# Felisyn protocol 022 revision 2: the plaids pilot as declared and frozen (public copy)

The declared design of protocol 022 revision 2, the plaids pilot, with every file its runs and its analysis depended on. This copy is dated 28 October 2026. A plaid is two drifting gratings shown on top of each other, and the pilot asked whether the cat circuit model combines them like a relay of its responses to each grating alone. The results are in the package [plaids-022](https://www.felisyn.com/research/plaids-022/README.md) and in [journal note 22](https://www.felisyn.com/journal/two-gratings-at-once).

- **Declared** at 2026-10-26T23:15:38Z, before any plaid run: the spec, here `frozen/spec.md`, SHA-256 `8f8878fc60e5378224def8866068d3e3a1ca03ca4a48b69fa69446f5bee662d3`.
- **Frozen** at 2026-10-27T18:39:13Z by the frozen manifest, here `frozen/manifest.json`, SHA-256 `a1e2e2d92a7bfe4855da5b2f03e97674bfaa8ef1a0849c8d60faabc116c8b3d0`. It gives the size and SHA-256 of all 57 files. The launcher refused to start the runs unless every server file matched its frozen hash.
- **This public copy** holds the frozen package in `frozen/`, in its own layout, with three lines replaced because they hold private strings, and with one licence file added. Nothing else differs. The frozen package itself stays unchanged in the project's private workspace.

Simulation only. The pilot ran one network instance of a 2 mm reduction of a published model of cat primary visual cortex. It is not a biological measurement and not biological validation. No token, wallet, exchange, key or real order was involved.

## How to check this copy

1. `frozen/manifest.json` is byte-identical to the frozen manifest. Its SHA-256 is the value above, recorded at the freeze, before the runs.
2. 54 of its 57 entries match the files in `frozen/` byte for byte. They include the spec (`spec.md`), the contract (`contract.json`, SHA-256 `128b5bba…`) and the analysis code that ran (`controller/analyze_plaids.py` and `controller/plaid_stats.py`).
3. The other three files differ from their entries only by the replacements listed below. Putting the original strings back gives the frozen files.
4. One file in `frozen/` is not in the frozen manifest: `third_party/pyNN/LICENSE`, added to this copy.
5. `manifest.json`, at the top, is this copy's own manifest. It gives the size and SHA-256 of every file here, this README included, and lists the three redactions with the SHA-256 of their frozen originals.

## Redactions

| File | Line | What was replaced | SHA-256 of the frozen original (in `frozen/manifest.json`) | SHA-256 here |
| --- | --- | --- | --- | --- |
| `frozen/test_queue_guard.sh` | 33 | a private temporary folder in `JOBTMP`, replaced by `redacted-private-tmp` (the test then uses its own temporary folder, as it does wherever that folder is absent) | `c172c011e312f7ca01fbc64078ac281c1f19650fdb7b8e3a21662f09cf73c832` | `7006b86a4e8bbb3cf673360b675ce4860f3925a9f5b38f136d404e66413e321b` |
| `frozen/test_stimulus_v3.py` | 16 | the local workspace path in `WS`, replaced by `<workspace>` | `d3a2e1466555a2838758d55baa6d50a928751d42e1bc480819040ead129213b6` | `0094598c9c86b15fd56379705686ce98b49ae2f139ac1b5639c290192cbc448d` |
| `frozen/tier2/receipt.json` | 4 | the research server's host name in field `host`, replaced by `redacted-host` | `6ed21ebd154e9315a92af7ae2e63673ee285db4bbd656d11187aa56fa2ab433c` | `a287ff76040ef75e98aba5493179b892aa7746f4a53110bb0627da59b4170cca` |

- **The redacted tier 2 receipt** is byte-identical to `receipts/tier2/receipt-20261027T182426Z.json` in plaids-022.
- **Both redacted tests** ran in their original form at the freeze.
- **Private names left in place.** The spec, the frozen manifest and several scripts name private project records: workflow runs by ID (`wf_…`), and files by their paths in the private workspace (under `work/` and `docs/`, also in the frozen manifest's `source` fields). Those records are not published. The files that name them are pinned by the frozen manifest, so they are left unchanged.

## Added file

| File | Bytes | SHA-256 | Source |
| --- | --- | --- | --- |
| `frozen/third_party/pyNN/LICENSE` | 21,097 | `20ef5a0029fd269919dca199876a7e435eb24e59e0ddc942fe8ddeb4078a0868` | [PyNN at tag 0.11.0](https://raw.githubusercontent.com/NeuralEnsemble/PyNN/0.11.0/LICENSE) |

The package's three PyNN files say "CeCILL, see LICENSE", but the frozen package has no PyNN licence file. This copy adds PyNN's own, the CeCILL free software licence agreement, version 2.0. The three PyNN files here are byte-identical to the same files at PyNN's tag 0.11.0.

## A known issue in the declared spec

**One sentence of the spec is broader than its measures.** Line 17 says that each of the three primary measures is exactly 0 for a relay, "whatever the network's tuning, gain or contrast saturation". That holds only for a relay as the spec defines it: the sum of the circuit's own above-blank responses to each grating alone. Cells that saturate when both gratings drive them are not a relay, and their saturation can move the measures away from 0. The design review had flagged the sentence before the declaration, and it was not corrected. The declared outcome texts do not rest on it: the GO text says that "the shared input stage is non-linear".

## What is in `frozen/`

| Part | What it is |
| --- | --- |
| `spec.md` | The spec as declared, revision 2. |
| `contract.json` | The contract of the two runs and their analysis, built from the spec by `make_contract_022r2.py`. |
| `manifest.json` | The frozen manifest, written by `freeze_022r2.py`. |
| `sequence-r2.json` | The declared stimulus order (seed 231), built by `make_plaid_sequence_r2.py`. |
| `controller/` | The analysis: `analyze_plaids.py`, the driver that ran once on 28 October; `plaid_stats.py`, with the measures, intervals, outcome rule and fixed texts; `frozen017.py`, which finds protocol 017's frozen code. |
| `server/` | The scripts of the pilot runs: the queue, launcher and fetch scripts, stimulus version 3, the runner and its LGN input cache (unchanged from protocol 021), the tier 2 check, and V0, the structural checks of the connection lists. |
| `server_v4/` | The connection export job: stimulus version 4 (the export hook only), `export_connections.py`, and byte-identical copies of the runner and the cache. |
| `tier2/` | The tier 2 check's inputs (version 2 of the stimulus code and earlier sequences), its passing receipt (redacted) and its log. |
| `test_*.py`, `test_queue_guard.sh` | The tests. The freeze ran `test_analyze_plaids.py`, `test_check_connections_v0.py`, `test_export_connections.py`, `test_freeze_022r2.py`, `test_plaid_stats.py`, `test_queue_guard.sh`, `test_sequence_r2.py`, `test_stimulus_v3.py`, and they all passed (`manifest.json`, field `tests_at_freeze`). |
| `third_party/` | Read-only reference copies of pinned Mozaik, imagen and PyNN files, with their licences, kept to check the design against the code the server runs. Nothing there is run. |

## Credits and licences

- **Model.** Antolík J, Cagnol R, Rózsa T, Monier C, Frégnac Y, Davison AP (2024), *A comprehensive data-driven model of cat primary visual cortex*, [PLOS Computational Biology 20(8): e1012342](https://doi.org/10.1371/journal.pcbi.1012342), published under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). The runs used its code, LSV1M v1.0 (MIT license), unchanged apart from the command-line overrides of the 2 mm reduction. That code stays on the research server, and none of its files is copied here.
- **Mozaik**, the simulation framework, at commit `f5c09a0e…`: CeCILL version 2.0, `frozen/third_party/mozaik/LICENSE`.
- **PyNN 0.11.0**: CeCILL version 2.0, `frozen/third_party/pyNN/LICENSE` (added, above).
- **imagen**, the pattern library, at commit `a4af248e…`: BSD license, `frozen/third_party/imagen/LICENSE.txt`.
- **NEST 3.4**, the simulator: GNU GPL, version 2 or later. It runs on the research server, and no NEST file is here.
- **This work.** The spec, the contract, the scripts and the tests are Felisyn's own.
