey-3 claim: No method, code or write-up was published as of 22 Oct 2026.
- The 15 verdict rows with a repository were re-checked; none changed.
What we could not resolve
- neural-canvas: screenshots cannot show whether the brain drives the movement, and the "0 spikes" status may mean that no neurons were painted rather than an idle brain (a hypothesis).
- Maxwell's Demon: the direction of its ratio is our reading; the author's files do not state it.
- flyconnectome-nulls: its research log, a cited file, gained new entries; we judged the cited content unchanged.
- All ledger numbers in this report are the authors' own; we did not re-run them.