RESEARCH NOTEBOOK

Journal

Notes on cat neural models, the evidence behind them, and the experiments we run.

All entries29 notes · 2026
  1. Three steps did not beat one

    Note 28 found that the two-signal rule sometimes holds when one signal leads by only 7.5 points. This note tests one remedy, declared beforehand: show the same pair for three steps in a row and decide from the mean of the three readings. Three new network instances were each built and run at least twice. Declared outcome: PASS, but at the minimum, with the intended action in 75 of 78 scored blocks where at least 75 were needed. The remedy itself did not help: in the 36 close blocks the mean of three steps decided for the stronger signal in 33, and the first step alone did so in 33 as well. Readings did not fade over the three repeats. All three misses were holds on the same side, where signal B led. A reading made after the result, and not declared, suggests a reason: the grey screen's reading lies nearer the B side, so one threshold leaves B less room. Equal signals held in 24 of 24 blocks and clear leads were acted on in 18 of 18. One of seven runs disagreed with its partner and was replaced as declared. Simulation only; no market signal and no action.

  2. A decision map on fresh instances

    Note 27 tested a rule that turns two simultaneous signals into one of three actions (A, B or hold), but on held runs and on three strengths only. This note repeats the test on runs that did not exist when it was declared, and on a finer grid. Three new network instances were each built and run twice; the rule and its threshold from note 27 were used unchanged on a 5 by 5 grid of strengths. Declared outcome: PASS, the intended action in 168 of 168 scored steps. The cells where the two strengths differ by only 7.5 points were declared a transition zone and described, not scored: there the rule took the stronger signal's action in 66 of 72 steps and held in 6. A relay that adds the single responses would have erred in 3 of 30 cases. Six of the 14 scored cells repeat conditions note 27 had already tested, so a PASS there was likely; the other eight use strengths the model had not been shown. Three instances, four steps per cell, no runs without the cortex's own connections, and no market signal or action. Simulation only.

  3. A rule for two signals at once

    Notes 22 and 25 showed that the cat circuit model does not add two superimposed gratings: the stronger one dominates. This note turns that into a decision rule for a controller that receives two signals at once, and tests it. Two signals are shown as two gratings at right angles; one number read from the circuit's response says which of three actions follows: A, B or hold. Every number of the rule was fixed on the pilot's one network instance and declared; the rule was then applied unchanged to the six instances of the confirmatory test. Declared outcome: PASS, the intended action in 336 of 336 test steps. A relay that adds the single responses would have erred in 5 of 24 cases. The test reuses held runs whose summary results were known and whose per-step readouts were already published, which is weaker than a fresh run; the rule also works on the pilot with the cortex's own connections removed, and no market signal or action is involved. Simulation only.

  4. What the suppressed cells received: the declared conductance follow-up

    Protocol 022 declared a follow-up before its pilot ran: rebuild, offline, the synaptic conductances each layer 4 cell received, after proving the rebuild on earlier runs. The replay passed all three declared validations on its first recorded attempt, reproducing more than 99.99 % of spike counts. The readings, exploratory by declaration and on one network instance: when a second, orthogonal grating is added, inhibition onto cells tuned to the first grating falls by about 15 % of what the first grating evoked, the opposite of the direction asked about, and excitation from other cortical cells falls by about 39 %. The declared reading of the input stage could not answer its question: the mean input from the LGN turned out not to depend on orientation. One declared count did not fit the declared stimulus set and was amended before any of the pilot's recorded spikes were read. The conductances are rebuilt from recorded spikes, so these readings describe the suppressed state and do not show its cause. Simulation only.

  5. The plaids result holds over six fresh instances: confirmed

    Protocol 023 revision 2 reran the confirmatory plaids test with a paired rule: every network instance is run twice on identical inputs and analysed only if the two runs agree on every spike train. Six fresh instances of the cat circuit model settled with an agreeing pair, after one slot fell to the reserve and the rule absorbed two one-off faults of the kind that had ended the first attempt. The declared outcome is CONFIRMED on all three measures: on average over the instances, the model circuit did not answer two superimposed gratings with the sum of its answers to each alone, in the pilot's direction and beyond the declared margin. As declared, this describes the whole model circuit on its one orientation map, input stage and cortex together, in simulation; it does not show that the cortex's own connections compute it, and it is not biological validation.

  6. The confirmatory test stopped at its first run: not interpretable

    Protocol 023 was to test the plaids pilot on six fresh network instances of the cat circuit model. Its first run finished, but a declared integrity check found one of 3,600 input cells firing differently from the reference, and the declared early stop ended the protocol. The declared outcome is NOT INTERPRETABLE, and no confirmatory reading was made. An exploratory diagnosis afterwards found that exact repeats of the run's inputs did not reproduce the difference, which points to an execution fault on the laptop; its cause is untested and its rate close to unknown. The plaids question over instances is still open.

  7. The model moved to a laptop: all 18,600 spike trains matched

    The research server was lost, so the 2 mm model moved to an 8 GB laptop, under a memory cap of a little under half the machine. Small fixes to its random numbers and arithmetic came first. The plaids pilot's intact run, repeated on the laptop, then reproduced all 18,600 neurons' spike trains over 35 seconds of model time, and the frozen analysis gave the same three measures. This is numerical agreement on one network instance. It is not new biology, and it is not a confirmation of the pilot.

  8. Two gratings at once: the circuit did not simply add them

    Protocol 022's plaids pilot has run. Shown two drifting gratings on top of each other, one network instance of the cat circuit did not answer with the sum of its answers to each grating alone: a grating at right angles suppressed the response to the preferred one, the stronger of two gratings dominated, and the decision readout, applied offline, leaned further toward the stronger grating than the sum of its single readouts. The declared outcome is GO. As declared before the runs, this describes the whole model circuit, input stage included, and does not show that the cortex's own connections compute it.

  9. A faster test that stopped at its own timing check

    Note 20 found the cat circuit to be a faithful relay when its inputs are spaced apart. The next test removed the gaps, so recent inputs overlap inside the circuit, and asked whether the cortex's own connections change how they mix. Its rules, fixed in advance, included two stop checks. Both fired at 70 ms steps. The declared fallback, 105 ms steps, failed the timing check again, so the design ended before any main run. Its question is still open.

  10. The circuit's first decisions: a faithful relay

    The decision test frozen in note 18 has run. At 210 ms gratings separated by grey gaps, the cat circuit passes each new input on almost perfectly and carries no material trace of the previous one into its next decision, in both network instances. Without its cortical connections, the decoded signal is compressed toward the middle and decisions lean slightly away from the previous input, so the cortex shapes the readout even at this timing. As declared, this design stops here; the next one was pre-registered before this result and is already running.

  11. Second test against cat recordings: shared fluctuations in firing

    We compared how strongly groups of neurons in the full model fluctuate together in their spontaneous firing with open multiunit recordings from cat area 17, under a protocol frozen before any of these measures was computed. Within each second, the model's groups share 8 to 64 times less fluctuation than the cats' recording sites, even in the cats' least synchronized states. In a secondary measure, the gap is largest on slow time scales. The cats were anaesthetized, so this describes a difference; it is not a pass or fail.

  12. Can the cat circuit make decisions? Calibrating and freezing the test

    We built Felisyn's first decision pipeline around the cat visual-cortex model, entirely in simulation. A synthetic market signal becomes a drifting grating, the model responds, and a fixed decoder turns its spikes into buy, sell or hold requests for a simulated executor. A calibration run showed that the model passes on a grating's orientation almost exactly. Before the real test runs, we froze how we will decide whether the circuit's own dynamics change its decisions, and what happens if they do not.

  13. First test against independent cat recordings

    We compared the full model's spontaneous membrane potential with intracellular recordings from cat visual cortex that the model was not built on, under a protocol frozen before any of these statistics existed. The model's membrane potential decorrelates about eight times faster than in the recordings, and nearby neurons are less synchronized overall. Above 20 Hz their synchrony is comparable. The recordings come from anaesthetized cats, so this describes a difference; it is not a pass or fail.

  14. Two ways to average: the 3.56 nS question

    The paper's text gives an average inhibitory conductance of 3.56 nS, while its Figure 4 code implies about 4.09 nS. Using our full run, both averages in the text are reproduced when every population counts about equally, and neither when the populations are weighted by size. The discrepancy most likely comes from how the averages were taken. This is an inference from one run; the code behind the text is not public.

  15. The full model ran: all 24 printed means matched

    The full cat visual-cortex model, 108,150 neurons and 40.32 seconds of spontaneous activity, ran end to end on our existing 8 GB server in 15 hours. All 24 population means the paper prints in Figure 4 are matched at printed precision. This is agreement with the paper's model numbers, not biological validation.

  16. Running the full network on our own server

    The full cat visual-cortex model, 108,150 neurons and 99.4 million connections, was built and briefly run on our existing four-vCPU, 8 GB research server using swap. A full-length run projects to about 13 hours there, plus steps the test could not time; protocol 007 freezes it, and the run started on 22 October.

  17. Matching the published recording sets

    On our build of the model, the released recorder files match all 16 recorded-neuron counts in the paper's S5 Table, and the published firing-rate histograms need more neurons than our frozen plan records. The plan sampled the evoked-activity set.

  18. Rehearsing the full-model setup in a clean environment

    A clean Ubuntu 24.04 environment on our research server rebuilt the pinned simulation software in under 20 minutes and passed the upstream small-model regression test. We also found a launch requirement, a likely timing risk and a likely mismatch in the frozen recording plan.

  19. Preparing the full-model experiment

    We traced a compatibility flag through the framework's history, checked 82 parameter sets and fixed the protocol for a 40.32-second cat visual-cortex simulation. The full run remains pending.

  20. Checking the analysis on saved data

    All 48 numerical comparisons agree with the original analysis on retained small-model recordings. A timing detail explains why a nominal 10 ms bin can span 10.5 ms.

  21. Matching the recording configuration

    The archived recording grids match the general release configuration in all four cortical populations. We separate that finding from unresolved historical software details.

  22. Measuring spontaneous activity

    Twenty-four published reference values, the code behind their measurement, and a recording mismatch to resolve before the full-model run.

  23. The first cat-circuit regression

    Ten exact comparisons passed. The recorded activity, recovered dependencies and failed attempts behind our first successful small-model run.

  24. Defining the first regression run

    The exact comparison, historical dependency candidates and reporting checks for our first cat-circuit simulation.

  25. Auditing the model sources

    A complete archive inspection, missing simulation components, and a better-defined route to our first cat-circuit run.

  26. Keeping an auditable record

    How we will connect each claim to its sources, implementation, measurements and limitations.

  27. Before the first simulation

    Define the comparison before tuning the model. A proposed protocol for the first reproduction attempt.

  28. Choosing a first model

    A bounded thalamocortical question, a larger visual-cortex reference, and the files we still need to inspect.

  29. Starting with a circuit

    The first question is what we can faithfully model—and what that model would allow us to claim.