Repository navigation
TPL-Lab v0.1.0
Public snapshot of a computational falsification lab for a speculative G₂/SU(3) ≅ S⁶ geometry of perception.
Result
On Allen Visual Coding Neuropixels spontaneous blocks (awake, no controlled percept): population spike-count clouds in visual cortex are not compatible with named S⁶. 12 / 12 area × session pairs refuted (two mice). Twice S⁶ won the ranking with score ~26 — the cutoff 3.0 refused. That also knocks down the stronger “sustained percept” claim on this observable.
Frozen rule
Tag prereg-v1 (0a5e794, 20 Aug 2026) is the analysis plan, locked before real spikes. This v0.1.0 tag includes results and docs; it does not change thresholds.
Reproduce
pip install -e ".[dev]"
pytest -q
python -m tpl.demo --seed 42Allen NWBs are not in git. See the README for curl + --nwb commands.
Not in this release
Raw .npy / NWB (gigabytes). IBL loader. Wake/anesthesia spike contrast. Causal non-associativity.